phenom in company of heroes

[taltamir]

I was reading about ocability of the Q9450 to see if it is worth the price premium over the Q6600. And I stumbled on this:
http://www.overclockersclub.co...ews/intel_q9450/14.htm

In this test it shows not only that the company of heroes FPS is severely CPU limited, but that a phenom can, well, double anything an intel has.

Thoughts?

 

[Markfw]

double ? one page had them winning by 20-30% as the resolution went up, and the rest of the screens, it got killed.

And that site is really annoying, as ads keep popping up so I can't read the review.

 

[taltamir]

i know, that is my point, it gets killed everywhere... except it narrowly wins in call of juarez and completely annihilates the intels with more then double the FPS on company of heroes... Why?

 

[Markfw]

No.... Intel wins everything except opposing fronts. At 1900x1200 on one page, AMD Phenom had 38, and Intel had 23. On all other screens Intel dominated. What benchmarks are you reading ???

 

[taltamir]

Originally posted by: Markfw900
No.... Intel wins everything except opposing fronts. At 1900x1200 on one page, AMD Phenom had 38, and Intel had 23. On all other screens Intel dominated. What benchmarks are you reading ???

No what? you say no and then you agree with me.
Unless you are disagreeing with the call of Juarez one... which is a page before:
http://www.overclockersclub.co...ews/intel_q9450/13.htm

Opposing fronts IS company of heroes (the expansion), and is the page I linked.

38 fps vs 23 fps is a huge difference and is what I am talking about. why is intel QX, and various quads oced to a whopping 4ghz all getting the same 23 fps on opposing fronts, and the phenom getting 38fps?

 

[Markfw]

On the Linked benchmarks, the Phemon looses at all but the 1900x1200, and then its 14 vs 12, big deal...

So on ONE benchmark ON ONE game the Phmon wins, and looses everything else, so ??? What is the point ? That it won ONE benchmark ? In one game ? Big deal...

 

[Cookie Monster]

Maybe the "native quad" nature of the phenom is showing its benefits?

 

[OCGuy]

Quick! Sell your Intels and buy Phenoms! :laugh:

 

[SlowSpyder]

It is rather odd that in that game the Phenom, which is considered to be so much slower ends up being very much faster. Some of those benchmarks show odd results, look at the GPU test in Crysis. As the resolution goes up the Phenom gains fps in some situations, and ends up very slightly faster then the Intel offerings. WIC shows the Phenom to be a lot better then most people here would tell you it is too. If you take the results shown on that website as truth, then if you game on a 22" monitor or higher, all I take away from it is that the Phenom is as good as any other option. Not bad for a Phenom 9600, which is middle of the Phenom range.

 

[taltamir]

the big deal is that it is an anomaly. Something I did not expect. If the benchmarks are to be beleived, then the phenom is not universally slower, but possibly faster for specific tasks. I want to know WHY it is faster in those tasks. Is it the "native quad" thing? is it due to a pipeline difference, is it a mistake in the test, etc.
And also I want to know how many other games are affected that way.

So it boils down to the following questions:
1. Is this a mistake? Can we get someone else doing a comparison of that specific game and seeing if they too get similar results.
2. why are ALL the intel cpus showing the exact same FPS on that game at that resolution, and only 1 FPS difference on lower resolutions? I don't see why a 2.4ghz Q6600, a 2.5ghz Q9300, and a 4ghz Q9300 all get the same FPS there, while a phenom nearly doubles it, is there some specific intel issue with that game. This is very odd, I have a strong feeling that this is not a CPU but a platform issue (aka, the game is GPU bound and someone used more ram when testing the phenom or something). Again, this needs an impartial third party test to verify.
The call of juarez test is weirder still, sometimes the CPUs are the same, sometimes they are all different, depending on the individual test ran (aka, the resolution)
3. If the results really due to a phenom, how many other games does this occur in? which ones?
4. If the results are really due to the phenom CPU, WHY does it do it? which exact architectural difference accounts for this. They are supposed to have advantages in certain server type tasks, maybe some games generate similar tasks.

I think we should focus in this thread on answer questions 1 through 4 in this order (3 and 4 could be irrelevant depending on what the answer is to 1 and 2).

Until I saw this review I had no doubt that the phenoms are just completely uncompetitive in every way shape or form, now I am left wondering. I really don't care if the results favor Intel or AMD, I am just curious at this... discrepancy between what I thought to be true, and what I observed.

 

[Rhoxed]

i dont have either of the games, nor a 8800GT at my disposal, or i would love to help.

Though i do have 2 phenoms, and many other games if you need any benches from those

Supreme Commander
GRID
Crysis (xfire does not work...?)
COD 4
Mass Effect
Assasins Creed

and a cpl others

 

[Keysplayr]

I have a Phenom 9550 X4, and a Q6600. Unfortunately I don't have Company of Heroes, or its expansion. Otherwise I'd be testing it out.

 

[Chiropteran]

This is very interesting. Has anyone been able to dig up benchmarks from other sources that either support or disagree with these results?

 

[Sylvanas]

This has been reported before but it is usually dismissed or glossed over, Phenom has shown promise especially in UE3 based games Source and other heavily multithreaded scenarios. It can be seen that other games like Crysis are clearly only using 1 core (see the Task manager screenshot), and thus CPU frequency is king. It could be that Company of Heroes makes proper use of all 4 cores like UE3 and we are only just scratching the surface. Remember, Phenom performance is more than clockspeed, there is NB speed, HT frequency, Memory latency (more prominent due on On-Die memory controller) and Ganged or Unganged memory mode (2x 64bit bus width for multithreaded scenarios or 1x128bit for single threaded).

 

[heyheybooboo]

As noted the purpose of the article was to assess a q9450 (333 x 8) when OCed to 3.7GHz (463 x 8). Since that was the focus it's probably best to use caution when making other comparisons because that was not the intent of the author.

The Phenom results at 19x12 were interesting but as Sylvanas points out that opens the 'fsb / ht / imc' debate which is difficult because the results may be unique to that Asus mobo with the Asus 8800gt.

There were other results in the article that were less than consistent. The OCed q9450 was slower than the stock version at 16x10 in a few benchies; and in Bioshock there was no drop off in framerates between stock and OC at higher resolutions.

The author kinda implied that he spent a day tweaking on the 9450 OC. It could be that the game testing had a little less focus ...

 

[Martimus]

Phenoms have much better memory management, and as resolution goes up, games end up much more memory intensive. It only makes sense that it would do better at functions that require access to system memory, since that is its strong point. At lower resolutions most of the textures and whatnot can be held within the video card, so Intels processing advantage will win out there. Company of Heroes probably just has large textures compared to other games.

 

[Accord99]

Originally posted by: Sylvanas
This has been reported before but it is usually dismissed or glossed over, Phenom has shown promise especially in UE3 based games Source and other heavily multithreaded scenarios.

The LostCircuits UT3 test is flawed because it does not remove the FPS cap from the benchmark.

http://www.xtremesystems.org/f...p=2824458&postcount=17
http://www.xtremesystems.org/f...p=2825836&postcount=28

 

[nbowman]

Those results are ... strange ...
-look at the WiC bench: the phenom gets 28 FPS all the way down, which is actually better than the intel processors at 1920 rez.
-as the OP pointed out, the CoH results are all phenom.

do they list their testing method anywhere? I didn't see it.

 

[Phynaz]

Originally posted by: Martimus
Phenoms have much better memory management, and as resolution goes up, games end up much more memory intensive. It only makes sense that it would do better at functions that require access to system memory, since that is its strong point. At lower resolutions most of the textures and whatnot can be held within the video card, so Intels processing advantage will win out there. Company of Heroes probably just has large textures compared to other games.

If you are implying that Phenom would be significantly faster at a game in which the graphics data are not loaded in the graphics card, then I'm afraid you would be incorrect.

Once you get to the point of texture swapping the cpu is basically out of the picture.

 

[nyker96]

Not sure I believe these results. If I get to see the P9600s trump Q9xxx series on every site's benchmarks I'd believe this result. Right now somewhat doubtful it could be possible unless amd has a new core.

edit: did a quick google, looks like most other sites shows Q9xxx ahead even Q6xxx ahead in game benchmarks, didn't look at anything else but in games, every other site shows Q9xxx the winner. Anyways, this site might have a glitch somewhere when they ran the tests.

 

[Martimus]

Originally posted by: Phynaz

Originally posted by: Martimus
Phenoms have much better memory management, and as resolution goes up, games end up much more memory intensive. It only makes sense that it would do better at functions that require access to system memory, since that is its strong point. At lower resolutions most of the textures and whatnot can be held within the video card, so Intels processing advantage will win out there. Company of Heroes probably just has large textures compared to other games.

If you are implying that Phenom would be significantly faster at a game in which the graphics data are not loaded in the graphics card, then I'm afraid you would be incorrect.

Once you get to the point of texture swapping the cpu is basically out of the picture.

That doesn't make any sense. The GPU doesn't have direct access to the system memory. The Phenom is significantly slower than the Q9450 in processing power, but significantly faster with both memory bandwidth and latency (access time). Of course a system with a Phenom would be better at memory intensive tasks. But to say that this will overcome its processing speed deficiency in most applications would be ignorant. To say that it isn't a factor at all is also extremely ignorant.

edit: I think I might be coming off in an attacking tone, but I don't mean to do that. I just couldn't think of the proper words for what I was trying to say. I would personally rather have a Q9450 than any Phenom currently available, but I also know that the Intel chip has its limitations, just like anything else.

 

[Phynaz]

It makes perfect sense. The cpu involvement of sending texture to a video card is for all practicle purposes zero.

The loading and usage of texture by the video card is what takes all the time.

 

[Martimus]

Originally posted by: Phynaz
It makes perfect sense. The cpu involvement of sending texture to a video card is for all practicle purposes zero.

The loading and usage of texture by the video card is what takes all the time.

Except the latency is about 3 times as much for an Q9450 than a Phenom. So access times will be slower in the Intel system. The CPU/Northbridge still has to do all of the accessing of the system memory. The GPU does not directly access the system memory. Here is a more indepth article on how a GPU will access system memory.

 

[Fox5]

Originally posted by: SlowSpyder
It is rather odd that in that game the Phenom, which is considered to be so much slower ends up being very much faster. Some of those benchmarks show odd results, look at the GPU test in Crysis. As the resolution goes up the Phenom gains fps in some situations, and ends up very slightly faster then the Intel offerings. WIC shows the Phenom to be a lot better then most people here would tell you it is too. If you take the results shown on that website as truth, then if you game on a 22" monitor or higher, all I take away from it is that the Phenom is as good as any other option. Not bad for a Phenom 9600, which is middle of the Phenom range.

My guess:
The video card's memory bandwidth/memory size is being stressed to its limits, forcing more data over the PCI-Express bus. At this point, AMD's superior memory controller and bandwidth comes into play.

Heh, I kind of want to see a low end shoot out now. Of course, even low end cards have 128MB to 256MB worth of ram, but it'd be interesting to see if cards with lower amounts of ram could be stressed and give an edge to phenom.

 

[taltamir]

Fox5, interesting theory, by that theory using a 1GB vram video card would cause the performance ratios to flip...