Jump to content



Photo

$2500 Build - Feedback?

build intel nvidia

  • Please log in to reply
42 replies to this topic

#31 Draconian Guppy

Draconian Guppy

    LippyZillaD Council

  • Tech Issues Solved: 3
  • Joined: 22-August 04
  • Location: Neowin

Posted 29 April 2014 - 19:49

it doesn't enable better processing across threads.

 

the problem today is that multicore games use a lot of one core, but then then barely make use of the other cores. This means for high end CPU's like i5 and i7's games today aren't bottlenecked by the CPU because the individual cores are so fast it doesn't bottleneck the GPU. DX12 will like mantle allow the performance to be spread out across the cores. meaning you get more equal usage of the cores so instead of one core having 70-90% utilization. it's more like 30, 20,20,20. or on more demanding games 50,50,50,50. though that's far in the future, cause that's a LOT of CPU power.

 

What's the difference on taxing a single core and keeping the other ones idle for something else, versus spreading the load on multiple cores?




#32 OP mastercoms

mastercoms

    Expert Microsoft Fanboy & C# Coder

  • Tech Issues Solved: 3
  • Joined: 21-May 13
  • Location: Marietta, Georgia
  • OS: W8.1U1 + Fedora 20
  • Phone: Lumia 928 WP8.1U1 Black

Posted 29 April 2014 - 19:52

I'll change my graphics card to a 770 and then get a better monitor, RAM, mouse, and keyboard.

What's the difference on taxing a single core and keeping the other ones idle for something else, versus spreading the load on multiple cores?

I believe then you can execute tasks synchronously with more cores.
So suppose you have Task 1, Task 2, and Task 3.
Single core:
Task 1 runs, task 2 runs, and then task 3 runs
Multicore:
Task 1 starts running on the main thread, task 2 is assigned to run to a second thread,and task 3 is assigned to run on a third thread. Then all 3 tasks run at generally the same time.

#33 HawkMan

HawkMan

    Neowinian Senior

  • Tech Issues Solved: 4
  • Joined: 31-August 04
  • Location: Norway
  • Phone: Noka Lumia 1020

Posted 29 April 2014 - 19:55

What's the difference on taxing a single core and keeping the other ones idle for something else, versus spreading the load on multiple cores?

 

Nothing if the single core (i5/i7) is able to do the work, everything if it isn't (AMD Phenom II)

 

there some other benefits of somewhat lower latency of doing 2.6 tasks simultaneously instead of serially as well. but nothing you'll notice for cpu tasks



#34 NinjaGinger

NinjaGinger

    Neowinian

  • Joined: 06-January 02

Posted 29 April 2014 - 19:58

Consider Corsair Dominator memory. Ive tried others and always come back to it.

#35 OP mastercoms

mastercoms

    Expert Microsoft Fanboy & C# Coder

  • Tech Issues Solved: 3
  • Joined: 21-May 13
  • Location: Marietta, Georgia
  • OS: W8.1U1 + Fedora 20
  • Phone: Lumia 928 WP8.1U1 Black

Posted 29 April 2014 - 20:04

Consider Corsair Dominator memory. Ive tried others and always come back to it.

I have seen good benchmark scores for Dominator Platinum memory, but I will have to see what will happen to my budget with that memory

#36 Colicab

Colicab

    DGDesign

  • Joined: 25-May 03
  • Location: Glasgow Scotland

Posted 29 April 2014 - 20:12

Get Steelseries for mouse. Razer mice craps themselves after a bit usage.

 

On the other hand, drop that keyboard. Get Razer Black Widow Ultimate (silent if you can order from razer store).

 

For the record: I have naga, naga hex, mamba, black widow ultimate, onza tournament.

Have Razer myself at mo, has been ok for a good while now tbh. But it replaced a Steel series and boy do I wish I`d kept it.

Got my Steelseries setup 6yrs ago and its still going strong. Love it.

Keep thinking I should replace my aging 7G keyboard or whatever its called, but it works and perfectly well.

So did the mouse, I just wanted to try something else and, yeh kinda wish I hadn`t



#37 Andre S.

Andre S.

    Asik

  • Tech Issues Solved: 10
  • Joined: 26-October 05

Posted 29 April 2014 - 20:15

Well it'll be a net win for actual multi core CPU's at least. well hyperthreaded will gain to since they use so little power form each core. but if you theoretically had a game that actually maxed each core, hyperthreading would then be a net loss. but that's all academic anyway since it's not an issue. 

I'm not sure you understand how hyperthreading works... if a game was using 100% of a non-hyperthreaded CPU it could easily be faster on a hyperthreaded CPU, just like video encoders currently do. The 100% of a hyperthreaded CPU is more than the 100% of a non-hyperthreaded CPU, because the latter wastes more time in thread switching than the hyperthreaded CPU does.



#38 HawkMan

HawkMan

    Neowinian Senior

  • Tech Issues Solved: 4
  • Joined: 31-August 04
  • Location: Norway
  • Phone: Noka Lumia 1020

Posted 29 April 2014 - 20:23

I'm not sure you understand how hyperthreading works... if a game was using 100% of a non-hyperthreaded CPU it could easily be faster on a hyperthreaded CPU, just like video encoders currently do. The 100% of a hyperthreaded CPU is more than the 100% of a non-hyperthreaded CPU, because the latter wastes more time in thread switching than the hyperthreaded CPU does.

 

Depends entirely on the calculations being done.

 

say for bucket rendering in 3DSMAX with finalRender which is highly thread optimized. say a 6 core i7 it would render faster with 6 buckets no HT than 12 bucket with HT. since each bucket it relies on as much raw power as the cpu an throw at it. 



#39 Andre S.

Andre S.

    Asik

  • Tech Issues Solved: 10
  • Joined: 26-October 05

Posted 29 April 2014 - 20:31

Depends entirely on the calculations being done.

 

say for bucket rendering in 3DSMAX with finalRender which is highly thread optimized. say a 6 core i7 it would render faster with 6 buckets no HT than 12 bucket with HT. since each bucket it relies on as much raw power as the cpu an throw at it. 

That sounds like a bug in 3dsmax. 



#40 HawkMan

HawkMan

    Neowinian Senior

  • Tech Issues Solved: 4
  • Joined: 31-August 04
  • Location: Norway
  • Phone: Noka Lumia 1020

Posted 29 April 2014 - 20:53

That sounds like a bug in 3dsmax. 

 

ummm why ? it's fairly natural.

 

each bucket will use 100% of a core. with HT one, you dn't have 100% of a core available. you have one core that pretends to be two. with overhead that means you have less total CPU power available. 

 

for other tasks that use a lot of small paralell tasks but very little cpu in each thread HT is a benefit. for rendering where each bucket will max the core it's on, it's not a benefit. 



#41 Andre S.

Andre S.

    Asik

  • Tech Issues Solved: 10
  • Joined: 26-October 05

Posted 29 April 2014 - 21:27

ummm why ? it's fairly natural.

 

each bucket will use 100% of a core. with HT one, you dn't have 100% of a core available. you have one core that pretends to be two. with overhead that means you have less total CPU power available. 

 

for other tasks that use a lot of small paralell tasks but very little cpu in each thread HT is a benefit. for rendering where each bucket will max the core it's on, it's not a benefit. 

100% of a hyper-threaded core is more than 100% of a non hyper-threaded core. For instance, while an integer operation is executing on one logical core, a floating point operation can be executed in parallel on the same core, since floating point and integer units are separate.

 

It's up to the software to take advantage of that. If a particular workload is pure FP or pure integer, for instance, it can issue a HALT instruction to the 2nd logical core to disable it, which effectively makes the core act as if non-HT. I don't know if 3dsmax does this. It's also possible that splitting the workload in smaller buckets is detrimental to caching performance. In any case, 3dsmax should know how to schedule itself for maximum efficiency on hyper-threaded CPUs; if enabling the feature drops its performance then that's a bug in the software.



#42 Javik

Javik

    Beware the tyrrany of those that wield power

  • Tech Issues Solved: 2
  • Joined: 21-May 12

Posted 29 April 2014 - 22:04

In the i5 v i7 debate something to also bare in mind is that i7 CPU's generally have more cache than i5's



#43 Draconian Guppy

Draconian Guppy

    LippyZillaD Council

  • Tech Issues Solved: 3
  • Joined: 22-August 04
  • Location: Neowin

Posted 30 April 2014 - 05:06

In the i5 v i7 debate something to also bare in mind is that i7 CPU's generally have more cache than i5's

But performance advantage in most cases is negligible