CPU/Mobo CPU bottleneck explained

Status
Not open for further replies.

hatter

Galvanizer
At Legion Hardware, a 4870X2 was paired with few CPUs (overclocked as well as underclocked).

This is what the result looked like




:rofl: So much for giving preference to Dual cores over quads for gaming.
Ok, CoH was extreme situation. The normal results were something like this :P



Conclusion was:

There were certainly a number of interesting results recorded when testing such a large range of processors with the Radeon HD 4870 X2. Perhaps the most interesting thing that we have noticed is how much faster the Core 2 Quad processors were when compared to the Core 2 Duo’s, this is not something that we have really seen before. This was really evident in Company of Heroes, where the Intel quad-core processors were roughly 100% faster than their dual-core parts.

While Company of Heroes was the most extreme example, all other games tested still gave the quad-core processors an advantage. Generally the performance margin between the Core 2 Duo and Phenom X4 processors was quite minimal, though more often than not the Intel dual-core processors did outperform the AMD quad-core parts. This is neither new nor surprising, as we have seen this time and time again in the past.

Although it was not all that surprising to see just how much the Athlon64 X2 architecture has aged since we last looked at it, we were still somewhat shocked at how slow the 3.46GHz 6400+ processor was. Obviously we do not expect that AMD Athlon64 X2 owners are going to race out and purchase a Radeon HD 4870 X2, but we still felt that including this processor series would be useful. Those still using an Athlon64 X2 should seriously consider upgrading their platform, even if they only plan to purchase a single current generation AMD or Nvidia graphics card.

Out of the five games used for testing, Devil May Cry was the only one that allowed the processors to find the limits of the Radeon HD 4870 X2. It would seem that this game is more GPU dependent than the others, as the Phenom X4 9850/9950 and Core 2 Quad Q6600/9650 processors were able to max out the graphics card. While the Core 2 Duo processors also did very well in this game, the overclocked 3.46GHz Athlon64 X2 6400+ processor was slower than the E6320.

Another game that strongly favored the quad-core processors was Enemy Territory Quake Wars. Here the Core 2 Quad Q9650 managed 133fps, while the Q6600 averaged 119fps. The Phenom X4 processors were next in line, with the 9950 averaging 115fps, and the 9850 reached 113fps. In fact the Phenom X4 9750 and 9650 processors were also able to outperform the E6850, which only narrowly edged out the 9550.

The Supreme Commander results were more like what we were expecting. Here the Core 2 Quad processors were only slightly faster than the dual-core versions, while the Phenom X4 processors were able to match the Core 2 Duo’s for the most part. However, Supreme Commander was another game that continually saw the Radeon HD 4870 X2 become faster, as the processors became more powerful and we never really saw the limits.

Finally, Unreal Tournament 3 painted an interesting picture, as there was a significant difference in performance between the fastest and the slowest processors tested. Once again the Core 2 Quad Q9650 was the king here, delivering 132fps at its default clock frequency of 3.0GHz, which was impressive when compared to the E6850 for example, which managed just 118fps at the same frequency. The best Phenom X4 processor was of course the 9950, which averaged 113fps, edging out the old E6700 by 2fps.

Again the Radeon HD 4870 X2 limits were not seen, as the results continued to improve with the processors clock speed. While we always suspected that the world’s fastest graphics card would require the world’s fastest processor, we know now for certain that it does. Those with lower-end Core 2 Duo/Quad processors can squeeze a great deal of performance out of the Radeon HD 4870 X2 simply by overclocking their processor past 3.0GHz, which is well within the limits of these processors.

Due to a lack of raw clock speed, the Phenom X4 processors did fall short in most tests, and even at 3.0GHz the 9950 struggled with the competition. Given that all tests were conducted at 1920x1200 with 8xAA/16xAF enabled, the results were impressive, and while the Phenom X4 processors did deliver acceptable performance, processors such as the Core 2 Quad Q9650 proved that the Radeon HD 4870 X2 still has much more potential just waiting to be unleashed.
Link: Legion Hardware
 
Will DDR2 to DDR3 difference doesnt affect the Scores ?!?! or

Is this Review valid for just CPU benchmarks ?
 
Fatal flaw: The tests are at 1920X, CPU bottlenecking is unavoidable with this card and that resolution.

Run the benchmarks at 2560x and then we'll see. I do know that the AT review spoke about this - the bottleneck is going to get worse with higher resolutions for some games, but the card will be at the appropriate load with a 4 megapixel display.
 
sangram said:
Fatal flaw: The tests are at 1920X, CPU bottlenecking is unavoidable with this card and that resolution.

Run the benchmarks at 2560x and then we'll see. I do know that the AT review spoke about this - the bottleneck is going to get worse with higher resolutions for some games, but the card will be at the appropriate load with a 4 megapixel display.

Completely agreed.

They are not trying to check the max available now.

Just check at the highest resolution for the maximum the 4870x2 can give.
 
sangram said:
Fatal flaw: The tests are at 1920X, CPU bottlenecking is unavoidable with this card and that resolution.

Run the benchmarks at 2560x and then we'll see. I do know that the AT review spoke about this - the bottleneck is going to get worse with higher resolutions for some games, but the card will be at the appropriate load with a 4 megapixel display.

Yes and another problem is most games these days can be maxed at whatever resolution, and they should have included crysis / Fear, which actually do stress the gpu to the limits. Any game having 100+ fps even at the highest resolution will never give a good idea about the cards power. In my TRI SLI testing only Crysis stood out to show the scaling even at my measly 2mp display.
 
Will this results hold good for Cards like 8800GT, 9600gt and 4850 ?

Then, better to invest in Q6600 with 9600gt/4850.
 
^^And what you lot seem to be forgetting is that most people don't have a resolution higher than 1920. So if that is your resolution (and it IS for most people) then the CPU is the bottleneck.

Benchmarks need to simulate the real world scenario, not just be an end unto themselves. Sure, as developers and architects you do all sorts of unrealistic tests, but that is to test the limits, not the common case. And the rule of design is optimize for the common case!

I agree that crysis should have been one of the tests though. It is a popular game.
 
^^No, unless you're testing at lower resolutions. At these test resolutions, those cards become the bottleneck.

The definition of a bottleneck is a slow component that affects your system, when it has components that can go faster. The bottleneck holds back performance.

That seems to be the case with this test. At 1920x the CPU is still bottlenecking the card (as is evident from the rapid increase in fps as the processor power increases), but at 2560x the card will become the bottleneck.

Crysis is a problem game, nobody knows exactly how it will scale with what hardware. Including Crytek, who refuse to divulge exactly what hardware will work with it. It's great that the tri-SLI config works with it, but most people simply don't have the budget or the power budget to run that kind of a configuration. I'm not sure it makes sense to invest in it for just one game.
 
^^ Well I know that the purpose of that test was to see how much today's CPUs compliment powerhouse like 4870X2, and it would have been more appropriate for folks at Legion hardware to "Also" test the card at 2560X res.

But the purpose of me posting this info and charts here was to show that the conventional wisdom that dual cores are better for gaming is not necessarily true. Quad cores (penryns) are in most instances either at par with dual cores (penryns) or better. It seems that finally quads are coming into their own and should be preferred over dual cores even if dual cores offer slightly higher clock speed.
 
Status
Not open for further replies.