nVidia 8 series and CPU scaling

[thilanliyan]

http://legionhardware.com/document.php?id=672

 

[MarcVenice]

Goes to show videocards come first, then cpu's. For playable framerates even the celeron and sempron were good enough, exept for x3. So if you're a gamer, you shouldn't spend more then 150$ or so, which almost buys you a x2 6000+ if you are also buying a 8800gts or 8800gtx. Anything more on the cpu and you should be looking to invest it in a better videocard instead.

 

[ShockwaveVT]

kinda wish they had included a 8800GTS 640MB, but overall a great look at the performance of many GPU/CPU combos

 

[aka1nas]

Would have been more useful if they had measured min framerates. Also, X3 is very CPU-intensive when playing the full game. I'm pretty sure they just ran the rolling demo benchmark like every other review site and basically just did a graphical test of the game. The full game is also simulating the other 120 or so sectors in the background and running any stock and user-added scripts as well.

 

[Piuc2020]

Yeah some games, especially RTS tax the CPU just as much as the GPU, run these same tests on Supreme Commander for instance and then we'll see.

 

[Keysplayr]

CPU isn't that important then? Hmm. Tell that to STALKER.

 

[lupi]

That's odd. When the 8800s first came out and were tested in SLI the reports said they were hitting a CPU blockade.

 

[ArchAngel777]

Originally posted by: keysplayr2003
CPU isn't that important then? Hmm. Tell that to STALKER.

Yeah, there are a few craptastic devs out there that think everyone can buy a new CPU every year.

The GPU makes more sense to tax rather than the CPU. When you replace a GPU, that is all you have to do. When you replace a CPU, often you need to replace the motherboard and memory that goes with it.

X3 and STALKER, IMO are bad examples. Besides, neither game interests me.

 

[Piuc2020]

Yeah I'm simply not buying those tests. Don't post flimsy articles until at least you get decent software to benchmark on... aside from Company of Heroes the rest were all craptacular choices for benchmarking, Far Cry? Come on my 9600SE/P4 computer ran Far Cry... the game was released in early 2004 for crying out loud. F.E.A.R. is an obviously GPU bound game... not much for the CPU to do in that game. I'll give X3 the benefit of doubt.

I suggest they get Oblivion, STALKER, Supreme Commander, Command & Conquer 3, Civilization IV, The Sims 2, Call of Juarez, etc and run tests on those games. If your website is so frigging poor the only game they can afford is Company of Heroes then please refrain from posting ****** and incomplete articles...

 

[RussianSensation]

Firingsquad 8800 cpu scaling:

AMD CPU scaling
Intel CPU Scaling

Almost any C2D processor is fast enough for 8800GTX at high resolutions. There is nothing new here. About the only game that really benefits from a faster cpu is Supreme Commander.

 

[NoStateofMind]

The Source engine is a very CPU intensive as well. At stock clocks my E6400 only pushes like 80FPS with a X1900XT. Contrary, when clocked at 3.2Ghz, I get 100-140FPS! CPU's are still important when choosing/making a gaming system IMO.

 

[myocardia]

Originally posted by: RussianSensation
Almost any C2D processor is fast enough for 8800GTX at high resolutions. There is nothing new here. About the only game that really benefits from a faster cpu is Supreme Commander.

You've obviously never used any flight simulators, specifically M$'s Flight Simulator X. Unless you think that a 7600 GT is really >85% as fast as an 8800 Ultra.

 

[n7]

This is nothing new.

The CPU has never been overly important.

Sure, something decent is requires, but GPU always makes a far larger difference, & considering how the average resolution being used keeps increasing, i don't see this changing anytime soon.

 

[Keysplayr]

Well, GPU's have certainly come a long way since the original Geforce. Touted to offload geometry calculations from the CPU. I guess that is happening more and more these days. But now there are many other things involved in newer games to keep the CPU busy. Increased AI would be one thing. I've always wondered what the primary role of the CPU is these days when it comes to gaming. How much of the geometry does it still handle? Does it run the game engines? Or is that handled by the GPU as well? Physics of course. Physics are becoming quite advanced. Painkiller was the first "blown away" physics game I can remember. I was simply amazed at the time.

 

[Piuc2020]

CPUs handle all the calculations that don't involve graphics at all, for example, a CPU will determine the amount of objects on screen, their size and their position and then pass them on to the GPU to render them. CPU are in charge of determining "what will happen next". The GPU is only in charge of rendering it all. If there are a million units fighting on screen, each with their own projectiles and abilities. The CPU has to calculate the trajectory of all those projectiles and the damage they do on a unit and what happens to that unit when it gets hits and it keeps track of the location and status of every unit. The GPU does is render those units and those projectiles so whether the characters are built with a single polygon and no shaders makes little difference to the CPU but lots of difference to the GPU.

It's possible to say that games today are not taxing the CPU enough and most CPUs can handle it and its true, advancement in games are only really advancing on the GPUs. RTS games are one of the few genres where CPUs are taxed heavily but even today RTS are still benefiting more from GPUs than CPUs. Physics calculations in current games are pathetically simple, everything is rigid and static and CPUs only have mundane gameplay calculations to worry about. This will all change with the upcoming physics revolution though, if Ageia Physx card pushes through however. If someone was to create a game that really gave a Core 2 Quad a workout by giving it a LOT of numbers to crunch, meaning millions (literally) of characters fighting on the same screen and with a helluva different properties and physics based combat for each then that game would be limited by the GPU being unable to render all those characters and stuff and no developer will make those characters look like N64 models just so the GPU can cope... do you understand me?

So the next big thing is going to be physics and an increment of things going on onscreen, CPUs will finally have a lot of number crunching to do, which is what they do best and with all the cores CPUs will be getting. We are really going to see much more dynamic and interesting games in a few years, even more so if the Ageia Physx card succeeds which I really hope it will.

 

[RussianSensation]

Originally posted by: myocardia

Originally posted by: RussianSensation
Almost any C2D processor is fast enough for 8800GTX at high resolutions. There is nothing new here. About the only game that really benefits from a faster cpu is Supreme Commander.

You've obviously never used any flight simulators, specifically M$'s Flight Simulator X. Unless you think that a 7600 GT is really >85% as fast as an 8800 Ultra.

I am not sure you linked the right graph because the game is far more GPU related since every ATI card is inferior to all current NV cards (I mean 6800 Ultra is as fast as X1950XTX....). But yes, 7600GT in that graph provides 85% of the performance of 8800Ultra, but both are getting <24fps? So if you want to play flight sims, you would first pick an NV card and then worry about the CPU Also this doesn't tell us how E6400 performs against say E6850.

LOMAC
1920x1200 4AA/16AF
8800GTX = 83.3fps
7900GS = 36.9fps (!!!)

Pacific Fighters
1920x1200 4AA/16AF
8800GTX = 55.6fps
7900GS = 27.2fps

You are getting more than double the performance with a faster graphics card. You are saying that you'd be more limited by the processor if you had E4400 in there?

 

[myocardia]

Originally posted by: RussianSensation

Originally posted by: myocardia

Originally posted by: RussianSensation
Almost any C2D processor is fast enough for 8800GTX at high resolutions. There is nothing new here. About the only game that really benefits from a faster cpu is Supreme Commander.

You've obviously never used any flight simulators, specifically M$'s Flight Simulator X. Unless you think that a 7600 GT is really >85% as fast as an 8800 Ultra.

I am not sure you linked the right graph because the game is far more GPU related since every ATI card is inferior to all current NV cards (I mean 6800 Ultra is as fast as X1950XTX....). But yes, 7600GT in that graph provides 85% of the performance of 8800Ultra, but both are getting <24fps?

Correct, because the speed of your video card makes nearly no difference, with FSX.

So if you want to play flight sims, you would first pick an NV card and then worry about the CPU Also this doesn't tell us how E6400 performs against say E6850.

Okay, you obviously know nothing about any gaming on computers, not just flight simulators. See, some games use an API that will show performance advantages with nVidia's cards, others use the API that benefits/plays faster with ATI's cards. That applies to every computer video game ever written, btw.

LOMAC
1920x1200 4AA/16AF
8800GTX = 83.3fps
7900GS = 36.9fps (!!!)

Pacific Fighters
1920x1200 4AA/16AF
8800GTX = 55.6fps
7900GS = 27.2fps

Those aren't flight simulators. They're games that have airplanes in them. There's a big difference. But, I do like how you chose only the resolution that's way above what a 7900GS can handle.

You are getting more than double the performance with a faster graphics card.

Yes you are, at 1920x1200, in gpu-limited games. This doesn't apply to flight simulators, and definitely not to the one that I mentioned.

You are saying that you'd be more limited by the processor if you had E4400 in there?

Actually, I'll go that one better. In Flight Simulator X, you'd get ~double the framerates with a C2D @ 4.0 Ghz, using a 7600 GS, as you would with an E4400, using an 8800 Ultra. <That's obviously using the same in-game settings. See, you really didn't know anything about flight simulators, did you?

 

[BenchZowner]

Let's add another one into the mix
BenchZone's GeForce 8800 Performance Scaling

 

[dug777]

Originally posted by: Piuc2020
until at least you get decent software to benchmark on... aside from Company of Heroes the rest were all craptacular choices for benchmarking, Far Cry? Come on my 9600SE/P4 computer ran Far Cry... the game

Your 9600SE would have run FarCry like a pile of utter crap

Even a 9800pro (which obliterated a 9600SE) could only manage 1024/2xAA/8XAF with everything cranked, and while i found that playable, it was probably averaging in the 30s, most people would have ditched the AA or started turning down settings.

 

[RussianSensation]

Originally posted by: myocardia

Okay, you obviously know nothing about any gaming on computers, not just flight simulators. See, some games use an API that will show performance advantages with nVidia's cards, others use the API that benefits/plays faster with ATI's cards. That applies to every computer video game ever written, btw.

You don't have to get all aggresive on me here. All I am doing is posting links to flying games and showing you that a graphics card matters more. I am not even disputing that NV is better in some games and ATI in others based on game engines. Long time ago I read that nvidia cards can do vertex prefetch or something along those lines that ATI cards cannot do in Pacific fighters which is why they are faster in that game. Also flight sims aren't shader intensive and are more texture limited due to vast land, sky, water and aircraft and ATI cards' main bottleneck has historically been its texture filtering relative to NVs. That still shows that a graphics card matters more because you'd rather pick an NV card and a slower cpu for pacific fighters for instance than the fastest cpu and an ATI card. But since you classified them as "flying" games and not flight sims, then benches for those games don't even matter to you; so let's move on.

Those aren't flight simulators. They're games that have airplanes in them. There's a big difference. But, I do like how you chose only the resolution that's way above what a 7900GS can handle.

If I had 8800GTX I'd play at the resolution that gives me playable framerates which 8800GTX can handle at 1920x1200 4AA/16AF. Why in the world would I play at 1024x768? Then I'd just get a 7900GS. 7900GS cannot handle those high resolutions and image settings for the above flying games because those 2 games are far more GPU dependent than CPU when comparing modern cpus (although they are cpu dependent as well, but GPU still matters more). If we were testing at 1024x768, the majority of games would then be CPU limited, but who plays at these resolutions exactly with modern GPUs?

Yes you are, at 1920x1200, in gpu-limited games. This doesn't apply to flight simulators, and definitely not to the one that I mentioned.

Ok if you want to get technical and call the games I posted "flying" games while your Microsoft flying game as a simulator, that's a mistake on my behalf. But how many people buy $500 graphics card and E6850 to play Microsoft flight sim only? The discussion in this thread is about gaming as a whole not about 2-3 games that 3% of the market plays. Top 3 genres for PC are FPS, massively multiplayer rpgs and strategy games. And then there is Microsoft Simulator that sells very well. But collectively, it probably took a lot of other flight games other than Microsoft's game to match sales of say Half Life 2. It's very easy to pinpoint 1-2 games to prove something like saying that X1950XTX cards are 2-3x faster than Nvidia 7900GTX when you enable 8AA. Well obviously if you want to pick extreme examples, you can show anything, but we are discussing gaming balance for a system on "average." You discounting 2 popular flight games only reduces the number of popular flying games and the reasonableness for investing even more $ into a faster cpu to play even fewer flight games, limited to hardcore simulators.

Actually, I'll go that one better. In Flight Simulator X, you'd get ~double the framerates with a C2D @ 4.0 Ghz, using a 7600 GS, as you would with an E4400, using an 8800 Ultra. <That's obviously using the same in-game settings. See, you really didn't know anything about flight simulators, did you?

Have you ever considered that flight sims' graphics engines are poorly coded? Ace Combat 6 looks better than 99% of all flight games created on PC and it runs very smoothly on Xbox360 while LOMAC looks far inferior in comparison and gives you 25fps with X1950XTX. In the demo of the game a poor IBM cpu on the 360 easily handles 10-15 airplanes in combat with slower graphics card than HD 2900XT!!!!! No matter how many flight sims I've tried playing, their graphics are always inferior relative to the same year FPS games on a PC. At the end of the day, the game should be fun and look good, but flight sims have the worst tradeoff for system requirements vs. graphics output at reasonable framerates. Even if I put together a Quad Core 4.0ghz with 8800GTX Ultra SLI system, there is no flight game that will look as good on a PC today as Crysis while still probably running at 30fps.

To me this shows that the biggest problem with flight sims on a PC is optimization. It largely appears flight sims on a PC are more an indication of poor game coding with respect to all other PC gaming genres. If flight sims had better textures of the ground and water than some of the top FPS, I'd understand the extreme hardware requirements, but they dont. Also don't tell me that Ace Combat is not a "flight sim" but a flying game and that's why the comparison doesn't apply. It is still a flying game, so I expect any other flight sim to look as good and run as smooth. It's not the consumers problem that you need horsepower for AI, etc. It's the developer's job to make the game look good. Alternatively, it's completely fair to compare an F1 car simulator to a racing game like DIRT. The F1 game might run worse and look worse and have just as stringent system requirements as say DIRT or Forza Motorsports 2, but it should be the developers' goal to develop a racing game that looks stunning while delivering smooth framerates on the same hardware, not the consumers. If it takes 1 game double the cpu speed and double the gpu speed to look as good as another game, what conclusion can you make? That's why HL2 engine is so great in that you didn't need a lot of horsepower to have good graphics while Halo port with laughable textures and graphics needed double the graphics power to run smooth and still looked worse.

I am also guessing that people who really really enjoy flight sims probably have fast systems because of flight sim's serious system requirements, so they wouldn't be concerned with cpu vs. gpu compromise to begin with, which is what GPU scaling articles are meant for. For everyone else, when flight sims might be <5% of all the games they play, we have to make compromises in our system and that's where the GPU completely trumps the CPU for performance benefit.

 

[deadseasquirrel]

People keep wanting to make blanket statements-- CPU speed doesn't matter; a GTX is bottlenecked by any CPU; faster CPU speeds will increase min framerates, etc.

The fact is, with regards to computer gaming these days, there is no hard and fast rule. You cannot say that a particular GPU is bottlenecked by an X2 because you upgraded to an e6600 and framerate doubled. What resolution do you play at? You might have shot up 2x the framerate at 1024x768, while another playing at 1600x1200 showed no benefit whatsoever. What game? Supreme Commander loves CPU speed a heckuva lot more than FEAR.

I've given published benchmarks a black eye before and I'll do it again now. Until someone takes the time to start doing extensive and thorough articles, you will never get the full picture regarding gaming benchmarks. More than 3 or 4 games need to be tested. More than 3 resolutions. More than a handful of CPUs (and certainly don't keep them all at stock mhz). And a good variety of GPUs need to be tested (even something from last gen, with Crossfire and SLI thrown in there as well). I want to see max, min, and average. I want more details on *what* they benchmarked... was it just a fraps run-through or what; what part of the game, what level, etc?

If you truly want to use benchmarks to help you determine what you should purchase for your system, then you need to look at benchmarks which come close to analyzing the data that is most similar to what you do with your system. Benchmarking racing games at 800x600 won't help you decide what CPU or GPU to buy if you only play flight sims at 2560x1600.

There is no 1 single correct answer. A GTX *will* be held back by an X2 in some games and at some resolutions. But definitely not all. CPU speed *does* matter to games. It's just that at some settings and on some games, it matters extremely little.

Blanket statements such as CPU speed doesn't matter or a GTX will be bottlenecked by any processor are incorrect at best, misleading at worst. If you're not considering resolution, game settings, the exact game, the CPU, and the GPU in your evaluation, then you're not looking at the full picture.

 

[coldpower27]

Originally posted by: deadseasquirrel
Blanket statements such as CPU speed doesn't matter or a GTX will be bottlenecked by any processor are incorrect at best, misleading at worst. If you're not considering resolution, game settings, the exact game, the CPU, and the GPU in your evaluation, then you're not looking at the full picture.

QFT. Blanket statements suck on the whole, and it's not just limited to computer hardware, their so no one hard rule, some games are more CPU intensive then others, some are more GPU dependent, etc etc.

 

[betasub]

Originally posted by: coldpower27
Blanket statements suck on the whole

Sig worthy, although I wish you'd gone for the full irony of something along the lines of:

"blanket statements always suck"

:laugh:

 

[Polish3d]

Originally posted by: MarcVenice
Goes to show videocards come first, then cpu's. For playable framerates even the celeron and sempron were good enough, exept for x3. So if you're a gamer, you shouldn't spend more then 150$ or so, which almost buys you a x2 6000+ if you are also buying a 8800gts or 8800gtx. Anything more on the cpu and you should be looking to invest it in a better videocard instead.

Yeah but if you save on CPU and then put that cash toward GPU and it gets you SLI, then you'll may need that faster CPU


Overall though, IMO it's much more fun to have a super-fast CPU in terms of overall experience with your system.

 

[apoppin]

Originally posted by: betasub

Originally posted by: coldpower27
Blanket statements suck on the whole

Sig worthy, although I wish you'd gone for the full irony of something along the lines of:

"blanket statements always suck"

:laugh:

Well ... now it is a blanket statement and it can be your own sig ... with credit to no one but "inspired by"

--and has anyone compared AMD 2900XT and CPU scaling to GTS's scaling ?

... if not ... send me a GTS, i'll check it out for you ... a QC CPU would be nice to compare also.