PS3: Unreal Tournament 2007


Recommended Posts

Ooops missed a post.

You were at Zero Hour?

Huh?

I said the Cell sucks for AI. And it does. I said it's great for physics, but that the architecture of the PS3 doesn't allow it to fully flex its muscle since according to the person I was arguing with it will be too busy doing all the geometry setup. So you're taking my comment out of the context of that conversation which isn't fair.

Also, that developer doesn't mention if they're running on an actual PS3 or a PC with a 7800GTX (which has oodles more memory bandwidth than the RSX)... and makes no mention of FSAA nor what resolution he's talking about.

The PS3 should be able to render 720p with HDR but no FSAA. Or it could do 720p with 2x FSAA but no HDR (assuming HDR means FP16 or higher). Of course it's entirely a matter of what you trade-off in order to get that.

No I wasn't but people that were there says so.

My guess is that it's running on the dev kit with a CELL at 2.4Ghz and a PC GeForce card in it.

The final dev kits ships sometime during December so unless that post was made very recently they're talking about the 2.4Ghz dev kit.

I said it's great for physics, but that the architecture of the PS3 doesn't allow it to fully flex its muscle since according to the person I was arguing with it will be too busy doing all the geometry setup. So you're taking my comment out of the context of that conversation which isn't fair.

Really? Sounds pretty straight forward that you said the Cell can't do physics.

The Cell can also not do physics or AI really at all.Branching destroys the Cell processor. Anything like physics or AI will have to be done entirely in the main CPU (the single-core 3.2Ghz Power-based chip).

Neowin

Edited by jmole

I for one would like to know how threedaysdwn know what the PS3 can and can't do seeing how it doesn't even exist yet (his own words by the way).

Thats true. PS3 doesn't exist yet. So lets wait till it actually ships or Sony publishes final hardware specs. In fact since its launching in Spring 2006, I am sure they at least have informed final specs to developers, shouldn't be far when they release specs publicly. So we can all compare in a fair manner...what say ? :yes:

On the other hand, according to udontneed2know it destroys XBOX360 in every aspect.

So why waste our time ? :D PS3 just won next gen console wars YEY !!!! :woot:

Threedays your rather pathetic these days. Backpeddle dude, backpeddle as hard as you can.

Your telling me a next generation system can't handle 720p with FSAA? Your sad, real sad if you actually believe that... trully believe that. Well you have no clue what your talking about so believing yourself isn't exactly difficult.

I bet you think just because the Xbox360 has 10MB of Embedded RAM thats why its able to have FREE AA right? lmao.

And the Cell is no worse at AI then the Xenon is, both are In-Order processors which is all that matters for AI.

But of course, you have to take into account that Cell will be able to run many different branchless tree-search algorithms that would choke and kill the Xbox360's generic processors thanks to the SPE's inherent capability to stream data in a logical data flow. There is also limited branch prediction using simple assembly language techniques you could utilize.

So in the end, once again, the Xbox360 is rather limited in an area thanks to its design compared to the PS3 which like in almost every single aspect of its creation, has built into it various different ways of performing passed a problem.

But anyway, its rather ridiculous to hear this guy even try anymore. Lair, 1080p with AA, MGS4 has AA at 1080p on Alpha Kits, UT2007 definately had AA at 720p, Heavenly Sword has AA at 1080p.

Just stop trying dude. Your not going anywhere but south in this whole thing.

Umm... where did Epic say that GoW was running at under 30FPS? Oh wait, you made that up.

God gave me this wonderful gift called Eyes. They have retinas and millions of wonderful things within that allows me to actually see things in this world. And what my eyes see, is Gears of War chugging to reach 30FPS. It will during cutscenes, like when they push flaming cars down a hill ( with pretty crappy physics I must add ) but once the action starts you can tell the game isn't running close to 30FPS and never has. Of course the game will reach 30FPs eventually, when they take away enough to compensate for the 360's inadequecies but it'll take awhile. Hell Sega of all masterminds still has Full Auto chugging along at the moment in a rather sad fashion. But then you have UT2007 running at a crisp 49FPS on alpha kits with maybe 6 months of development time behind it and many many many many months to go on final development kits to further optimize the code. And Epic flat out said it will look much better when the games released.

2) You're comparing the Cell's non-IEEE compliant single-precision number to the Xenon's IEEE compliant double-precision number. For the PS3, the DP number is a measely 20 GFLOPs or so (estimated as IBM reports the cell with 8 SPEs at 4Ghz does 26 GFLOPs)

3) As for the RSX, the Geforce FX 5800 had more computational power than a Radeon 9800 Pro. Which was more useful in games?

You know, you have the gall to call me out for BS ( when its not ) but then your spout on about this horrid mess of ignorance? I probably just lost about 50 IQ just trying to reach the level of intelligence it would take to read this drivel. Did you really, seriously, just compare a Geforce FX 5800 to a Radeon 9800 Pro? Seriously? ...... I'll just laugh and shake my head at that one.

And dude ... come on, if you really think those 1Tflop numbers out of Microsoft are Double precision you need your head examined with a quickness. Seriously.

You know what, nevermind. Not even worth it. I'll just continue shaking my head and put you on ignore or something lol. Your pretty much wasting mine and everyone else time.

I'll just leave you with some words from someone who does know a thing or two about game development, since I obviously don't have a clue : laugh :

es, the game is targeted to run at 1080x720, a.k.a. 720p resolution if you have a high-definition display. We asked Epic Games vice president Mark Rein, who was also present at the demo, about the target frame rate. “It’s simply too early to say, but we’re aiming for it to be as high as possible,” he responded. For the record, the 720p, totally un-optimized build running on hardware less powerful than the final PS3 spec was clipping along at 49fps.

Totally un-optimized game code running on hardware less powerful then the final PS3 specs clipping along at 49FPS @ 720p. Now go to bed kid.

Edited by udontneed2know

fsaa was created to reduce jagged edges. when you're running at 1080p i don't think it's necessary at all. at that point you should instead focus on hdr, physics, ai, etc.

anyways don't speculate on unreleased hardware. didn't people think it was impossible for the ps2 to do 1080i?

Edited by psyko_x

No need to get angry at each other. Please, continue the debate. I've learned alot and I wish to learn more. Feed my brain... diminish its hunger for tech knowledge... :pinch:

Really? Sounds pretty straight forward that you said the Cell can't do physics.

Again you took my comment out of context. I said if the RSX is doing all the geometry work, then the Cell can't do physics (in the way that the Xenon can because the Xenos can MEMEXPORT the vertex data to the UMA memory).

If the Cell is doing all the geometry work and all the physics work, then you're wasting the GPU and splitting the Cell's resources among the two tasks. I was asking "undontneed2know" what work he expected the Cell to do and what work he expected the GPU to do - and I never got a clear answer.

I think the answer is - for most games, the GPU will do most of the work just like in a current PC. A few developers (those with the most time and that are closest to Sony) will try to split work between the GPU and the RSX. And it will cost them hugely during development and possibly not be worth the effort.

It's all speculation. I never said it wasn't. But if you're going to speculate that the PS3 will be faster because Sony has bigger numbers - I'm going to speculate that the 360 is faster because the architecture makes more sense.

What's more, if PS3 games end up looking largely just like 360 games, I think people like "udontneed2know" are going to be very disappointed. They have unrealistic expectations that the PS2... err, PS3 will be outputting "Pixar-like" graphics or some such.

Three, your acting like FSAA and high resolutions are something extraordinary that require extreme amounts of power just to accomplish. I've been playing games with FSAA x 4 at 1600 x 1200 on my PC since the early days of the Radeon 9200 Pro. Your whole argument is just flat out stupid I'm sorry to say.

The Cell processor can do a number of things, including what the Heavenly Sword developers are doing which is have Cell handle the AA process after the graphics are drawn on the screen. Simple, resourceful, and something Cell could do at FSAA x 16 with only the use of a few SPE's.

The fact of the matter is that the Cell processor is a programmable GPU. You can pretty much throw out the fact that Cell is a basic PC CPU whos only job is to throw instructions at the GPU given to it and control the operations of an OS. The Cell is an incredibly capable GPU. So given that fact, and the fact that from the ground up the Cell processor is a highly efficient, incredibly flexible architecture that gives the programmer absolute FULL control over every aspect of the architecture. From the FlexIO being customizable to the EIB being customizable to the SPE's being customizable to the PPE being customizable to the SPE's having direct control over the DMA memory managers and full access to the System RAM when avaliable. The programmer finally has full control over the machine they are speaking to, and the fact that Assembly is highly utilized throughout the Cell language that gives programmers the ability to talk directly to the system as well. So there are many dozens of ways to figure out a way around any problem that will ever arise in the programming process of PS3 development.

I'd also like to add that the RSX will be programmable as well in a limited fashion through the multi-way programmable shader architecture. Which adds a level of flexibility as well for the GPU.

Anti Aliasing could be done through the GPU or through Cell. Geometry can be done through the GPU or Cell ( or share the process, possibly every model being pushed throubh by the RSX while Cell handles various special effects, lighting, Anti-Aliasing, morphing, mapping, etc. ). Physics will be handled by the SPE's but either in the bulk tradition or in a streamed processing fashion.

Many ways to tackle every problem within the PS3 and the Cell has more then enough Cache to handle all these different processes and instruction layouts without a hitch.

What's more, if PS3 games end up looking largely just like 360 games, I think people like "udontneed2know" are going to be very disappointed. They have unrealistic expectations that the PS2... err, PS3 will be outputting "Pixar-like" graphics or some such.

Have you listened to a single word from PS3 developers? Lair is performing miracles that WERE only possible before in movie studio CGI workstations. 10 layers of textures, 170,000 triangles per model at 30FPS @ 1080p. Heavenly Sword was real-time with full AI implemented in the video shown and that was done SINGLE THREADED. The playable Gundam World at Sony Summit was done THROUGH ONLY THE PPE AND 1 SPE and the models looked exactly like those of the CG presentation shown at E3. MGS4's models and HDR lighting ability is right up there with anything done in the movies, and considering the Cell processor is actually capable of real-time raytracing algorithms you could perhaps see that implemented in less intense games like Fighting titles. Who knows. My expectations are high because I see a brand new architecture in the PS3 that gives programmers the ability to not talk to the processor, but better Sing to the processor and make it do whatever they need it to do.

I'd also like to add that since the PS3 will use the Linux OS in its HDD's we'll see many thousands of various tools built by independant programmers and developers that PS3 developers could simply go online and download to use on their own. The biggest tool though thats come out recently from IBM and Sony are those SPE Optimizers. That tool allows the developer to not really worry too much about programming into the script what exactly the SPE's will need to do and how exactly they will do it. That extremely useful tool will do that for the developer on its own. Like for instance if the programmer has a heavy multi-threaded neural net AI algorithm within the program the SPE Optimizer will rearrange the LS Storage within a few of the SPE's to give the ones handling the branchless algorithm some more storage for the purpose. Stuff like that is automated now thanks to that tool. And well, for Cell the Compiler is the most important tool of all and up until November developers had to hand write their own compilers for Cell which was probably the most difficult thing of all. But with SN Systems professional compiler that is thrown out the window and programmers can go buy some wigs for all that missing hair they pulled out.

Edited by udontneed2know

I just want to add to "udontneedtoknow"'s excellent post by saying that Sony had originally intended to have a second Cell processor as the GPU (but decided to use the RSX for price and avialalibility reasons).

It's partly because of the Cell's flexibility that its difficult to program. If you have a fixed paradigm, its easier to do stuff, but what you come up with is usually less efficient and less as you wanted it. The Cell lets programmers truley make the game in the designers' heads.

Three, your acting like FSAA and high resolutions are something extraordinary that require extreme amounts of power just to accomplish. I've been playing games with FSAA x 4 at 1600 x 1200 on my PC since the early days of the Radeon 9200 Pro. Your whole argument is just flat out stupid I'm sorry to say.

Now you're just plain lying.

The Cell processor can do a number of things, including what the Heavenly Sword developers are doing which is have Cell handle the AA process after the graphics are drawn on the screen. Simple, resourceful, and something Cell could do at FSAA x 16 with only the use of a few SPE's.

Bull****.

The fact of the matter is that the Cell processor is a programmable GPU. You can pretty much throw out the fact that Cell is a basic PC CPU whos only job is to throw instructions at the GPU given to it and control the operations of an OS. The Cell is an incredibly capable GPU. So given that fact, and the fact that from the ground up the Cell processor is a highly efficient, incredibly flexible architecture that gives the programmer absolute FULL control over every aspect of the architecture. From the FlexIO being customizable to the EIB being customizable to the SPE's being customizable to the PPE being customizable to the SPE's having direct control over the DMA memory managers and full access to the System RAM when avaliable. The programmer finally has full control over the machine they are speaking to, and the fact that Assembly is highly utilized throughout the Cell language that gives programmers the ability to talk directly to the system as well. So there are many dozens of ways to figure out a way around any problem that will ever arise in the programming process of PS3 development.

Come back when you have even the slightest grasp of the technology you're talking about.

I'd also like to add that the RSX will be programmable as well in a limited fashion through the multi-way programmable shader architecture. Which adds a level of flexibility as well for the GPU.

Umm... you mean just like every other GPU that supports SM 3.0?

Anti Aliasing could be done through the GPU or through Cell. Geometry can be done through the GPU or Cell ( or share the process, possibly every model being pushed throubh by the RSX while Cell handles various special effects, lighting, Anti-Aliasing, morphing, mapping, etc. ). Physics will be handled by the SPE's but either in the bulk tradition or in a streamed processing fashion.

Do you actually believe these words mean something or do you just assume that non-technical readers will gloss over this and agree with you because they like what you're saying?

Have you listened to a single word from PS3 developers? Lair is performing miracles that WERE only possible before in movie studio CGI workstations. 10 layers of textures, 170,000 triangles per model at 30FPS @ 1080p. Heavenly Sword was real-time with full AI implemented in the video shown and that was done SINGLE THREADED. The playable Gundam World at Sony Summit was done THROUGH ONLY THE PPE AND 1 SPE and the models looked exactly like those of the CG presentation shown at E3. MGS4's models and HDR lighting ability is right up there with anything done in the movies, and considering the Cell processor is actually capable of real-time raytracing algorithms you could perhaps see that implemented in less intense games like Fighting titles. Who knows. My expectations are high because I see a brand new architecture in the PS3 that gives programmers the ability to not talk to the processor, but better Sing to the processor and make it do whatever they need it to do.

This is what I mean. You're deluisional if you believe any of that. Now the Cell does AA and raytracing? What next, it will make you breakfast?

I'd also like to add that since the PS3 will use the Linux OS in its HDD's we'll see many thousands of various tools built by independant programmers and developers that PS3 developers could simply go online and download to use on their own.

You mean just like happened with the PS2? :laugh:

The biggest tool though thats come out recently from IBM and Sony are those SPE Optimizers. That tool allows the developer to not really worry too much about programming into the script what exactly the SPE's will need to do and how exactly they will do it. That extremely useful tool will do that for the developer on its own. Like for instance if the programmer has a heavy multi-threaded neural net AI algorithm within the program the SPE Optimizer will rearrange the LS Storage within a few of the SPE's to give the ones handling the branchless algorithm some more storage for the purpose. Stuff like that is automated now thanks to that tool. And well, for Cell the Compiler is the most important tool of all and up until November developers had to hand write their own compilers for Cell which was probably the most difficult thing of all. But with SN Systems professional compiler that is thrown out the window and programmers can go buy some wigs for all that missing hair they pulled out.

Dude that was the most hilarious thing I've read all day. You should write a script for Star Trek or something.

"Sir, we've got a problem with the SPE optimizers. They've overloaded the LS Storage and crippled the multi-threaded neurel nets. I don't know how much more of this she can take!"

I think udn2k has a good library of keywords and/or of sentences which he mixes & matches to post on this forum...not that I know/understand any/all of it...but some of his posts sure don't make any sense. Including Sony's "special tricks" for cell.

Do you actually believe these words mean something or do you just assume that non-technical readers will gloss over this and agree with you because they like what you're saying?

there, you nailed it. I am probably one of those. But still can't see how developers find pure assembly based coding easier than xbox360 !

Three, your acting like FSAA and high resolutions are something extraordinary that require extreme amounts of power just to accomplish. I've been playing games with FSAA x 4 at 1600 x 1200 on my PC since the early days of the Radeon 9200 Pro. Your whole argument is just flat out stupid I'm sorry to say.

The Cell processor can do a number of things, including what the Heavenly Sword developers are doing which is have Cell handle the AA process after the graphics are drawn on the screen. Simple, resourceful, and something Cell could do at FSAA x 16 with only the use of a few SPE's.

got a link to cell doing fsaa16?

The fact of the matter is that the Cell processor is a programmable GPU. You can pretty much throw out the fact that Cell is a basic PC CPU whos only job is to throw instructions at the GPU given to it and control the operations of an OS. The Cell is an incredibly capable GPU. So given that fact, and the fact that from the ground up the Cell processor is a highly efficient, incredibly flexible architecture that gives the programmer absolute FULL control over every aspect of the architecture. From the FlexIO being customizable to the EIB being customizable to the SPE's being customizable to the PPE being customizable to the SPE's having direct control over the DMA memory managers and full access to the System RAM when avaliable. The programmer finally has full control over the machine they are speaking to, and the fact that Assembly is highly utilized throughout the Cell language that gives programmers the ability to talk directly to the system as well. So there are many dozens of ways to figure out a way around any problem that will ever arise in the programming process of PS3 development.

fact is cell a powerful enough gpu as ati or nvidia, thus it was scrapped only around a year ago so that Nvidia could make RSX for it. Cell is not a basic CPU, theres a difference between Inorder and Outoforder CPUs. Cell is not an incredible GPU. its a CPU similiar to FireGL cards which are bad for gaming but good for 3d imaging. Cell processor is not really highly efficient, if it were, it would have branch predictions, which it doesnt, making it inefficient compared to Intel and AMD processors. FlexIO has nothing to do with Cell, its just a connection between RSX and Cell. SPEs dont have branch predictors which makes it inefficient compared to Intel and AMD processors in terms of game programming, but if you say SPEs are only used for graphics, that means the games will be chuggy. Actually compared to Xbox 360 the code layering is pretty thick as described by alot of japanese developers and john carmack and john nevell, who have said the gfx and programming layers are very thin so its easier to program for the xbox 360 even if the PPC main core is the same as the cell processor.

I'd also like to add that the RSX will be programmable as well in a limited fashion through the multi-way programmable shader architecture. Which adds a level of flexibility as well for the GPU.

Anti Aliasing could be done through the GPU or through Cell. Geometry can be done through the GPU or Cell ( or share the process, possibly every model being pushed throubh by the RSX while Cell handles various special effects, lighting, Anti-Aliasing, morphing, mapping, etc. ). Physics will be handled by the SPE's but either in the bulk tradition or in a streamed processing fashion.

mind you rsx cannot do hdr and anti aliasing at the same time, it doesnt have the capabilities, thats why the developers for heavenly sword have said instead of using Fp16 which is minimum standard for good graphics, they are using Int8, whciih is ok but not next gen

Many ways to tackle every problem within the PS3 and the Cell has more then enough Cache to handle all these different processes and instruction layouts without a hitch.

Have you listened to a single word from PS3 developers? Lair is performing miracles that WERE only possible before in movie studio CGI workstations. 10 layers of textures, 170,000 triangles per model at 30FPS @ 1080p. Heavenly Sword was real-time with full AI implemented in the video shown and that was done SINGLE THREADED. The playable Gundam World at Sony Summit was done THROUGH ONLY THE PPE AND 1 SPE and the models looked exactly like those of the CG presentation shown at E3. MGS4's models and HDR lighting ability is right up there with anything done in the movies, and considering the Cell processor is actually capable of real-time raytracing algorithms you could perhaps see that implemented in less intense games like Fighting titles. Who knows. My expectations are high because I see a brand new architecture in the PS3 that gives programmers the ability to not talk to the processor, but better Sing to the processor and make it do whatever they need it to do.

170,000 Triangles is nothing for a setup rate model when it was claimed the setup rate is at 500 million triangles per second. Heavenly sword was forced to reduce from FP16 to Int8 according to its developer in beyond3d forums. Gundam looked on par with FEAR with no physics. if you look at MGS4 and look at entire screenshots, the background texture resolution was really low and blurry and all concentration had been given to Snake's model, so in overall sense, it wasnt impressive, it would have been if backgrounds were highly textured as well, compare that to gears of war where everything is detailed from walls to characters to the ground. Cell is not capable of real-time ray Tracing, Xenos, the xbox 360 processer in a slideshow by ATI said it is, Sony has not talked about ray tracing at all in cell or rsx. I think ur expectations are too high but I will sympathize with you.

I'd also like to add that since the PS3 will use the Linux OS in its HDD's we'll see many thousands of various tools built by independant programmers and developers that PS3 developers could simply go online and download to use on their own. The biggest tool though thats come out recently from IBM and Sony are those SPE Optimizers. That tool allows the developer to not really worry too much about programming into the script what exactly the SPE's will need to do and how exactly they will do it. That extremely useful tool will do that for the developer on its own. Like for instance if the programmer has a heavy multi-threaded neural net AI algorithm within the program the SPE Optimizer will rearrange the LS Storage within a few of the SPE's to give the ones handling the branchless algorithm some more storage for the purpose. Stuff like that is automated now thanks to that tool. And well, for Cell the Compiler is the most important tool of all and up until November developers had to hand write their own compilers for Cell which was probably the most difficult thing of all. But with SN Systems professional compiler that is thrown out the window and programmers can go buy some wigs for all that missing hair they pulled out.

in bold

Wow Three, that was a staggering counter-argument. You provide nothing but bland words of " liar " without actually saying anything.

Pretty typical if you ask me.

If you think 170,000 triangles per model isn't impressive ( especially with 100 or so of those models flying around on the screen at once ) then you should never again speak of Gears of War which is only around 20,000 triangles per model.

fact is cell a powerful enough gpu as ati or nvidia, thus it was scrapped only around a year ago so that Nvidia could make RSX for it. Cell is not a basic CPU, theres a difference between Inorder and Outoforder CPUs. Cell is not an incredible GPU. its a CPU similiar to FireGL cards which are bad for gaming but good for 3d imaging. Cell processor is not really highly efficient, if it were, it would have branch predictions, which it doesnt, making it inefficient compared to Intel and AMD processors. FlexIO has nothing to do with Cell, its just a connection between RSX and Cell. SPEs dont have branch predictors which makes it inefficient compared to Intel and AMD processors in terms of game programming, but if you say SPEs are only used for graphics, that means the games will be chuggy. Actually compared to Xbox 360 the code layering is pretty thick as described by alot of japanese developers and john carmack and john nevell, who have said the gfx and programming layers are very thin so its easier to program for the xbox 360 even if the PPC main core is the same as the cell processor.

Scrapped a year ago? Try way back in early 2002 when Nvidia was called in. Not because a dual Cell framework wouldn't have been ideal, but simply because it would have been far too difficult on game developers to rework their entire programming thought process for graphics. Going with a more known API from Nvidia was the ideal solution so Nvidia came in and along with Sony built a custom GPU that will work hand in foot with the Cell processor. You guys should try reading something every now and again, this stuff has been known for ages.

The graphic capabilities of Cell are pretty extreme in various programming methods. Such as RAYES. REYES basically subdivides down to micropolygons and then computes vertex intensities from neighboring vertex colors. You could adapt a depth-first subdivision algorithm to the SPEs. This would basically bypass the traditional triangle fragement generation, processing backend used in most GPUs today and leverage the vertex processing capabilities of the Cell SPEs. More heavily utilized Vertex programming at the geometrical level will also be another strong point and is only slowed down if the SPU-GPU bandwidth was limited, which it isn't. Utilizing SPUFballs or small applications at the matrix 32-bit level ought to be easy as well. Software rendering, something that has been wiped out due to todays GPU's is back into the picture thanks to the Playstation 3 and Cells capabilities. Software rendering could bring about some exciting changes in the way programmers work their magic.

And Cell can't do ray-tracing? You guys are about 6 months behind the times. Heres an IBM engineer / programmer to explain it to you.

Sheepie I agree with you that many folks over hype the power of GPUs and under estimate the graphics potential of Cell.

I was the lead developer on a team of three that wrote the IBM Cell terrain ray-caster that you mentioned. While Daniel Cohen did develop ray-casters on IBM PVS HW back in the 90s our effort was completely independent from his work.

Our work has shown that one 3.2 GHz Cell processor can ray-cast and compress 1280x720 super-sampled images at frame rates better than 30 frames/sec without the assist of a GPU. We start with a 16 bit (5/6/5) texture and a 16 bit digital elevation map and render into a 128 bit (32 bit float/channel) frame buffer. The atmosphere is procedurally generated on the fly with multiple octaves of Perlin noise and the finished images are JPEG "like" compressed and delivered by Cell all in real-time.

On the topic of Ray-tracing, Philipp Slusallek already has a full blown ray-tracer running on Cell and showed an early version of it at Siggraph 05. More interesting stuff on the way....

mind you rsx cannot do hdr and anti aliasing at the same time, it doesnt have the capabilities, thats why the developers for heavenly sword have said instead of using Fp16 which is minimum standard for good graphics, they are using Int8, whciih is ok but not next gen

Wheres your proof of this? WHERE?! lol. Fact is the only proof we have right now is games pushing AA and HDR with no problems whatsoever. I would just like some proof, a shred atleast. But you'll find none because there is none. Saying this architecture is incapable of two standard things like HDR and AA is ridiculous at best and shows how little you actually know about how powerful these new machines really are.

Heres some more Ray-Tracing on Cell stuff for you.

Recently I came across a link on www.gpgpu.org that I found interesting. It described a method of ray-tracing quaternion Julia fractals using the floating point power in graphics processing units (GPUs). The author of the GPU code , Keenan Crane, stated that “This kind of algorithm is pretty much ideal for the GPU - extremely high arithmetic intensity and almost zero bandwidth usage”. I thought it would be interesting to port this Nvidia CG code to the Cell processor, using the public SDK, and see how it performs given that it was ideal for a GPU. First we directly translated the CG code line for line to C + SPE intrinsics. All the CG code structures and data types were maintained. Then we wrote a CG framework to execute this shader for Cell that included a backend image compression and network delivery layer for the finished images. To our surprise, well not really, we found that using only 7 SPEs for rendering a 3.2 GHz Cell chip could out run an Nvidia 7800 GT OC card at this task by about 30%. We reserved one SPE for the image compression and delivery task. Furthermore the way CG structures it SIMD computation is inefficient as it causes large percentages of the code to execute in scalar mode. This is due to the way they structure their vector data, AOS vs SOA. By converting this CG shader from AOS to SOA form, SIMD utilization was much higher which resulted in Cell out performing the Nvidia 7800 by a factor of 5 - 6x using only 7 SPEs for rendering. Given that the Nvidia 7800 GT is listed as having 313 GFLOPs of computational power and seven 3.2 GHz SPEs only have 179.2 GFLOPs this seems impossible but then again maybe we should start reading more white papers and less marketing hype.

RSX with a 128-bit Z Buffer will be a very nice little companion for Cell, and Cell will be a very nice companion for RSX. This crap you guys are spouting on about, unable to do HDR and AA is laughable.

Software rendering is even making 360 programmers excited since no machine has ever been fast enough to accomplish it up until the PS3.

hat said, i'm also quite certain (as stated previously) that a GPU will beat the Cell hands-down at GPU-style activities. I'm not saying that the Cell is a slow processor, or that the GPU is better technology, only pointing out that a GPU is custom-built hardware with a specific purpose. At that purpose, it will beat a Cell. That is all I am saying. That's an opinion, and everyone's welcome to have their own.

Getting beyond that, I am asking: If we ignore the GPU completely for a bit, what kind of rendering can we do on the Cell? It's been a long time since we've had a machine fast enough to even consider software rendering. There are numerous techniques that were thrown away when the industry moved to hardware rendering. Perhaps it's time to dust some of those off?

SPE's do have branch prediction, only limited. Heres IBM on the subject.

A few remarks:

- It is an easy misperception to think of an SPE as a SIMD(vector) only processor. IT IS NOT. Just because we have unified the register file does not mean you have to use SIMD. ( An x86 with SSE or a 970 with VMX is still perfectly ok to use for scalar code and so is the SPE). Compilers support scalar code just fine.

- This SPE implementation has only trivial hardware branch prediction (because large branch prediction structures do not provide a good return on investment in terms of area and power ... better to have more SPEs). Still we do care a lot about idle cycles due to branches, so there is a nice branch hint instruction (architecturally a no-op) that notifies the instruction fetch unit that at address x a branch will end up at address y. Also we support the select instruction that often allows one to eliminate a branch altogether.

- With respect to AI, I don't have the answers, but I have good hope. I think AI is usually not bound by computation but by memory access penalties (on high-frequency processors). I can imagine tree search algorithms for the SPEs that absolutely rock, by getting a lot of memory accesses in flight concurrently. There may be a patent out there by M. Necker and myself that describes some of this for the case of routing table accesses ( also a kind of tree search ).

Anything else?

Edited by udontneed2know

Wow Three, that was a staggering counter-argument. You provide nothing but bland words of " liar " without actually saying anything.

Pretty typical if you ask me.

If you think 170,000 triangles per model isn't impressive ( especially with 100 or so of those models flying around on the screen at once ) then you should never again speak of Gears of War which is only around 20,000 triangles per model.

Scrapped a year ago? Try way back in early 2002 when Nvidia was called in. Not because a dual Cell framework wouldn't have been ideal, but simply because it would have been far too difficult on game developers to rework their entire programming thought process for graphics. Going with a more known API from Nvidia was the ideal solution so Nvidia came in and along with Sony built a custom GPU that will work hand in foot with the Cell processor. You guys should try reading something every now and again, this stuff has been known for ages.

The graphic capabilities of Cell are pretty extreme in various programming methods. Such as RAYES. REYES basically subdivides down to micropolygons and then computes vertex intensities from neighboring vertex colors. You could adapt a depth-first subdivision algorithm to the SPEs. This would basically bypass the traditional triangle fragement generation, processing backend used in most GPUs today and leverage the vertex processing capabilities of the Cell SPEs. More heavily utilized Vertex programming at the geometrical level will also be another strong point and is only slowed down if the SPU-GPU bandwidth was limited, which it isn't. Utilizing SPUFballs or small applications at the matrix 32-bit level ought to be easy as well. Software rendering, something that has been wiped out due to todays GPU's is back into the picture thanks to the Playstation 3 and Cells capabilities. Software rendering could bring about some exciting changes in the way programmers work their magic.

And Cell can't do ray-tracing? You guys are about 6 months behind the times. Heres an IBM engineer / programmer to explain it to you.

Wheres your proof of this? WHERE?! lol. Fact is the only proof we have right now is games pushing AA and HDR with no problems whatsoever. I would just like some proof, a shred atleast. But you'll find none because there is none. Saying this architecture is incapable of two standard things like HDR and AA is ridiculous at best and shows how little you actually know about how powerful these new machines really are.

Heres some more Ray-Tracing on Cell stuff for you.

RSX with a 128-bit Z Buffer will be a very nice little companion for Cell, and Cell will be a very nice companion for RSX. This crap you guys are spouting on about, unable to do HDR and AA is laughable.

Software rendering is even making 360 programmers excited since no machine has ever been fast enough to accomplish it up until the PS3.

SPE's do have branch prediction, only limited. Heres IBM on the subject.

Anything else?

1) Nvidia started on 7800 technology in december 2003, if it was before 2003 then they would have gone the ATI route and made a totally deperate Console specific technology which is not the case from given RSX specs

2) Sony has yet to state that Ray Tracing can be done while ATI has released slides saying real time 64 bit ray tracing can be done according to developer wishes. You dont seem to know Nvidia if you still think its G70 technology can do FP8 or FP16 HDR without any kind of AA, if they try 2xAA they need to reduce textures, whcih you dont need to do with ATI cards since 9800 series, if you want proof

http://www.digit-life.com/articles2/video/g70-part3.html look at the Far Cry bench and Bold statement about HDR and AA

3) RSX is as much a companion as the Nvidia card was for the P3 733 Mhz chip in Xbox

4) What is accomplished in PS3 is only physics and CGI graphics vs the Xbox CPU but when you factor in the fact that the ATI card in Xbox 360 is a GPGPU, which Nvidia cant claim because thier G70 isnt, and that it can do real time pre and post fetching from any core of the 3 in CPU as slaves to render and process CGI and ingame graphics, you have a system which in real time performance by around 2007 will make Xbox 360 games look better than PS3 games even if the physics of PS3 will get better than Xbox 360 games.

anything else.

Wow Three, that was a staggering counter-argument. You provide nothing but bland words of " liar " without actually saying anything.

Pretty typical if you ask me.

If you think 170,000 triangles per model isn't impressive ( especially with 100 or so of those models flying around on the screen at once ) then you should never again speak of Gears of War which is only around 20,000 triangles per model.

Scrapped a year ago? Try way back in early 2002 when Nvidia was called in. Not because a dual Cell framework wouldn't have been ideal, but simply because it would have been far too difficult on game developers to rework their entire programming thought process for graphics. Going with a more known API from Nvidia was the ideal solution so Nvidia came in and along with Sony built a custom GPU that will work hand in foot with the Cell processor. You guys should try reading something every now and again, this stuff has been known for ages.

The graphic capabilities of Cell are pretty extreme in various programming methods. Such as RAYES. REYES basically subdivides down to micropolygons and then computes vertex intensities from neighboring vertex colors. You could adapt a depth-first subdivision algorithm to the SPEs. This would basically bypass the traditional triangle fragement generation, processing backend used in most GPUs today and leverage the vertex processing capabilities of the Cell SPEs. More heavily utilized Vertex programming at the geometrical level will also be another strong point and is only slowed down if the SPU-GPU bandwidth was limited, which it isn't. Utilizing SPUFballs or small applications at the matrix 32-bit level ought to be easy as well. Software rendering, something that has been wiped out due to todays GPU's is back into the picture thanks to the Playstation 3 and Cells capabilities. Software rendering could bring about some exciting changes in the way programmers work their magic.

And Cell can't do ray-tracing? You guys are about 6 months behind the times. Heres an IBM engineer / programmer to explain it to you.

Wheres your proof of this? WHERE?! lol. Fact is the only proof we have right now is games pushing AA and HDR with no problems whatsoever. I would just like some proof, a shred atleast. But you'll find none because there is none. Saying this architecture is incapable of two standard things like HDR and AA is ridiculous at best and shows how little you actually know about how powerful these new machines really are.

Heres some more Ray-Tracing on Cell stuff for you.

RSX with a 128-bit Z Buffer will be a very nice little companion for Cell, and Cell will be a very nice companion for RSX. This crap you guys are spouting on about, unable to do HDR and AA is laughable.

Software rendering is even making 360 programmers excited since no machine has ever been fast enough to accomplish it up until the PS3.

SPE's do have branch prediction, only limited. Heres IBM on the subject.

Anything else?

GOW has 20,000 on each character because the focus is on the environment, which each element caries 10-100,000 setup rate triangles per second, not just the character. just like MGS4, get any pic of him leaning against the wall and you will know, compare the wall and compare him, stagerring difference

926596_20050916_screen005.jpg

compare the character of Snake with the Wall behind him. notice the difference in quality

you know your comparing a game thats been in development for about 5 months compared to a game thats been in development for over a year right? Whats the first thing you do when you create an engine? Thast right, you get the models up and going. Then you work on the environmental models, then you work on your AI and Physics routines ( physics wise especially the MGS4 video absolutely annihilates anything seen in anything GoW related ). There isn't necessarily a different in quality just a difference in necessity. They wanted to showcase the PS3's ability to do self-shadowing and HDR lighting and all that good stuff on the models, which is something 10x more complex then a static building. Those textures will be ramped up considerably as the game actually reaches something further then 10% completion time.

If you want to compare GoW to something just head on over to Unreal Tournament 2007. Same engine as GoW, same resolutions, same effects, same AA abilities. It just runs a hell of alot smoother on the PS3 utilizing unoptimized code.

) Nvidia started on 7800 technology in december 2003, if it was before 2003 then they would have gone the ATI route and made a totally deperate Console specific technology which is not the case from given RSX specs

Your debating publicly known information dude. The RSX has been in co-development since early 2002. Go read something for the love of GOD lol. They announced back in 2004 that they have been working together for 2 years on the GPU for the Playstation 3. The G70 will have very little similiarities when its all said and done with the RSX.

2) Sony has yet to state that Ray Tracing can be done while ATI has released slides saying real time 64 bit ray tracing can be done according to developer wishes. You dont seem to know Nvidia if you still think its G70 technology can do FP8 or FP16 HDR without any kind of AA, if they try 2xAA they need to reduce textures, whcih you dont need to do with ATI cards since 9800 series, if you want proof

You know 2 other major corporations worked on the Cell processor besides Sony. I just provided you not one but two quotes about Cell and Ray-Tracing. READ MAN lol. And you keep referring back to Nvidias passed projects. The RSX is a new thing man, learn to live with that.

To end, some of the most exciting possibilities concerning the RSX don't come from analysis of the G70 itself at all, but from talk of certain changes the 90nm revision of G70 is rumored to feature. In addition to theorizing that the 90nm revision of the chip will sport a full eight pixel quads (32 pixel pipes) and 10 vertex shaders, Josh Walrath at Penstar Systems indicates that NVidia may be reworking the anti-aliasing unit in the 90nm revision in order to allow for multi-sample anti-aliasing and HDR lighting to be implemented concurrently. There is also talk that texture filtering may be raised back up to pre-NV40 series levels. If true, we might expect any or all of these architectural improvements to find their way into the RSX - something that all those anticipating the PS3 should find fairly exciting.
Mr. Roman informed them that NVIDIA has been working on aspects of Sony Computer Entertainment Inc.'s next generation system for the past 2 years and that the GPU for the PS3 will be a custom version of our next generation GPU.

Nvidia flat out said this in plain English at E3. The RSX is a custom built GPU that only has rather basic architectural similarities to the G70 series. Everything else will be custom, EVERYTHING ELSE. So you can pretty much throw all your basic arguments out the window.

) RSX is as much a companion as the Nvidia card was for the P3 733 Mhz chip in Xbox

This coming from the guy who actually thought Nvidia had been working on the RSX for less then a year. Come on, basic common sense. RSX was built from the ground up to be just that, a companion GPU for the Cell architecture.

4) What is accomplished in PS3 is only physics and CGI graphics vs the Xbox CPU but when you factor in the fact that the ATI card in Xbox 360 is a GPGPU, which Nvidia cant claim because thier G70 isnt, and that it can do real time pre and post fetching from any core of the 3 in CPU as slaves to render and process CGI and ingame graphics, you have a system which in real time performance by around 2007 will make Xbox 360 games look better than PS3 games even if the physics of PS3 will get better than Xbox 360 games.

Good lord. lol. And what games are being showcased now or in the future even show that the Xbox360 is a real competitor to current PC technology? Hmmm .. absolutely nothing. Whereas the PS3 games being showcased on unfinished hardware with relatively little development time are destroying anything shown in todays world in physics, graphics, whatever. I'll be sure to quote you on that though when 2007 comes around and the 360 is still trying to catch up to the PS3 launch titles lol.

Oh, also another important note is that nothing shown for the PS3 has of yet even used the RSX for crying out loud. Its all been done with either the dual Geforce 6800 setup avaliable in the Beta kits or the Geforce 7800GTX within the Alpha kits. The RSX will be utilized in the final development kits. You'll see a rather dramatic lift in performance just from that simple fact.

Oh and if the 360 is supposed to be capable of so much with this free AA and HDR lighting and all this other BS your talking about, why exactly have absolutely zero games utilized these features and why are so many games chugging along. Quake 4 was ported over to the 360 and if your talk is correct then well, something went HORRIBLY wrong. That game absolutely CHUGS on the 360, this supposed super graphics machine of yours. And this coming from ID, supposely from what you Microsoft people think, who are graphic masterminds.

As Sony put it so well recently in a PSM interview, " the proof is in the pudding ". The proof of what a system is capable of is announced loudly by what is actually showcased onthat system. From what we have seen from all launch titles and upcoming 360 games, the 360 isn't capable of anything that isn't easily matched on current PC's. And with the launch titles, Quake 4 actually handles one hell of alot worse on the 360 compared to its PC counterparts. The proof is in the pudding, indeed.

But wait you say, Quake 4 may have been a rush job. Well, UT2007 on the PS3 is a rush job too with unoptimized code running at 720p @49FPS on unfinished development kits not even using the specialized GPU for the PS3 with atleast 7 more months of pure development time to come. The proof .. is in .. the pudding ...

Oh, and to kill off another one of your false claims.

http://www.behardware.com/art/imprimer/599/

Please proceed to the bottom of the page listing the capabilities of each processor. And then please keep your eyes firmly within the HDR FP16 section of the table showcasing that even the 7800 GT and GTX are capable of this feature.

And now please, shut up about it.

Edited by udontneed2know

you know your comparing a game thats been in development for about 5 months compared to a game thats been in development for over a year right? Whats the first thing you do when you create an engine? Thast right, you get the models up and going. Then you work on the environmental models, then you work on your AI and Physics routines ( physics wise especially the MGS4 video absolutely annihilates anything seen in anything GoW related ). There isn't necessarily a different in quality just a difference in necessity. They wanted to showcase the PS3's ability to do self-shadowing and HDR lighting and all that good stuff on the models, which is something 10x more complex then a static building. Those textures will be ramped up considerably as the game actually reaches something further then 10% completion time.

If you want to compare GoW to something just head on over to Unreal Tournament 2007. Same engine as GoW, same resolutions, same effects, same AA abilities. It just runs a hell of alot smoother on the PS3 utilizing unoptimized code.

Your debating publicly known information dude. The RSX has been in co-development since early 2002. Go read something for the love of GOD lol. They announced back in 2004 that they have been working together for 2 years on the GPU for the Playstation 3. The G70 will have very little similiarities when its all said and done with the RSX.

You know 2 other major corporations worked on the Cell processor besides Sony. I just provided you not one but two quotes about Cell and Ray-Tracing. READ MAN lol. And you keep referring back to Nvidias passed projects. The RSX is a new thing man, learn to live with that.

Nvidia flat out said this in plain English at E3. The RSX is a custom built GPU that only has rather basic architectural similarities to the G70 series. Everything else will be custom, EVERYTHING ELSE. So you can pretty much throw all your basic arguments out the window.

This coming from the guy who actually thought Nvidia had been working on the RSX for less then a year. Come on, basic common sense. RSX was built from the ground up to be just that, a companion GPU for the Cell architecture.

Good lord. lol. And what games are being showcased now or in the future even show that the Xbox360 is a real competitor to current PC technology? Hmmm .. absolutely nothing. Whereas the PS3 games being showcased on unfinished hardware with relatively little development time are destroying anything shown in todays world in physics, graphics, whatever. I'll be sure to quote you on that though when 2007 comes around and the 360 is still trying to catch up to the PS3 launch titles lol.

Oh, also another important note is that nothing shown for the PS3 has of yet even used the RSX for crying out loud. Its all been done with either the dual Geforce 6800 setup avaliable in the Beta kits or the Geforce 7800GTX within the Alpha kits. The RSX will be utilized in the final development kits. You'll see a rather dramatic lift in performance just from that simple fact.

Oh and if the 360 is supposed to be capable of so much with this free AA and HDR lighting and all this other BS your talking about, why exactly have absolutely zero games utilized these features and why are so many games chugging along. Quake 4 was ported over to the 360 and if your talk is correct then well, something went HORRIBLY wrong. That game absolutely CHUGS on the 360, this supposed super graphics machine of yours. And this coming from ID, supposely from what you Microsoft people think, who are graphic masterminds.

As Sony put it so well recently in a PSM interview, " the proof is in the pudding ". The proof of what a system is capable of is announced loudly by what is actually showcased onthat system. From what we have seen from all launch titles and upcoming 360 games, the 360 isn't capable of anything that isn't easily matched on current PC's. And with the launch titles, Quake 4 actually handles one hell of alot worse on the 360 compared to its PC counterparts. The proof is in the pudding, indeed.

But wait you say, Quake 4 may have been a rush job. Well, UT2007 on the PS3 is a rush job too with unoptimized code running at 720p @49FPS on unfinished development kits not even using the specialized GPU for the PS3 with atleast 7 more months of pure development time to come. The proof .. is in .. the pudding ...

Oh, and to kill off another one of your false claims.

http://www.behardware.com/art/imprimer/599/

Please proceed to the bottom of the page listing the capabilities of each processor. And then please keep your eyes firmly within the HDR FP16 section of the table showcasing that even the 7800 GT and GTX are capable of this feature.

And now please, shut up about it.

im actually comparing non gameplay realtime rendering vs gameplay if you reall wanna know

no, when you work on the environment you focus on pixel/vertex shaders and texture rendering, thats the look of the game, the feel of the game (when you hold the controller in ur hand) comes from the AI and physics, the look is different from the feel of it. Its logical, are u logical?

actually this was a demonstration, a demonstration of real-time non gameplay always has alot of textures used for it. It might look better but not that much. It will just look as if gears of war is running on one core with all graphics, physics and AI on the same core, so u will expect GOW to be in improved as well.

Can you show me this INFORMATIOn that you have that RSX has been in development since early 2002? I would be pleased if you showed it

Ummm in september 2005 the nvidia CEO said to PSM " RSX uses the same theory and architecture of the G70, its just faster than the 7800 GTX" and yes its faster, 7800 GTX is 430 Mhz while RSX will be 550 Mhz. RSX has lower memory bandwidth than the 7800 GTX, and ALOT lower then the 7800 GTX 512 MB.

Actually Q4 is an example of bad porting , compare that to COD2, 60 FPS with HDR FP16 and 4xAA at full screen while PGR3 is FP16 HDR with 2xAA , same goes for Kameo so you do have games on Xbox 360 using HDR and AA at the same time with no loss in bandwidth, bandwidth will be lost in RSX because its based on the G70 architecture.

btw in that link you gave you forgot this

ATI and NVIDIA each support each a certain amount of different types of HDR. Here are the main differences:

- ATI supports FX10 blending, NVIDIA does not.

- ATI supports multi-sampling antialiasing in FX and FP, NVIDIA does not.

- FP16 texture filtering is free for NVIDIA not for ATI

Overall, ATI has more flexibility than NVIDIA for the number of HDR formats supported. However, as NVIDIA was the first to develop this technology, FP16 is the main format used. This can be a bit problematic for the Radeon X1K, which doesn?t normally filter FP16 textures and as long as shaders aren?t adapted to combat this deficiency.b>

Now you would think that would be bad for the ATI card in Xbox 360 but Xbox 360 has the 10 MB EDRam chip which nullifies the last advantage for Nvidia and gives ATi Xbox 360 card an advantage in all 3 HDR rendering areas, especially number 3 where it says ATI can AA in FX and FP while Nvidia cannot

Nvidia is capable of Fp16 i never said its not, I just said it will have JAGGIES.

Next....

This discussion leads nowhere as long as the final PS3 specs aren't in the wild. As soon as the first CELL workstations are sold, and the first devs that have no affiliation with Sony speak, it's all just speculation. We all know Sony usually uses very exotic hardware, so all I'm sure of right now is that it's wrong to assume the PS3 architecture is even similar to the 360's. They probably did some strange things as usual (even the PSP is full of mysteries, even with the official Sony SDK you have next to no idea what's really in that small black box).

On the raytracing issue: There's only a single chip available on the planet that's capable of realtime raytracing (SaarCOR RPU [1]). Neither the RSX, nor the CELL, nor any chip in the 360, not even ART PURE chips [2] can do that in any sane way. Of course, the CELL will be fast at raytracing, and modern GPUs can help processing frames faster (see Nvidia Gelato [3] for reference), but that's not usable for games. And we'll definitely not see realtime raytracing with a quality even close to MentalRay anytime soon.

[1]: http://graphics.cs.uni-sb.de/~woop/rpu/rpu.html

[2]: http://www.artvps.com/products.ihtml?page=pureoverview

[3]: http://film.nvidia.com/page/gelato.html

threedaysdwn = microsoft lapdog. :whistle:

Excuse me?

Please do your homework before making useless troll posts like this. I think I've made it quite clear that my discussions of the PS3 architecture have absolutely nothing to do with my employer. If you don't believe that, then that's your choice. Just keep the personal insults to yourself.

Furthermore, I have made no conclusions about the PS3's real-world performance (which would be impossible since it doesn't exist yet) - and only refuted the invalid or unsubstantiated claims of people like "udontneed2know."

Claims that the PS3 is architecturely superior to the Xbox 360 are completely unfounded in reality, just as are claims to the contrary.

I prefer to discuss facts, not opinions. It's a fact that rendering at 1280x720 in FP16 with 2x mutlisample AA will require 2-4 times as much bandwidth as the PS3 has available (assuming 60FPS). It's something in the neighborhood of 90 GB/sec. The RSX GPU has 22.4GB/sec of bandwidth to its VRAM. According to "udontneed2know" (though I've seen no documentation from Sony) - the RSX can also access the system memory using some sort of UMA or AGP-like interface. Even given the most optimistic numbers (22GB/sec + 25GB/sec for VRAM plus system RAM), you're still at about half the requird bandwidth.

You can cut the framerate down, reduce the size of the Z-buffer, make trade-offs in texture filtering, etc. But you're going to have to give up something. Even Sony can't defy basic laws of arithmetic.

This topic is now closed to further replies.
  • Recently Browsing   0 members

    • No registered users viewing this page.
  • Posts

    • Why it's almost impossible to produce a smartphone in the United States by Hamid Ganji If you look at the back of some Apple products, you can see the famous phrase “Designed by Apple in California, Assembled in China.” This phrase appears on products from one of the largest smartphone brands in the United States. These products are designed in the U.S., but their manufacturing takes place in China, India, Vietnam, or even Brazil. But why can’t Apple, as one of the largest American tech companies, produce its iPhones on U.S. soil? The idea for this topic came to me after the Trump Foundation launched a smartphone called the T1 and claimed that it was designed and built with American values in mind. However, this claim did not last long, as it was revealed that Trump’s phone was actually a rebranded HTC U24 Pro, with only a gold case and minor internal component changes. You see? Even a phone that is supposed to represent American values is manufactured in China. With a gross domestic product (GDP) exceeding $32 trillion, the United States is currently the world’s largest economy, while China ranks second with around $20 trillion. On the other hand, the United States is by a wide margin the global leader in various technological fields, and American companies spend hundreds of billions of dollars annually on research and development. From Apple and Google to Microsoft, Lockheed Martin, Boeing, and others, American tech and industrial giants lead their foreign competitors in many sectors. The United States also has no shortage of smartphone brands. Apple, Google, and Motorola are among the major brands in the smartphone market, collectively holding a significant share. However, the vast majority of their products are manufactured outside the United States. So why is it that the world’s largest economy, home to the most advanced technology companies and industrial powers, cannot produce a smartphone on its own soil? Let’s explore this question together. Even threats to impose tariffs won’t work After Trump entered the White House as the 47th President of the United States, his administration adopted strict tariff policies. One of these policies was the imposition of a 25% tariff on smartphones manufactured outside the United States. Trump said he “had a little problem” with Apple CEO Tim Cook over producing smartphones outside the U.S. So he thought that threatening a 25% tax on imported phones might force Apple to bring manufacturing back to the United States. “I have long ago informed Tim Cook of Apple that I expect their iPhones that will be sold in the United States of America will be manufactured and built in the United States, not India, or anyplace else,” Trump wrote on Truth Social. Image via The White House Although Apple currently manufactures some of the iPhone’s chips in the United States with TSMC's help, it still shows no willingness to shift full iPhone production to the country. At the time, renowned Apple supply chain analyst Ming-Chi Kuo wrote on X, “In terms of profitability, it’s way better for Apple to take the hit of a 25% tariff on iPhones sold in the US market than to move iPhone assembly lines back to the US.” However, manufacturing a smartphone in the United States is not as easy as it might seem, and many technical and economic barriers are involved. The lack of necessary manufacturing hubs There is a clear reason why many companies prefer to manufacture their products in China. China has established itself as the main global manufacturing hub for international companies, and over the past few decades, large contract manufacturers have emerged there, allowing companies like Apple to outsource production. One such example is Foxconn, which also manufactures some Apple products in India. Building the infrastructure required to produce smartphones in the United States would require tens of billions of dollars in new investment. Factories would need to be built, essential manufacturing equipment would have to be installed, and, most importantly, a skilled workforce capable of operating these systems would need to be recruited and trained. The United States currently lacks the core infrastructure needed to manufacture smartphones, and for this reason, many companies prefer to outsource production to Chinese contractors rather than spend tens of billions of dollars to build that infrastructure, which is significantly more economically efficient. Additionally, building such infrastructure in the United States could take up to a decade, ultimately leading to a significant increase in the product's final price for consumers. Shortage of trained labor in the U.S. compared to China Decades of serving as a global manufacturing hub have allowed China to build a massive talent pool in the production sector that is almost unmatched worldwide. Today, if a company chooses to manufacture its products in China, it can be confident that the workers involved in production have years of experience in their respective roles and are capable of producing high-quality goods with minimal errors. Even if we assume that tens of billions of dollars were invested in building smartphone manufacturing infrastructure in the United States, finding skilled workers would remain highly challenging. Apple CEO Tim Cook visiting the iPhone 6 assembly line in China in 2014. Image: Tim Cook on X In a 2015 interview on CBS’s 60 Minutes, Tim Cook said the main reason Apple isn’t producing in the US is a lack of skills. "China put an enormous focus on manufacturing, in what you and I would call vocational kind of skills. The US over time began to stop having as many vocational kinds of skills. I mean you could take every tool and die maker in the United States and probably put them in the room that we're currently sitting in. In China you would have to have multiple football fields,” Cook said. Also, in 2017, at the Fortune Global Forum in Guangzhou, Cook once again emphasized the importance of highly skilled Chinese workers. “China has moved into very advanced manufacturing, so you find in China the intersection of craftsman kind of skill, and sophisticated robotics and the computer science world. That intersection, which is very rare to find anywhere, that kind of skill, is very important to our business because of the precision and quality level that we like. The thing that most people focus on if they’re a foreigner coming to China is the size of the market, and obviously, it’s the biggest market in the world in so many areas. But for us, the number one attraction is the quality of the people,” Apple CEO said. Higher labor costs in the United States Producing almost any product in the United States is more expensive than in many other countries, and one of the main reasons is the higher cost of labor in the U.S. According to the Bureau of Labor Statistics, median weekly earnings of full-time workers in the United States were $1,235 in the first quarter of 2026. Meanwhile, the average annual salary in China's private sector in 2025 was RMB 71,590 (US$9,961). In many parts of the world, the weekly wage of an American worker is equivalent to several months of income. Another important factor to consider is that in the United States, the workforce capable of working on a smartphone assembly line is highly specialized and therefore commands higher-than-average wages. According to an estimate by Bank of America, producing an iPhone in the U.S. is technically possible, but “iPhone cost can increase 25% purely on higher labor cost in the U.S.” However, this 25% increase applies only if final assembly is performed in the United States while components are still sourced from China or elsewhere. In this case, the price of a base iPhone would rise from $799 to around $1,000. But in another scenario, if Apple were to produce the required components for the iPhone within the United States, production costs could increase by more than 90%. Trump’s dream for a “Made in the USA” iPhone might never come true In a free-market capitalist economy, one of the primary responsibilities of any CEO is to maximize profit. Using Apple as an example, Tim Cook’s role is to maximize the company’s profits so that it can fund research and development for new products and invest in areas such as artificial intelligence, while also keeping shareholders satisfied. Therefore, it is entirely understandable that Apple would choose not to bring its manufacturing back to the United States and instead keep production in countries where labor is cheaper, and products can be manufactured at a lower cost, thereby maximizing its profit margins. What is your opinion about manufacturing smartphones in the United States? If you are an American citizen, would you be willing to pay hundreds of dollars more for a smartphone made domestically in the USA? Let us know in the comments.
    • Cheers everyone for the replies. It's been very useful. 👍
    • Compared to the 7735HS it is around 25-30% slower in multi-threaded tasks (according to Google search) I did a review of the 7735HS Beelink SER6 Max in 2023, but thinking about it, it's not comparable to the 7730U. For the example you gave about how it will be used, the 7730U is actually an excellent choice for its power and battery efficiency.
  • Recent Achievements

    • Reacting Well
      JuvenileDelinquent earned a badge
      Reacting Well
    • One Month Later
      Excellence2025 earned a badge
      One Month Later
    • Week One Done
      Excellence2025 earned a badge
      Week One Done
    • Week One Done
      flexorcist earned a badge
      Week One Done
    • Week One Done
      Woland13 earned a badge
      Week One Done
  • Popular Contributors

    1. 1
      +primortal
      503
    2. 2
      +Edouard
      194
    3. 3
      PsYcHoKiLLa
      151
    4. 4
      Steven P.
      72
    5. 5
      FloatingFatMan
      67
  • Tell a friend

    Love Neowin? Tell a friend!