TechSpot: AMD FX-8350 and FX-6300 Piledriver Review

About this time last year, AMD's new Bulldozer-based FX series launched to bright-eyed system builders who expected the new architecture to challenge Intel's increasingly comfortable position in the upper-end processor market. Unfortunately, Bulldozer wasn't all it was cracked up to be. Its performance fell short of the then 9-month-old Sandy Bridge processors and in some cases, even failed to surpass the Phenom II range.

Following Bulldozer's mediocre reception, AMD insisted that the new architecture was still young and would serve as a "solid building block" for the FX series. Although hotfixes such as one that addressed an SMT inefficiency have boosted Bulldozer's performance slightly, little has changed with AMD's FX series in the last year -- until now, anyway, with today marking the arrival of the company's second-generation FX offerings.

AMD is refreshing its desktop processors with Piledriver, an enhanced version of Bulldozer that focuses on improving instructions per clock and frequency -- something we witnessed earlier this month when we tested the company's new Piledriver-powered Trinity APUs. In other words, instead of a major overhaul, Piledriver picks up where Bulldozer left off, which may disappoint those who wanted AMD to abandon the architecture.

Read: AMD FX-8350 and FX-6300 Piledriver Review
This article is brought to you in partnership with TechSpot

Report a problem with article
Previous Story

Microsoft reveals more on Xbox SmartGlass

Next Story

Lenovo giving free Windows 8 notebooks to "Sky Warriors"

24 Comments

Commenting is disabled on this article.

I see all the well documented and scathing attacks on this hack review were deleted. The majority of comments shown just 48 hours ago were extremely critical of the poor methodology in this testing. Why were they del;ted. You are obviously shills for Intel. Dissgusting intellectual lzainess and over reliance on synthetic benchmarks , insufficent real-world application testing. What happened to Battlefield III , a game that is designed for multprocessors.why were no sim type games ,where database technology is employed and the FX chips excell in this way. Win Rar should NOT even be included it is a dying application and very poorly designed. win zip is a threaded app where the FX was at the top of the heap.
Also that memory app that you ran isan old single threaded app and others are avialable and far better in design. Single threaded apps are a dying breed, yet the majority of the games and apps you chose to run were based on the dying technology. There was not adequate balance in your testing. The tester is once again flawed in his testing design and intelllectually lazy in my opinion.
I do want a retest done as I stated before in one of the critical remarks that you censored from the comments here. If you can't take the heat then don't do a bogus test. to make you guys like good , You did a great disservice to the user base here. I used to respect Tech Spot, I can't say the same now. Tom's Hardware site did a much ore thorugh test and designed it in abalanced way with more real-world testing than the review here.

I bought an FX-6100 when it first came out and, for the price, it's been absolutely amazing. I love it and would buy it again.

Disappointing, but not unexpected. From what I've read, Piledriver was never going to be a massive improvement over Bulldozer, Steamroller will fill that gap next year (hopefully), but even so, there doesn't really seem to be any compelling reason to buy AMD over intel at the moment. Even if the FX8350 can outperform a similarly priced Core i5, the power consumption is still more, which makes Intel the cheaper option longer-term, while being about as good on the price/performance scale.

I hope AMD can pick up the pace again, I really do. I've got good hopes for Steamroller, since I'm going to be looking for a new build around then, but I hope that they can compete with Haswell.

So on single-threaded performance they still lag far behind Intel, but on multi-threaded performance they're competitive. Who wants a CPU that's only competitive in heavily multi-threaded scenarios? Single-threaded performance is still what matter most of the time. And as Anandtech pointed out, it doesn't look like AMD will ever catch up on Intel on that level, at least if current trends continue.

I don't know how AMD put themselves in this hole, after dominating Intel for such a long time (Athlon 64 vs P4 era), but they need to step up their game big time.

Dr_Asik said,
So on single-threaded performance they still lag far behind Intel, but on multi-threaded performance they're competitive. Who wants a CPU that's only competitive in heavily multi-threaded scenarios? Single-threaded performance is still what matter most of the time. And as Anandtech pointed out, it doesn't look like AMD will ever catch up on Intel on that level, at least if current trends continue.

I don't know how AMD put themselves in this hole, after dominating Intel for such a long time (Athlon 64 vs P4 era), but they need to step up their game big time.

You do understand that there are 2 types of people, one who need multi-threaded workloads daily and the other that need single-threaded? If I deal with multi-threaded workloads daily, hell I'll go with the cheaper part that lacks single-threaded as I just don't need it, simple as that

alwaysonacoffebreak
You do understand that there are 2 types of people, one who need multi-threaded workloads daily and the other that need single-threaded?
No, because that's wrong. Everyone benefits from better single-threaded performance because most applications and services, including those you use every day like all your background services, video players, browsers, games, etc. are limited by single-threaded performance.

Dr_Asik said,
No, because that's wrong. Everyone benefits from better single-threaded performance because most applications and services, including those you use every day like all your background services, video players, browsers, games, etc. are limited by single-threaded performance.

Video games and video players are single-threaded? Kthxbye. Nothing to talk to you about anymore.

alwaysonacoffebreak said,

Video games and video players are single-threaded? Kthxbye. Nothing to talk to you about anymore.

I didn't say they were single-threaded, I said they were limited by single-threaded performance. See benchmarks where Intel CPUs dominate in games. You fail at reading comprehension.

Dr_Asik said,
I didn't say they were single-threaded, I said they were limited by single-threaded performance. See benchmarks where Intel CPUs dominate in games. You fail at reading comprehension.

Please do tell me where does any of the Intel CPU's dominate anything here:

http://media.bestofmicro.com/X.../battlefield%203%201680.png

Games are mostly GPU dependent not CPU, there are maybe 3-4 CPU heavy games, that's it. One of them is Skyrim

http://media.bestofmicro.com/Y.../original/skyrim%202560.png

Yeah Dominates. The differences are minimal in games, even if they were bigger you wouldn't notice a thing.

Dr_Asik said,
I don't know how AMD put themselves in this hole, after dominating Intel for such a long time (Athlon 64 vs P4 era), but they need to step up their game big time.

I'd assume R&D budgets are the problem mainly. AMD never really had an opportunity to capitalize on the Athlon 64's because of Intel's shady OEM agreements in the EU (despite P4 sucking so hard, if you wanted a pre-built PC, that's all you could find; Ath64s were reserved for system builders and custom machines). The €1b fine wasn't nearly enough reparations for the losses AMD suffered as a result of Intels practices, and since then they've always been relatively strapped for cash.

I think that AMD have sort of bet it all on Bulldozer, they can't afford to say "whoops, we screwed up, so we're going back to the drawing board", and they're proceeding with Bulldozer and making iterative refinements. Hopefully Steamroller will offer the kinds of performance improvements we're looking for from AMD, but by then Haswell will be out from Intel as well, and there's a risk that AMD will be on the back foot again

Dr_Asik said,
No, because that's wrong. Everyone benefits from better single-threaded performance because most applications and services, including those you use every day like all your background services, video players, browsers, games, etc. are limited by single-threaded performance.

Um, only if you are stuck in the world of Linux or OS X.

Windows NT and especially background services and its driver model are highly threaded in design. Even if a single driver was poorly written with no threading, the OS would still be able to send it off to its own CPU as it managed other process threads and drivers.

The trick in the equation is how large are the chunks of execution being multiplexed. For the majority of users, AMD will do just fine by them, and in day to day operations single threaded code hitting hard is not something you find outside of gaming or high end computational software that are going to max out all the Cores, like rendering a Video.

This is also not an easy formula, as the i7 with hyper-threading enabled, even with all CPUs and virtual CPUs maxed out could out perform a 8 Core AMD CPU.

So for gaming and high end users, YES you are correct, stick with the CPU that has the best single Core performance. However, if you are not a gamer, and a casual user, the chance you will be able to measure the performance is questionable.

(Look at PassMark test for example. This is more like what an average user would find, and a nice 6 or 8 core CPU is going to be faster for them than an i3 or i5, even though the single core execution is far slower. OSes have a lot of things going on in the background, and users are more conditioned to have several Apps running to be taking advantage of the extra Cores that will offset the extra speed the primary application would be getting.)

Also if you are talking about Linux or OS X, they do poorly in SMP performance in general, and with the massive amount of OS and Kernel hard locks they are less able to use the additional cores than Windows.

(On, OS X, unless all software is written with the newer SMP APIs, an older Application or even part of the OS itself can introduce rather nasty Locks that drop even a nice 8 Core Mac Pro to 1 or 2 Cores. Based on average user usage, this is happening around 20-40% of the time they are using the system, and even when it is scaling up to use the full 8 CPUs, the overhead throws a 20-30% hit overall in system performance to gain the use of the additional CPUs.)

alwaysonacoffebreak said,

Please do tell me where does any of the Intel CPU's dominate anything here:

http://media.bestofmicro.com/X.../battlefield%203%201680.png

Games are mostly GPU dependent not CPU, there are maybe 3-4 CPU heavy games, that's it. One of them is Skyrim

http://media.bestofmicro.com/Y.../original/skyrim%202560.png

Yeah Dominates. The differences are minimal in games, even if they were bigger you wouldn't notice a thing.

Good job picking meaningless GPU-bound benchmarks.

So for gaming and high end users, YES you are correct, stick with the CPU that has the best single Core performance. However, if you are not a gamer, and a casual user, the chance you will be able to measure the performance is questionable.
Well, very casual users (like my mom) are fine with 4-year old Athlon X2s at 30$ really, so of course for these people performance doesn't really matter. But as soon as you're gonna be waiting on a CPU-bound operation, chances are it'll not be something that scales well to multiple cores. So in the general case users are better served with the CPU that has better single-threaded performance. Yes the more cores you have the more stuff can run in parallel and that should in theory help with your 70 background services, but if you look at what they're doing they're actually using 0,01% of the CPU or less most of the time, so they could actually all just run on one core and be fine. But once in a while, one of them might cap one core 100%; that's the one time you'll be CPU-bound and you'll wish you'd have better single-threaded performance then. The AMD will shine when you're capping all your cores with a task that scales linearly with cores (like some video encoders), so if you're building a machine specifically for that, it can be the best performance/price ratio. But not in the general case.

Why, yes. A day has come when 4 GHz boundary has been broken and it still can't outpace Intel properly, but just keep up, with certain compromises.
While Intel is busy cooking Haswell and 22nm process fine arts, AMD is literally begging you to buy the new stuff. It's so cheap that I wonder if 23% "improvement" will be their next quarter profit loss.
Radeon engineers throwing their expertise into Llano (and hopefull continuing to do so amidst the aggressive layoffs - the only thing that's aggressive about AMD) is what keeps them above the waterline still.

The worst part was the last paragraph: for those with a high end phenom II quad core (that's what, 2 years old now.), AMD still doesn't have a viable upgrade for you.

I want to make an AMD build next time I upgrade, I really do. But they really need to get a decent CPU out.

Read all the reviews, so sad that AMD could not catch up. These days, it's ARM that's pushing Intel - to go lower power.

a1ien said,
Read all the reviews, so sad that AMD could not catch up. These days, it's ARM that's pushing Intel - to go lower power.

Indeed do read. Even tomshardware, the biggest intel fansite out there, says that FX is going into the right direction. Overall it's 91% of the performance of an 3770k which is far from this price point.

alwaysonacoffebreak said,

Indeed do read. Even tomshardware, the biggest intel fansite out there, says that FX is going into the right direction. Overall it's 91% of the performance of an 3770k which is far from this price point.

Even if the overall speed would be the same it's still distributed about twice the cores, meaning each core is half the speed of an Intel one and that's very bad since most applications aren't optimized for multicore or don't scale properly. Also the price isn't really that much far if you compare it to a 3570k that is basically a 3770k without the (useless?) hyperthreading.

francescob said,

Even if the overall speed would be the same it's still distributed about twice the cores, meaning each core is half the speed of an Intel one and that's very bad since most applications aren't optimized for multicore or don't scale properly. Also the price isn't really that much far if you compare it to a 3570k that is basically a 3770k without the (useless?) hyperthreading.

I agree, most programs are not optimized for AMD's arch yet But the ones that are show some good progress for AMD and the software creator. Also the price difference was a comparsion taken from Semiaccurate since they compared 3770k to 8350 not the 3570k. I think the i5 is like 20 bucks more expensive than the 8350? And that's the place where you must make up your mind what to you do most, single or multithreaded workloads. And ofcourse if you're (no you as in you but in general) living in your moms basement not paying a dime then she might get mad for the the electricity bill next month, I don't have that problem so I barely care for the 125W TDP.

Oh my! I totally forgot about the higher power usage!

From Anandtech benchmarks the setup system with the 8350 on full load used 95 more watts, almost twice the power used by the 3570k system. Intel always declares very pessimistic TPD estimated values but I wouldn't have never expected such an huge difference in the real values, how is that even possible?

With such an huge power usage difference it wouldn't take many hours to fill the 20$ in the price gaps if you use CPU-intensive applications, not to mention the money you'd have to spend on a better cooling system since twice the power consumption means twice the thermal power to dissipate as well.

Unless I read that benchmark wrong even remotely considering that processor, unless the motherboards are much cheaper, now just seems a completely retarded idea. Even if the motherboards were much cheaper though I still wouldn't bet their reliability since they'd have to handle that crazy wattage range.

francescob said,

Even if the overall speed would be the same it's still distributed about twice the cores, meaning each core is half the speed of an Intel one and that's very bad since most applications aren't optimized for multicore or don't scale properly. Also the price isn't really that much far if you compare it to a 3570k that is basically a 3770k without the (useless?) hyperthreading.

Hyper-threading has pros and cons. However, for the majority of consumers, it is a good thing.

(Unless you are pushing the CPU Thermal range or dealing with code that the CPU is hitting prediction at high rates, hyper-threading does offer additional threading paths that work well.)

Some of the best demonstrations on are older CPUs when it was introduced on Pentium 4s, the multi-tasking performance was visible noticeable.

You are spot on with your assessment of where AMD is right now, as they expected multi-core/parallel/Async coding and processing to be far more used by this timeframe than it is.

If we were truly gaining speed from pure addition of Cores/CPUs, AMD would have the advantage, AMD has the capability to easily offer 16 core CPUs in the low i7 price range. Since this isn't the current state of software, they have no advantage to pushing anything beyond their 8 core CPUs for consumers, and in highly threaded software and concurrent operations, they do well.

However, as you state, the single core execution of AMD is running at about 50% Intel right now, and Intel can hit the appearance of 8 Core threading with an Quad Core with HyperThreading that is faster even in multi-threaded applications than the 8 core AMD CPUs.

AMD planned on the world being to the point of hitting single core execution limits in 2012 and multi-core is the answer when that happens. However, we aren't there, and they have lost their lead in this area to Intel.

AMD also didn't expect Intel to get better performance out of upper end Atom processors or get the power efficiency Intel is getting out of the i5/i7 series processors.

At the very least AMD thought they would be nailing Intel with superior integrated GPU performance, and the newer HD 4000 from Intel is lagging behind AMD's integrated offering, but not enough that it is worth dumping so much CPU performance over.


In a way I feel bad for AMD, and see where they were heading with lots of cores and low power and good graphics, but the end result is not implemented in a way that is attractive to builders, and not enough of a discount most of the time.

There are some AMD products that can hit the lower end i3 and can hit Atom, but outside of that slim area, AMD has nothing.


alwaysonacoffebreak said,

I agree, most programs are not optimized for AMD's arch yet But the ones that are show some good progress for AMD and the software creator. Also the price difference was a comparsion taken from Semiaccurate since they compared 3770k to 8350 not the 3570k. I think the i5 is like 20 bucks more expensive than the 8350? And that's the place where you must make up your mind what to you do most, single or multithreaded workloads. And ofcourse if you're (no you as in you but in general) living in your moms basement not paying a dime then she might get mad for the the electricity bill next month, I don't have that problem so I barely care for the 125W TDP.

It isn't just AMD's architecture, it is the exponential increases in performance gained from multi-core designs are not here yet.

If you look at even well done OS and Application technologies there are still some hard limits that haven't been dealt with fully, and the ones that have are not widely in use or fully credible solutions.

Sony had the same problems with the PS3, not only were game engines and developers not geared for getting the most out of the 6 Cell cores, but Sony's development tools and based OS was horribly inefficient at async and parallel computing. Sony got a bit better, but still never reached the potential that is there.

Microsoft did better with the Xbox 360, as it is also capable of 6 effective threads, being a tri-core design with hyper-threading type technology. However, they also hit some of the same issues, but being Microsoft had a better OS and compiler technology to start with.


AMD/ATI also has had the same problems with their GPU designs, in theory they throw in a lot of pipelines, yet Nvidia still ends up killing them with fewer stream processors and throwing more MHz at the problem.

This again is a multi-processing issue.

Microsoft has some of the best work on parallel and async languages and compiler technologies right now, but some of them are just not ready for production, and there is still a HUGE gap in developers handing threads on their own better.

(Check out F# and other things Microsoft R&D is working on with compiler technologies that help developers that still having trouble doing threading well.)

With Windows 7, Microsoft also assumed that more cores would be the focus of where we would be today. They specifically redesigned the SMP technology in the NT kernel in Win7 to group and handle more CPUs more efficiently than any OS seriously tried to deal with.

NT and other OSes could in 'theory' handle a 1000 CPUs, but the overhead at even 16 CPUs in managing threads would overcome any gains in the additional CPUs, which is why Clustering and other segmented technologies were a better choice than dumping in several CPUs in a Server.

**Even specialized Linux builds for SMP start choking on overhead at 8 CPUs, with serious scalability problems at 16 CPUs. Sadly, the move to clustered servers and not dealing with SMP better in Linux is one reason the industry has retarded SMP progress that affects development in the Server and Windows worlds, as well as even GPU circles.

Microsoft has detailed how to get better SMP performance, that even Linux can use; however, changes like this are easy on NT, but cascade through Linux and are a major rewrite.

Windows 7 has little overhead loss of managing threads with 64 processors and not a lot of overhead in managing even 256 processors in the Server Editions. So this makes desktop 64 CPU configurations very feasible and 256 processor Servers feasible and faster than a traditional 8 CPU/Core setup of 32 separate servers where intercommunication is the bottleneck.


So Microsoft expected to see more low powered and multi-core CPUs in 2012 as well, or they would have left the SMP changes to be delivered in the Windows 8 timeframe. We just don't have them, at least not reasonably. AMD could be putting out 16 core systems, but since they are not giving users any more performance than their 8 core models, it is not worth discounting the prices, and leaving them for servers when they can charge significantly more.

Anyway, it isn't just and AMD architecture problem, it is a failing in the industry as whole, and Intel has done well to capitalize on this, and use the time to be ready for higher end multi-core drops when we need the extra cores. (Think of things like a 16 core Atom processor in a Tablet, that turns on cores as needed and scales up for better performance, yet can idle on a tiny amount of power. If Intel gets Atom a bit closer to ARM in power consumption, this would kill off ARM advantages. However, the same could be true of ARM multi-core designs.)

thenetavenger said,

Hyper-threading has pros and cons. However, for the majority of consumers, it is a good thing.

(Unless you are pushing the CPU Thermal range or dealing with code that the CPU is hitting prediction at high rates, hyper-threading does offer additional threading paths that work well.)


Sorry if I was ambiguous, what I meant was that having hyperthreading may have no advantage compared to the 3570k because if you buy a K-series CPU (specific for overclocking) you often have to turn hyperthreading off to achieve higher speeds so it becomes kinda useless. At that point you could just get the 3570k rather than wasting the money for the extra nominal 100MHzs of the 3770k or just buy the vanilla non-K processors that are cheaper and come with VT-d.

With all those pins it would be a perfect hair clip. See? There's always more you can do with AMD CPUs that you can't with Intel's.