CPUs & Real-world Gameplay Scaling

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,634
CPUs & Real-world Gameplay Scaling - What do different video cards and CPUs mean to gamers in World of Warcraft, F.E.A.R., Battlefield F2, and Call of Duty 2? We played all the games with 6 different CPUs and 4 different video cards to find out the difference in real-world gameplay.

We have to give a nudge in the direction of the AMD Athlon 64 line of processors in real-world gaming. Obviously the differences we have seen here are not as exaggerated as those you might see by using canned and synthetic benchmarks. In fact real-world gaming shows a different story than will be told by benchmarks alone. Yes, both Intel and AMD make good processors for playing games, and don’t let anyone else tell you otherwise. We have also done incredible amounts of gaming on OEM systems that are powered by both Intel and AMD CPUs in our [H] Consumer desktop system evaluations that will show you both Intel and AMD are good for gaming. So our Blue Vs. Green story told here is hardly fueled by only one month of gameplay.

Please first read our article and discuss it here. Be aware that any trolling or flamebaiting will get you banned. So leave the fan-boy comments at home please.
 
Great job on the article. One thing I wish had been added is CS:Source. It is one of the most popular online games, and is extremely CPU bound. This becomes especially apparent with nicer video cards, they are held back signiifcantly even by 2.4ghz Venice.

Other than that, good article.
 
I just took a quick skim across, but so far this looks like some really valuable information. I'm going to read about this more in-depth pretty soon. Of course, since I'm a WoW addict, I focused more on what was said about it.

I'd like to let Brent and Kyle know that this FPS limitation on dual-core CPUs does not seem to be apparent anymore. I think it might have have been a driver related issue, but before upgrading to 84.21, it the framerate never went much beyond 60 FPS, even at the log-on screen. Now I'm experiencing a max of 85 FPS (I have vsync on with a refresh of 85 Hz) in less intensive areas such as the log-on screen, character selection screen, looking at a wall, etc. (hey, only got one 7800 GT :p).

Can anyone else verify?
 
Thanks for the kind words. CS:S was kicked around for a bit, but was cut just simply due to the time it was taking to get this done. It took over a month to compile all that data and put it in a slightly reasonable format.
 
Yoshiyuki Blade said:
I just took a quick skim across, but so far this looks like some really valuable information. I'm going to read about this more in-depth pretty soon. Of course, since I'm a WoW addict, I focused more on what was said about it.

I'd like to let Brent and Kyle know that this FPS limitation on dual-core CPUs does not seem to be apparent anymore. I think it might have have been a driver related issue, but before upgrading to 84.21, it the framerate never went much beyond 60 FPS, even at the log-on screen. Now I'm experiencing a max of 85 FPS (I have vsync on with a refresh of 85 Hz) in less intensive areas such as the log-on screen, character selection screen, looking at a wall, etc. (hey, only got one 7800 GT :p).

Can anyone else verify?

Good to know! Will look for other feedback. And yes you are correct, there was a driver release after our testing was complete and it was just simply too time intensive to run all of the results over. I still think we show what needs to be shown though in relation to dual core.

Brent gets the BIG KUDOS for keeping his nose to the grindstone!
 
Damm fine read!

I think a lower end CPU in the mix may have been worth the x-tra time in building the reivew.. but you guys are writeing it and not me... ;-)

(like a 512kb cache CPU of some sort, 3500+ mabey

EDIT: [/hugs 2.9Ghz 1mb cache 3700+]
 
I(illa Bee said:
Damm fine read!

I think a lower end CPU in the mix may have been worth the x-tra time in building the reivew.. but you guys are writeing it and not me... ;-)

(like a 512kb cache CPU of some sort, 3500+ mabey

EDIT: [/hugs 2.9Ghz 1mb cache 3700+]

Thanks!

Yeah, that was kicked around too, and maybe looking back if I had to do it again I would scale them a bit more. Then again, AMD is bunching their product line up pretty well too! The conclusions I will stand behind though.

Looking into caches, in real gaming, it virtually makes no difference overall at this point in time with the games we tested. Spend your money on GHz, and if you happen to get an extra 512KB of cache with the CPU, call it a bonus! ;)

Your 3700+ above is more than enough CPU for just about any video card setup you want to run. Congrats on that OC. Sweetness.
 
Hmmm... I think I found something interesting about the problem with WoW and dual-core. By selecting "Hardware Cursor" AND "Smooth Mouse" in the video options, I noticed a cap of 64 FPS. By disabling "Smooth Mouse", the fps shot back up to the refresh of the monitor (again, I'm either looking at a wall or at the log-on screen so nothing bogs down the framerate).

Perhaps it wasn't a driver issue at all... but the damned "smooth mouse" option! :p
Was that option selected during the evaluation?
 
Now...maybe I breezed through this article a bit too quickly or something...but what about Quake 4 and the 1.5 patch? There is a huge performance win with dual core (r_enableSMP 1)...not sure why this was not included...it would definately offset the conclusion of "There just simply aren’t any major 3D game titles that utilize dual core processors in any way that is advantageous." which is simply not true.

hifi
 
Interesting, I am actually a bit surprised that only a 3.2GHz P4 is required to avoid CPU bottlenecking with a single vid card. Until now I believed that a hi-end card such as the 7800GTX 512 would leave a 570/670 (3.8GHz) trailing behind.

How long will it take before dual-cores have a clear advantage over single-cores in gaming? Will probably wait until next year before I get my first dual-core system.
 
What's the point of testing single vs dual core on two games that don't seem to be multithreaded - why didn't you test quake 4 with the patch which has been shown to have large performance improvements with dual core?

Interesting that the multi-core aware nvidia and ati drivers didn't improve performance at all as that's not what previous reviews have shown. Does this functionality have to be enabled? Did you enable it?
 
The is the article of the year without exaggeration. So much for dual core CPUs that all vendors are trying to push onto us. No benefits whatsoever , all you get is a big electricity bill and less overclockable system. When upgrade time comes I'm definitely buying single core CPU. As far as I know no game company is developing a game that can benefit from dual core CPU's , because it makes the programming logic much harder and serves only a small percent of customers. We have a similar situation with XBox360 - it has several CPUs and no games that use more than 1 of them.
 
This definitely answers ALOT of question I have been looking for answers for. I always wondered if the newer GPU's would more then make up for a CPU..

It looks like if I save a little money on the CPU I can get a GPU and get much better performance for it.

Its kinda funny how there is little to no difference between the crazy expensive FX-60 and a x2 3800 in some cases.. Yes I know there ARE changes in the type of AA, but to me that doesn't look like a huge difference.

and of course, these numbers might be lower then what we can currently buy. The 7900 series cards are better, even if by a little-bit.

I guess we wont see an X900XTX cpu comparison, or maybe an Oblivion benchmarky?
 
Nice, now I have a review to point to for all the !!!!!!s saying you "have" to have dual core. People think playing games and having IM and Winamp running in the background needs dual core. :rolleyes:
 
Wow, great article. I was looking for something like this for a while to see real world performance with current processors. Great read and my hat goes off to Kyle and Brent for the hard work.
 
hifi said:
Now...maybe I breezed through this article a bit too quickly or something...but what about Quake 4 and the 1.5 patch? There is a huge performance win with dual core (r_enableSMP 1)...not sure why this was not included...it would definately offset the conclusion of "There just simply aren’t any major 3D game titles that utilize dual core processors in any way that is advantageous." which is simply not true.

hifi

Thanks for your thoughts. Sorry we did not cover exactly what you see as pertinent. The scale of top selling games of 2005 provided a framework for what games we used. Q4 was no where near the top.
 
Zap said:
Nice, now I have a review to point to for all the !!!!!!s saying you "have" to have dual core. People think playing games and having IM and Winamp running in the background needs dual core. :rolleyes:

There are definitely some real-world benefits to dual core processors, it is just that currently none of them are seen in current gaming...unless you are heavily multitasking as you hinted. AMD going to dual core with their Athlon X2 series brought me back to using an AMD-based machine. The same as Intel's HyperThreading kept an Intel CPU in my box for years. I know that most of the time that I don't come to actually utilizing all the power on hand, but the dual cores certainly allow for a smoother computing experience when you have a multitude of desktop applications running all the time. Smooth & Creamy Computing....or SCC. ;)
 
Wow, I've been waiting for something like this for a while. Nice read, should come in very much handy come my next upgrade.

On a sidenote, it's interesting to see that the performance difference between AMD and Intel has narrowed very much. I'm guessing this is because with today's games the limiting factor is pretty much the GPU (as can be seen in your FEAR benchies - faster processors provided absoloutely no increase in performance). Quite interesting to see this happening, I somewhat expected that with the 7800/X1800 series the bottleneck would once again shift to the CPU (ala the early 6800/X800 days).

Just curious, how come you used the 800 series of dual-cores, not the 900s? I'm curious to see how the extra cache of the 900s help...
 
InorganicMatter said:
Just curious, how come you used the 800 series of dual-cores, not the 900s? I'm curious to see how the extra cache of the 900s help...

Thanks!

Trying to keep this "real-world" was a major focus instead of making it a "technology expo" for Intel or AMD. I think more readers will be served by the choices we made.

Honestly, some of this stuff we sat around and talked about for a LONG time. This was such a massive undertaking in terms of man-hours that we just had to make some hard decisions to limit in certain areas. I wanted to have this published in December, but it was truly hard to schedule something like this. (I think it is a great story for the week of GDC though.) I wish we would have had the ability to do 12 processors and 10 video cards, but with this real-world gameplay testing, which is just not feasible for a webpage our size.
 
Thanks for the article, and the time it took to put all this together. I am always on the lookout for another excellent [H]ard|OCP writeup. But even with as much information as posted, the answer to a few questions I have is not there. Since I am in-process of building my new rig now, I will be spending plenty of time testing all my OC potential... very similar to the processor scaling you have completed. Hopefully I can answer a few of the questions myself over the next few weeks.

The questions missed:
  1. EQ2 shows as 100% cpu all the time in task manager. Is there a large benefit from CPU mhz increases? Dual-core? (of course many will have a game that was missed)
  2. With all of the recent discusions on the forums of physics in game, are some graphics-quality features CPU limited? I do not mean general features (Balanced or High) but more like enabling active terrain interaction, or water effects.
  3. How does the CPU scaling affect SLI rigs? Or, does SLI help remove a GPU bottleneck so that you *can* benefit from faster CPU's?
  4. What difference is there in real world gaming of changing only the memory speed and timing? DDR 400mhz (2-3-2-6) vs 500mhz (3-4-4-8) I know, this was a CPU speed comparison; OC'ing is all about CPU speed but memory speed follows.

Thanks again!

Not in my sig since it is not yet built, but am in process of building an Oppteron 165, dual 7900gtx, raptor 150 system. Should be a good system to test perform of processor overclocks since the GPU might not be holding me back much. On the other hand, I play at 1920x1200 widescreen, so might still be the big squeeze.
 
Thanks for a great article and for finally putting light on some of the dual core problems. I have had the same issues with my dual core setup and about ready to put my single core chip back in. Did you apply the Microsoft dual core hotfix in your systems?
 
Diseaseboy said:
Thanks for a great article and for finally putting light on some of the dual core problems. I have had the same issues with my dual core setup and about ready to put my single core chip back in. Did you apply the Microsoft dual core hotfix in your systems?
Yeah good point. Did you apply the M$ hotfix, and install the AMD driver?
 
Fantastic article. I wish I could stop all the "what about XX game and XX video card and XX CPU?!?! wtfz0r!!!!111" that will inevitably follow these posts.
Having said that, I truly didn't expect much CPU limitation with the game titles selected. They simply are known to be primarily GPU bound in almost all cases. However, I suspect that Quake4 may show better CPU scaling, HL2 has shown to be CPU intensive, and I wonder if RTS titles would scale with CPUs better as well. My game of choice lately is Empire at War, and I have Battle for Middle Earth 2 sitting on my hard drive waiting for me to finish EAW. EAW has been a real resource hog on my 3.4c, even with video details turned down. It would be interesting to see an article like this attack some of the games with reputations of being CPU smashers.
 
InorganicMatter said:
Yeah good point. Did you apply the M$ hotfix, and install the AMD driver?


I am going to let Brent answer this as he had all the hands on. That said, we have been going around with the MSHotfix and AMD drivers at [H] Consumer now for months and months and to put it succintly, they have never fixed anything for us. :(
 
Would it be possible to compile a numerical graph that compares all the results from all the games on one page? It's great that you guys undertook all the work that you did to bear out these results, I'm just wondering if you might be able to take one more step to make the information even a bit more accessible.

Great article. I love this sort of article, because it always saves me money on CPU upgrades! I guess my Winchester will be around at least until AM2 is well established and hopefully performing better than 939. Or maybe I'll go over to the blue team, but either way this little Winnie has still got legs and the [ H ] just proved it.
 
jebo_4jc said:
Having said that, I truly didn't expect much CPU limitation with the game titles selected. They simply are known to be primarily GPU bound in almost all cases.

Maybe they are simply known (or from what you have been exposed to on almost every other "video card site" on the net) to be primarily GPU bound in what is shown by benchmarks not real gameplay?

I know some of you guys think I am nuts about this, but I am telling you that benchmarks and timedemos are no longer reliable as a true measuring stick outside of a metric that is easy to get for making pretty graphs on a tech website.
 
superkdogg said:
Would it be possible to compile a numerical graph that compares all the results from all the games on one page? It's great that you guys undertook all the work that you did to bear out these results, I'm just wondering if you might be able to take one more step to make the information even a bit more accessible.

Great article. I love this sort of article, because it always saves me money on CPU upgrades! I guess my Winchester will be around at least until AM2 is well established and hopefully performing better than 939. Or maybe I'll go over to the blue team, but either way this little Winnie has still got legs and the [ H ] just proved it.


Feel free to do that for yourself and consider this permission to repost it or publish it. If you do, please let us know so we can link it up. :)

Thanks for the kind words, we were hoping this article would truly mean a lot to our readers in helping them make a good upgrade/buying decision.
 
Excellent Read. Very helpful information for the high-end, mid-range, or low-end performnace consumer. Good work Brent.
 
Wow great read, Kyle and Brent!

I was at a crossroads a little while ago when deciding between getting an X2 3800+ or a second 7800GT CO SE. I chose the GT when Monarch had it on sale for $240 after the rebate. I mostly game on my system, so I was happy with the performance gain the second GT offered. But still, there was that "what if" thought regarding the dual core, but that has been happily curbed thanks to this article. I can now say with 100% relief that a dual core won't be residing in my system until at least tax time next year, and now I am even that much more happy with the purchasing decision I recently made. Keep up the great work!
 
riblet said:
Thanks for the article, and the time it took to put all this together. I am always on the lookout for another excellent [H]ard|OCP writeup. But even with as much information as posted, the answer to a few questions I have is not there. Since I am in-process of building my new rig now, I will be spending plenty of time testing all my OC potential... very similar to the processor scaling you have completed. Hopefully I can answer a few of the questions myself over the next few weeks.

Thanks! Sure, I'll see what I can answer for ya

[*]EQ2 shows as 100% cpu all the time in task manager. Is there a large benefit from CPU mhz increases? Dual-core? (of course many will have a game that was missed)

Other games show as 100% in task manager as well, EQ2 in fact most games do.

We did not specifically test EQ2 single core vsl dual core so I don't have information on that, I would make an educated *guess* there will be no difference between single and dual core and that frequency will matter most as it did with the other games.

[*]With all of the recent discusions on the forums of physics in game, are some graphics-quality features CPU limited? I do not mean general features (Balanced or High) but more like enabling active terrain interaction, or water effects.

Anything that will be accelerated by the GPU will be GPU limited. Things like NPCs, AI, Physics will be CPU limited, and also if the graphics just aren't that intense that the GPU is basically twiddling its fingers, then it will be CPU limited.

[*]How does the CPU scaling affect SLI rigs? Or, does SLI help remove a GPU bottleneck so that you *can* benefit from faster CPU's?

SLI will remove GPU limitations so that you are more CPU bound, though we did not do any direct testing of this.

[*]What difference is there in real world gaming of changing only the memory speed and timing? DDR 400mhz (2-3-2-6) vs 500mhz (3-4-4-8) I know, this was a CPU speed comparison; OC'ing is all about CPU speed but memory speed follows.

None, notice our P4 3.4EE had higher clocked memory and bus, no difference.

I play at 1920x1200 widescreen, so might still be the big squeeze.

At 1920x1200 you will be more GPU limited.
 
Diseaseboy said:
Thanks for a great article and for finally putting light on some of the dual core problems. I have had the same issues with my dual core setup and about ready to put my single core chip back in. Did you apply the Microsoft dual core hotfix in your systems?

Can you point me to this hot fix? I did not specifically go to MS's site and download a specific hotfix.

I did however have all Windows Updates installed.
 
Yoshiyuki Blade said:
Hmmm... I think I found something interesting about the problem with WoW and dual-core. By selecting "Hardware Cursor" AND "Smooth Mouse" in the video options, I noticed a cap of 64 FPS. By disabling "Smooth Mouse", the fps shot back up to the refresh of the monitor (again, I'm either looking at a wall or at the log-on screen so nothing bogs down the framerate).

Perhaps it wasn't a driver issue at all... but the damned "smooth mouse" option! :p
Was that option selected during the evaluation?

Yes, Smooth Mouse was enabled in all tests, I'll check that out.
 
Diseaseboy said:
http://support.microsoft.com/?id=896256

This corrected my speedup/slowdowns in GTA:SA. Make sure you update to the latest dual core processor drivers from AMD too.

We have had little luck with those fixes across a myriad of games. So don't expect the HotFix to be too hot or AMD drivers to fix a ton of game implementation issues. :(
 
Thanks for the review... it definately helped me decide what I want in my new rigs (Wifey and mine.) It's just the AXP's are getting long in tooth, and the games I play (16x12, 0xAA, 4xAF on mine, 12x10 0xAA, 0xAF hers,) we typically see numbers in my teens.

Knowing what's the baseline in the LAN and Online games we actually play (and you nailed them 4 out of 5,) gives me an idea what to spend my hard earned money on. Kudos to Kyle & Brent.
 
Another excellent read,

Makes me feel confident that with my current 7800GT/Opteron165 I will experience fantastic gameplay for at least another year. I always get disheartened when a new line of <insert computer components here> comes out and that new rig you just dished out $2K for seems obsolete.

As previously stated it would be nice to see some SMP games put in the mix like Quake 4, or even the mmo City of Villains/Heros.

Of course there is always OC'ing to further extend the future of your components :)
 
Sure spells it out. Games are not written for Dual Core CPU's. I built Optern 175 with 7800GT for Wife to replace P4 3.2 & 9600 Pro. She still quite happy. Word opens faster than ever. Ok she plays some games too (Civ 4, C&C Generals).

I made sure it had 2g of Ram. I think that was the biggest impact.

I would like to see an CS2 (Photoshop CS2) dual vs single core. Any Links?

DM
 
Back
Top