Why does the 3.1GHz/R6970 1G iMac best the 3.4Ghz/R6970 2G iMac in Portal 2?

[glitch44]

Barefeats just did a GAMING SHOOTOUT:2011 iMacs vs others.

I'm curious why the Core i5 iMac got fractionally higher FPS than the Core i7 in Portal 2 at both 1920x1080 and 2560x1440?

1920x1080
iMac 3.4 = 125
iMac 3.1 = 127

2560x1440
iMac 3.4 = 83
iMac 3.1 = 84


iMac 3.4 R6970 = 'mid 2011' iMac 3.4GHz Core i7 with Radeon HD 6970M (2G)
iMac 3.1 R6970 = 'mid 2011' iMac 3.1GHz Core i5 with Radeon HD 6970M (1G)

If the game wasn't using the extra 1G of video ram, I'd still expect the fractionally higher FPS to go to the Core i7, or at least be even.

 

[5piN]

One word "Hyper-Threading" .....okay maybe two...

It's a well known fact that hyper-threading hurts fps in most games.... be interesting to see the result run with hyper-threading disabled on the i7

 

[spacepower7]

The extra 1GB of vram didn't help either

i7 was bto with 2GB vram
i5 was standard with 1GBvram

 

[kfscoll]

Barefeats just did a GAMING SHOOTOUT:2011 iMacs vs others.

I'm curious why the Core i5 iMac got fractionally higher FPS than the Core i7 in Portal 2 at both 1920x1080 and 2560x1440?

1920x1080
iMac 3.4 = 125
iMac 3.1 = 127

2560x1440
iMac 3.4 = 83
iMac 3.1 = 84


iMac 3.4 R6970 = 'mid 2011' iMac 3.4GHz Core i7 with Radeon HD 6970M (2G)
iMac 3.1 R6970 = 'mid 2011' iMac 3.1GHz Core i5 with Radeon HD 6970M (1G)

If the game wasn't using the extra 1G of video ram, I'd still expect the fractionally higher FPS to go to the Core i7, or at least be even.

The game is clearly GPU-bound. In other words, no matter how much more CPU power you throw at it, the GPU is maxed out and simply can't push any more pixels. Therefore the scores are virtually equal since the systems use the same GPUs and the extra VRAM isn't of any benefit with this game at these resolutions. The actual difference between the game scores is statistically insignificant.

I'd expect that most modern games are going to be GPU-bound on the new iMacs. This would likely be the case even if a single-GPU desktop card were used...at least if the games were run at the highest quality settings. Only in high-end multi-GPU systems will you see the CPU bottlenecking the video cards.

 

[5piN]

The game is clearly GPU-bound. In other words, no matter how much more CPU power you throw at it, the GPU is maxed out and simply can't push any more pixels. Therefore the scores are virtually equal since the systems use the same GPUs and the extra VRAM isn't of any benefit with this game at these resolutions. The actual difference between the game scores is statistically insignificant.

Valid point but i believe OP was wondering why the i5 3.1 scores marginally higher then the i7 3.4, which theoretically should have been the other way round.

http://vr-zone.com/articles/does-core-i7-hyper-threading-helps-/6160.html?doc=6160

 

[iSayuSay]

Maybe the test should be done more than once for the same computers, games and graphic settings

That way you can evaluate whether there were significant difference between two, we´re talking about statistic here, taking one data does not make it a valid information.

Could be the i7 was under more stress due to background apps or something. Even identical GPU installed on different computer can yield different result .. but close to each other. No computer is exactly identical, even with same spec on paper.

And i5 has been fast enough for any gaming today, any faster processor would be obsolete for this case. And so the i7 did. Once again, the test should be conducted frequent times to notice the difference

 

[sth]

1-2 fps is a normal margin of error in benchmark tests like these and can be attributed to many different factors.

The reason why the faster CPU makes no difference is because the tests are GPU-limited.

PS: Does anyone have the exact clock speeds Apple uses for the 9670M?

 

[barefeats]

Barefeats just did a GAMING SHOOTOUT:2011 iMacs vs others.

I'm curious why the Core i5 iMac got fractionally higher FPS than the Core i7 in Portal 2 at both 1920x1080 and 2560x1440?

Hi. It's BareFeats. It does seem counter intuitive but the difference is so minor it's academic. It does argue that if you are running games at high settings where they are GPU bound, the GPU is more important than the CPU.

The Core i7 flexes its muscles when you run a CPU intensive app like Cinebench or Geekbench where it's 39% faster than the Core i5 or LuxMark in "CPU only mode" where it's 49% faster.

 

[tsugaru]

Wirelessly posted (Mozilla/5.0 (iPhone; U; CPU iPhone OS 4_3_3 like Mac OS X; en-us) AppleWebKit/533.17.9 (KHTML, like Gecko) Version/5.0.2 Mobile/8J2 Safari/6533.18.5)

I would normally say that the results are within reason, but I'm thinking there's something afoot here.

A - i7 is about 9% faster clockwise
B - i7 has 8MB of L3 cache, vs 6MB in the i5
C - i7 has double the VRAM, which apparently is the only difference (same GPU clocks, shaders etc)

I would retest this. I just got my i7 yesterday. Once I get all my stuff front my older iMac on here, I'd like to test this personally.

 

[glitch44]

Hi. It's BareFeats. It does seem counter intuitive but the difference is so minor it's academic. It does argue that if you are running games at high settings where they are GPU bound, the GPU is more important than the CPU.

The Core i7 flexes its muscles when you run a CPU intensive app like Cinebench or Geekbench where it's 39% faster than the Core i5 or LuxMark in "CPU only mode" where it's 49% faster.

Thanks for responding, Barefeats! Didn't know you were in the Macrumor forums.

 

[50548]

Hi. It's BareFeats. It does seem counter intuitive but the difference is so minor it's academic. It does argue that if you are running games at high settings where they are GPU bound, the GPU is more important than the CPU.

The Core i7 flexes its muscles when you run a CPU intensive app like Cinebench or Geekbench where it's 39% faster than the Core i5 or LuxMark in "CPU only mode" where it's 49% faster.

When is BareFeats testing the new iMacs with some other games? In particular, I'd like to see tests with titles that may push the 1GB VRAM limit...of course Mac titles, not PC games...

 

[tsugaru]

Wirelessly posted (Mozilla/5.0 (iPhone; U; CPU iPhone OS 4_3_3 like Mac OS X; en-us) AppleWebKit/533.17.9 (KHTML, like Gecko) Version/5.0.2 Mobile/8J2 Safari/6533.18.5)

I'd go as far to say that their testing was flawed. At least in this case. There is no justification for how the i5 bests the i7, GPU bound or not, the i7 is faster and has more cache.