Multithreading in Relation to Gaming

tooz

Distinguished
Apr 26, 2009
29
0
18,530
I'm still on a search for a CPU that is right for me...and now I have some more questions. I know that past three or four cores, games don't really respond to the amount of cores, or at least that's what this article said: http://www.tomshardware.com/reviews/multi-core-cpu,2280-11.html. I'm quite new to processors and chips of right now (I know how it works, but from a technical standpoint rather than knowing what's the best chip out on the market, and I have a REALLY limited knowledge of threads). And so now I am looking at how the number of threads may impact performance in gaming. Does multithreading really provide that much advantage now or in the future in gaming, or in other words, will games take advantage of multithreading abilities? Also, I have a really quick question: will current processors like the i7 and PII handle games coming out for the next, say 5 or 7 years? Thanks for any help.
 

smithereen

Distinguished
Oct 4, 2008
1,088
0
19,310
5-7 years is a long time - I'd aim a little lower. For gaming, the sweat spot right now is a Phenom II X3 720/Core 2 Duo e8400, but they are less futureproof than, say, a Core i7 920. An AM3 chipset with a PhII 720 would probaby be the best value and most upgradable gaming platform right now in the medium-high end.

Also, that benchmark is biased agaist triple-cores because it is using a double-cheeseburger quad, which has a bottleneck in that the 3rd and 4th cores are seperated from the first and second by the FSB. In a Phenom II triple, the cores are all on the same die, so the performance will increase almost linearly. Most games utilize two cores, leaving the third for Windows and other things running in the background.
 

tooz

Distinguished
Apr 26, 2009
29
0
18,530

Alright thank you for the input. So are saying AMD is better for gaming at the price, or is the i7 better overall and is more futureproof? So than should I just ignore the i7, or should I look into it?
 

smithereen

Distinguished
Oct 4, 2008
1,088
0
19,310
It depends on your budget. In gaming, for the forseeable future, Core i7s will be overkill outside of top-of-the-line multi-GPU configurations. The only real advantage the i7 has in gaming is that it handles SLi-Crossfire overhead better, but you will almost always be GPU-bottlenecked with any modern processor, especially after you overclock them. All modern CPUs overclock *very* well, especially Intels and AMD Black Editions.

The i7's advantage is applications like encodign and compressing that are very, very multithreaded and like high instructions per clock, for everything else, Phenom IIs and Core 2 Quads should be good for the foreseeable futrue, but the Phenoms are better value and aren't on a dead-end platform quite yet.

Look into i7s only if you have money to burn, I think.
 

tooz

Distinguished
Apr 26, 2009
29
0
18,530

Thanks a bunch for all the help!
 
i7's aren't that much more expensive, especially with costs of DDR3 dropping.

We're starting to see more games take advantage of more cores. I wouldn't recommend 2 cores anymore, and I think even 3 won't be enough in two years. If you go the X3 route, make sure you have an upgrade path to a X4, just in case.
 


Do ou mean biased towards a triple core? Because normally the FSB is seen as a bottleneck with the C2Qs.

but from what I have seen since most games don't rely heavily on the CPU at all then the performance bottlnecks at the GPU. Except for Source based games where even a Phenom II X4 and a C2Q are about on par. In some cases a C2Q does better due to better IPC but thats mainly the 45nm C2Qs.

Core i7 itself is a good way to go for at least 3 years and since it is supposed to have dropin support for Westmere I would go that way. Plus it will probably not bottlneck a SLI/CF system until the next gen of cards comes out next gen of cards comes out next year and with drop in Westmere then you just upgrade the GPUs again and you are set.

But currently a good few games support quads+ and soon more and more will. I bet we will see them coming sooner too since it does have advantages such as being able to offload physics and particle effects to the CPU and letting the GPU focud on the polygons and textures.

Even a older C2Q Q6600 will last another 2-3 years for gaming but thats streatching it a bit.
 


You contrdict yourself: A benchmark that's negatively influenced by the presence of a Front Side Buss can *NOT* be bias'd against an AMD processor. AMD (and Intel Core I7) do not have a FSBs. No FSB = No Bias

I'm sure we can all accept an argument that it's performance is perfectly good for the intended usage (which it clearly is), especially in light of the fact that Graphics are the main determinant in gaming performance. But the simple truth why the AMD processor does not measure up on those benchmarks is because it's not as fast.
 

smithereen

Distinguished
Oct 4, 2008
1,088
0
19,310


You misunderstand me. The review I'm talking about simply took a Core 2 Quad and manually disabled cores. Obviously, performance did not scale linearly after the third and forth cores were enabled, and the FSB bottleneck appeared. No comment was made by the reviewer, other than to say "applications don't seem to take advantage of the third core as much as the forth and second".
 
No... I understand very well, I think: You're referencing an article - a link to which you don't provide - in which a reviewer intentionally disables cores of an Intel quad to show FSB limitations... It is quite well known, I think, that an FSB based system delivers less bandwidth than one with an integrated controller. Yet meaningless on a desktop machine, since very few consumer applications come vaguely close to the bandwidth limitations of a Front Side Buss design, let alone an integrated one.

And what applications did the reviewer in the Article Which Is Not Provided use? Are they suitably multithreaded? Or are they typical desktop apps, which are normally not multithreaded at all? You *DO* understand that the application itself has great bearing on multithreaded performance, do you not?

You say the reviewer made no further comment, and that all he did was play with cores on an Intel C2Q. If there is/was no comparison to an AMD processor, where is the basis for your own assertion? If there *is* a comparison to an AMD processor, are we to infer that the Intel quad "won" anyhow? And that therefore the review was biased?


This smells like random information, pulled out of context, and then misused to support a preconceived idea. Why? There are holes in your logic large enough for a 12 year old to drive a truck through.
 

smithereen

Distinguished
Oct 4, 2008
1,088
0
19,310
I was addressing the OP, who provided a link to the conclusion of an article which he referred to for information on the benefit additional cores provide in gaming. I merely informed him that all currently available triple-core processors, and both the Core i7 and Phenom II quads, would scale noticeably better than a Core 2 Quad, and that additional cores above two would be more beneficial than the 'review' implied.

Here is the entire article.
 

smithereen

Distinguished
Oct 4, 2008
1,088
0
19,310
I displays a lower boost in performance from 2-3 cores than from 3-4 in fully multithreaded applications. A Core 2 Quad uses the FSB to communicate between the two separate dies. When the third core is added, bottlenecking appears, reducing the performance boost. When a forth core is added, to a lesser degree even more bottlenecking occurs. This makes the article biased against Phenom 2s and Core i7s, which are monolithic quads and triples, especially triples, as the bottleneck is more noticeable - less computing power is being added to compensate for the bottleneck.
 

tooz

Distinguished
Apr 26, 2009
29
0
18,530
Yea, after doing some more research, I see your guy's point, too. http://www.tomshardware.com/reviews/phenom-ii-940,2114-19.html shows some performance differences, especially in WiC (which I love). So if there are other games out there that use the i7 architecture like WiC, then I guess the jump to i7 will be worth it. So now my only question is how long is Intel planning to support LGA-1366 with i7? I heard something about a "Westmere" chip coming out, and I am wondering whether that will be supported by the i7 architecture. Thanks.
 



You have it backwards: AMD's quads and triples, and Intel's i7 do not have a Front Side Buss - therefore they do not have the bottleneck you are referring to. The C2Q *does* have a front Side Buss, so it suffers in certain memory/throughput related benchies.

 

smithereen

Distinguished
Oct 4, 2008
1,088
0
19,310
That's what I've been trying to say!!! But article the OP cited uses a C2Q, so the triple and quad core benches sufferer from a FSB bottleneck. The article features no i7s or Ph2s, but merely draws erroneous conclusions about the benefits of triple and quad core processors, because of the multiple-die architecture.
 
You did it *again*

But article the OP cited uses a C2Q, so the triple and quad core benches sufferer from a FSB bottleneck.


The Triples and Ph/i7 Quads Do Not Suffer From A FSB Bottleneck Because They Have No FSB. The C2Q is the one with this particular limitation.
 

smithereen

Distinguished
Oct 4, 2008
1,088
0
19,310
I know that!!! The article merely made additional cores seem less effective by using a C2Q. Therefore, I merely told the OP that this bottleneck would not be present in a Ph/i7.
 

Dekasav

Distinguished
Sep 2, 2008
1,243
0
19,310
Ok, the article used a Core 2 Quad and manually disabled cores. Hence, the article sees a bandwidth limitation in the 3 core variant (as it IS a C2Q, it uses the FSB, and it bottlenecks slightly).

They are NOT testing an AMD tri-core, they're testing a C2Q with a disabled core.
 

randomizer

Champion
Moderator
I think I understand the logic here, even if it's a bit strange. Only AMD currently makes tri-cores, and none of them use an FSB. You wouldn't go and buy a C2Q to disable a core and run 3. If you want a tri-core you'd buy an AMD one.

The benchmarks unfairly discredit tri-cores because they don't use a real tri-core in the review. A novice user would see 3 cores there and assume that an AMD triple cripple would have similar scaling. Had they used an i7 or Phenom II the scaling would have been different and more "fair".

Lets hope that made sense, I can ramble on a bit sometimes.