Choosing the right CPU for your GPU(s)

K

kevineugenius

[H]ard|Gawd
Joined
Dec 9, 2006
Messages
1,415
My experience with overpowered CPUs has been pathetic. They never deliver performance gains that I expected. I was reading through an ATI 3870x2 vs Nv 9800gx2 review and it was mentioned that the CPU wasn't beefy enough to deliver performance in some of the games they were testing. I have seen numerous people hiding behind the term "bottleneck" with reference to their CPU, claiming that the CPU is not going to allow the graphics cards to work well enough. In this instance, the CPU in question was a Phenom 2.6ghz, which I understand isn't the fastest critter on the planet, but is it really going to cause a bottleneck in the GPU?

My understanding is that the CPU doesn't handle the same processes of information that the GPU handles... that's the point of having them separate. If I suddenly turn on 4x AA in a game where there was no AA before, is it going to increase the CPU load by a substantial amount?

In short: I never over-spend on CPUs. I overspend on GPUs. Is my philosophy wrong? If you're going to say I'm wrong, please back it up with hard data, or links to the data. I've done a few tests myself, such as overclocking an e6600 by 33% which only gave me a 2% framerate increase in my CS:S stress test. I'm looking for the truth of the matter, not opinions.
 
According to Nvidia you are doing the correct thing. But there is something to be said for having too much gpu power in comparison to your cpu power. Alot of the cpu bottlenecking depends on the resolution you run also. At 1920x1200 and up, the cpu becomes less of a concern in comparison to the gpu.
 
I can't remember where I saw the test. Maybe it was on Toms Hardware :)rolleyes:) but they showed that a balanced system was still slower than an unbalanced system. For example a X2 6000+ with a 3870 was slower than a X2 4400 with a 8800GTX. I always put more stock into the GPU than the CPU.
 
kevineugenius said:
My experience with overpowered CPUs has been pathetic. They never deliver performance gains that I expected. I was reading through an ATI 3870x2 vs Nv 9800gx2 review and it was mentioned that the CPU wasn't beefy enough to deliver performance in some of the games they were testing. I have seen numerous people hiding behind the term "bottleneck" with reference to their CPU, claiming that the CPU is not going to allow the graphics cards to work well enough. In this instance, the CPU in question was a Phenom 2.6ghz, which I understand isn't the fastest critter on the planet, but is it really going to cause a bottleneck in the GPU?
Click to expand...
Not likely, probably plenty of CPU performance.

kevineugenius said:
My understanding is that the CPU doesn't handle the same processes of information that the GPU handles... that's the point of having them separate. If I suddenly turn on 4x AA in a game where there was no AA before, is it going to increase the CPU load by a substantial amount?
Click to expand...
Correct, that should have little to no effect on CPU.

kevineugenius said:
In short: I never over-spend on CPUs. I overspend on GPUs. Is my philosophy wrong? If you're going to say I'm wrong, please back it up with hard data, or links to the data. I've done a few tests myself, such as overclocking an e6600 by 33% which only gave me a 2% framerate increase in my CS:S stress test. I'm looking for the truth of the matter, not opinions.
Click to expand...
Correct, I don't even bother overclocking my E6600, I know it's not my bottleneck. I leave it running cool and stock with low quiet fan speeds :)

While in princible what you say is a good general rule, there is a small handful of games like SupCom that are CPU hungry, multicore cpu hungry. Is this the future trend? *shrug*
 
It depends on the game / simulator as well. For example, FSX shows little gain with a more powerful GPU (assuming you have a decent GPU to begin with), but OCing the CPU or running a faster CPU/more cores can yield a significant performance increase.

Since I spend alot of time in FSX, I am probably going to move from my E4300 to a E8400 in the near future and pass over the GPU upgrade since that will probably show the best gain for my particular needs.
 
Couple of googlings: Prey is Doom3/Quake4 engine right? Fairly modern.
http://www.legionhardware.com/document.php?id=603&p=4
1.59ghz Core2 vs. 3.33Ghz Core2
FPS delta of 1.4 FPS

http://www.bit-tech.net/gaming/2007/08/30/bioshock_gameplay_graphics_and_performance/10
More modern game. Bioshock, UT3 engine, 1920x1200
2.13ghz Core 2 vs. 3.0ghz quad
.4fps difference in average
2fps difference in min

But like I said before, there are a handful of games like SupCom that do need a lot of CPU. So it's in your best interest to have somewhat a balanced system.
 
Do you mean I don't have to replace my Q6600@3.2GHz when Nehalem comes out to rock some games?
 
Menelmarar said:
Couple of googlings: Prey is Doom3/Quake4 engine right? Fairly modern.
http://www.legionhardware.com/document.php?id=603&p=4
1.59ghz Core2 vs. 3.33Ghz Core2
FPS delta of 1.4 FPS

http://www.bit-tech.net/gaming/2007/08/30/bioshock_gameplay_graphics_and_performance/10
More modern game. Bioshock, UT3 engine, 1920x1200
2.13ghz Core 2 vs. 3.0ghz quad
.4fps difference in average
2fps difference in min

But like I said before, there are a handful of games like SupCom that do need a lot of CPU. So it's in your best interest to have somewhat a balanced system.
Click to expand...

+1 to this post. CPU doesn't matter at all at high resolutions when you reach a certain point. That point seems to range from a 2GHz to a 2.4GHz Core 2, dual or quad doesn't matter so much. And like he said, the exception is simulator and strategy games that do a lot of CPU math that have nothing to do with rendering.

The one thing to note about SupCom is that beyond the "frame rate" which everyone tests, theres a simulation rate behind the scenes. You can be getting a great framerate but the simulation rate may be bogged down, especially with a lot of AIs in a game. There are mods which can show your simulation rate if I recall.
 
Legion and Bi-Tech are fairly reputable sites, what they say i have seen before on others too.

This is a conversation that i remember from when Conroe came out.
here was one little twist to it.

How many of you run an Antivirus? a Firewall? an IM client? a browser? etc...
Benchmarks dont really capture these, this is where more ram and a faster CPU play.
remember that windows can simultaneously run many apps, so the more you have in your task manager, the more of a load youre putting on your CPU, and the less meaningful those benchies become.
 
I just never go all out and get top of the line stuff just a waste of dolla.

I have never seen a CPU be a true bottle neck in gaming becuase a true bottle neck stops all performance increases. Most of the time if you were to buy a better card you would get better results regardless of your cpu.

If you go from a 7800gt to a 8800gt and get a 0 fps boost then you would have a true bottle neck and I have never seen it happen. I think you would need like a p2 or lower.
 
allow me to pose my scenario. I recently built a new PC with the following:

q6600
8800gts 512 (g92)
8 gigs of ram
150 raptor

which replaced what's currently in my sig. which btw the 8800gts in the sig is the old 640mb version

playing EVE online side by side my new build gets at least double the FPS of my old build. as far as I know my new card is the equiv of a 8800gtx roughly. the old card would normally get about what 10 to 15 fps less than a gtx? so with a new cpu and wider memory bus I get double the frames with a minor vid update. old = about 30 fps, my new pc gets about 80 fps.

what would you say I'm missing then if what you guys are saying is true?
 
This is the main reason I'm not a big fan of overclocking. It doen't necessarily add much performance for the potential problemsThis is the main reason I'm not a big overclocker. It doesn’t necessarily add much performance for the potential problems. I’d much rather have greater stability over a little performance.
 
heatlesssun said:
This is the main reason I'm not a big fan of overclocking. It doen't necessarily add much performance for the potential problemsThis is the main reason I'm not a big overclocker. It doesn’t necessarily add much performance for the potential problems. I’d much rather have greater stability over a little performance.
Click to expand...

Oh, don't misunderstand, a faster CPU is great for a lot of things, but gaming isn't one of them. Also I'm not sure why you think instability is inherent to overclocking. The first thing anyone does when overclocking is stability test the thing using the more torturing and sensitive applications there are.
 
heatlesssun said:
This is the main reason I'm not a big fan of overclocking. It doen't necessarily add much performance for the potential problemsThis is the main reason I'm not a big overclocker. It doesn’t necessarily add much performance for the potential problems. I’d much rather have greater stability over a little performance.
Click to expand...

hmmm in the intel forums there is a list of all the C2D chips that have been OCed.
some high percent there. A 5% OC is more problem than its worth, usually, but. (depends on the person) mine is 30%+ on the intel.

various benchmarks show benefits, real life benefit was also immediately seen.
 
Since having to downclock my oc'd Opty 165 (from 2.4 to 2.1) COD4 gets hitches in times of frantic activity (in mp), but mostly I still game, very well I might add, at 1920x1200 max settings (in game).

The more powerfull the cpu the less times (from my perspective) you'll experience sudden drop (system lag) in games.
 
I am more than happy with my Q6600, I expect it to last a long time. I am content with a "mild" OC to 3.0GHz, anything more is not worth it for my uses, including crysis. :)
 
CPU's are usually more important for ai intensive games like real-time strategy games or when scripted walls/buildings are envolved, but the CPU power required for those games is always very minor compared to the best CPU's available.

Graphic cards and speedy RAM are more important for smooth gameplay.
 
Megalomaniac said:
How many of you run an Antivirus? a Firewall? an IM client? a browser? etc...
Benchmarks dont really capture these, this is where more ram and a faster CPU play.
remember that windows can simultaneously run many apps, so the more you have in your task manager, the more of a load youre putting on your CPU, and the less meaningful those benchies become.
Click to expand...

Ah yeah, I forgot about this point. Basically a guy asked me why I don't recommend anyone (even enthusiasts) buy "Intel Extreme" processors, and I just didn't have enough numbers off the top of my head to satiate him.

GG all, thanks for the backup.
 
how come my p4 northwood at 3.2ghz cant handle max settings even if i stuck a 8800gts card(agp hypertheticly) in it?
 
jonneymendoza said:
how come my p4 northwood at 3.2ghz cant handle max settings even if i stuck a 8800gts card(agp hypertheticly) in it?
Click to expand...

Because that's equiv to like a 1.6GHz Core 2, which is far below the critical point at which CPU ceases to matter at high resolutions.

But really, max settings in what, exactly? And you seem to be talking hypothetically here, so maybe it could handle max settings in whatever you're talking about.
 
Syndicated_Death said:
I've still yet to hear anyones thoughts on my scenario.
Click to expand...

I looked at your post but theres not much to say. There is little performance data available for Eve Online out there, so its anybodys guess to how it scales. You did not state which GTS you started with, 320MB version, 640MB? Its not uncommon to see an up to 50% boost going from an old GTS to a new one, depending on the game. As to your scenario of almost tripling framerates, it seems unlikely. I don't see any reason why Eve would be very CPU bound. All the hard thinking is done server side, AI, calculation of data for charts and graphs, etc.
 
MMOs in general tend to be heavily CPU-reliant as well like RTS games, should explain the case with EVE Online (I have no idea if it's multithreaded though).
 
jonneymendoza said:
how come my p4 northwood at 3.2ghz cant handle max settings even if i stuck a 8800gts card(agp hypertheticly) in it?
Click to expand...

My Albatron PX915P4C mobo is socket 478 and a PCI-E x16 slot, so I may be able to see how my 3.06northwood handles one of my 8800GTX's.
 
Trombe said:
MMOs in general tend to be heavily CPU-reliant as well like RTS games, should explain the case with EVE Online (I have no idea if it's multithreaded though).
Click to expand...

I haven't really been able to locate any accurate benchmarks on MMOs. What do you theorize the reason is for the CPU reliance you have seen? They do not have to calculate AI or physics or anything. So I wonder what would be causing it.
 
NKDietrich said:
I looked at your post but theres not much to say. There is little performance data available for Eve Online out there, so its anybodys guess to how it scales. You did not state which GTS you started with, 320MB version, 640MB? Its not uncommon to see an up to 50% boost going from an old GTS to a new one, depending on the game. As to your scenario of almost tripling framerates, it seems unlikely. I don't see any reason why Eve would be very CPU bound. All the hard thinking is done server side, AI, calculation of data for charts and graphs, etc.
Click to expand...

I did state which gts I started out with... you just missed it... I said the older 640 version... Believe it or not, for a FACT that is what I'm getting in frame rates. 30 on the old setup and 80 on the new.
 
Trombe said:
MMOs in general tend to be heavily CPU-reliant as well like RTS games, should explain the case with EVE Online (I have no idea if it's multithreaded though).
Click to expand...

I had suspected that EVE was very cpu limited but I wasn't aware to what extent until I built the new computer. and as you said it is true of RTS games and MMos.. so with all of this in mind I think alot of the real question boils down to depending on what games you play what impact will the cpu have on my game play or even be a bottleneck.
 
Syndicated_Death said:
I had suspected that EVE was very cpu limited but I wasn't aware to what extent until I built the new computer. and as you said it is true of RTS games and MMos.. so with all of this in mind I think alot of the real question boils down to depending on what games you play what impact will the cpu have on my game play or even be a bottleneck.
Click to expand...

Don't get me wrong, I'm not saying you're lying, but I'm having a hard time fathoming why Eve would be that reliant on a CPU that you could get +150% framerate based on your CPU. That is on par with what you'd expect to see from that upgrade in a completely CPU-based renderer.

Are you fairly certain those framerates were taken from roughly equivalent situations? Because this intrigues me, I wish I could find more information, but no one seems to benchmark Eve.
 
The easiest test to find out if you are being cpu bottlenecked. Set you resolution to 1024x768. Then play a scene in a game. Then reset the resolution to 1280x1024. If you don't see a drop in the fps then your cpu is bottlenecking your graphics card. Prime example, I can increase the resolution on an older comp, A8NSLI Premium w. FX-55 and 2G of ram. playing FEAR and the fps stays with in a couple of frames until I go over 1680x1050 then the frame rates drop fast. That is why on my A8N32SLI Deluxe w/ opty 185 w/4G of ram or my M3A32 MVP Deluxe w/Phenom 9600 BE w/ 4G ram I have higher frame rates up till 1680 x 1050. Oh video card used in the test was a HIS 3870 IceQ3 and I had AA and AF turned off.
 
I think MMO's use a bit of cpu power because they decrypt the incoming packets and convert the data into each players direction and action (the gpu does the model rendering but the positioning and action state are cpu side), along with scripted npc's and monster ai.

It takes a bit of minimum cpu power to run RTS and MMO's, but once you have enough power to run it smoothly there is no benefit of having anything faster, aside from managing your background processes.
 
NKDietrich said:
Yeah and I'm not even counting the fact that it has two cores, since that doesn't really matter so much for games.
Click to expand...
That's where I might have to disagree. Most modern game engines are at least utilizing 2 cores pretty effectively. Maybe not 4 cores yet, Crysis makes use of 4 cores a little bit, so does SupCom. But neither one are fully multi-threaded enough to fully balance load across 4 cores IIRC the reviews. But definitely dual.
 
dcds1 said:
The easiest test to find out if you are being cpu bottlenecked. Set you resolution to 1024x768. Then play a scene in a game. Then reset the resolution to 1280x1024. If you don't see a drop in the fps then your cpu is bottlenecking your graphics card. Prime example, I can increase the resolution on an older comp, A8NSLI Premium w. FX-55 and 2G of ram. playing FEAR and the fps stays with in a couple of frames until I go over 1680x1050 then the frame rates drop fast. That is why on my A8N32SLI Deluxe w/ opty 185 w/4G of ram or my M3A32 MVP Deluxe w/Phenom 9600 BE w/ 4G ram I have higher frame rates up till 1680 x 1050. Oh video card used in the test was a HIS 3870 IceQ3 and I had AA and AF turned off.
Click to expand...

i did a similar test on cod4 on my p4 rig i mentioned earlier. at 1024 up to 1680x res teh fps stays wroughly the same but if i dip down800x600 its about 10fps higher.

i have a 7800 gainward bliss Overclocked and i feel that im not using the full power of it in games i play due to my cpu
 
You must log in or register to reply here.
Back
Top