Matching CPU, Bus speed and Ram to a Graphics Card

[Delorentis]

Hi all,

Firstly - I am a noob so *shy wave* hope I can make a contribution to the informative discussions here.

I have a major query - one that has become more of an issue for me since I have recently upgraded both my graphics card and CPU. It struck me that having a guide to match CPU, M/B bus speeds and RAM figures to the graphics cards (even if they were guidelines and not entirely accurate) would have been more useful to help me get bang for my buck.

Does one of these exist (any of you know?) or could an intrepid and far more industry wisened person have a go at it? Is it foolish to even think this could be of any use?

Cheers!

 

[Diplo]

That's a good question, and one I'd be interested in hearing answered. Has anybody every actually come up with a graph or chart where you could gage the perfect CPU/GPU pairing? It would be really useful if you could look up your rough CPU speed and then see which cards would make the most of it. Likewise, if you're upgrading, you could see what CPU would do justice to the flashy GPU you own...

Of course, people are too busy bashing Nvidia and down-playing ATI to probably notice your thread

 

[Randell]

Of course, people are too busy bashing Nvidia and down-playing ATI to probably notice your thread

for somebody who's been a member for 2 days, don't you think it's too early to be slagging off helpfulness of the forum?

Persoannly I thinks its just too much hard work for people too do. i.e. check lot's of different cpu/fsb speeds against differnt video cards. Its bad enough checking one video card against another in a variety of games.

My basic rules of thumb are -

don't buy the bleeding edge in anything as their is always a price premium that isnt necessarily cost effective

always get good fast memory i.e. max out the fsb in your system first (i.e. fast DDRAM) - but again don't buy the bleeding edge

buy the best of what you can afford in each hardware category, and match what seems cost effective. i.e. if your system is quite old, a medium to high cpu with a mid-range/fast GPU is probably better than a slow CPU and a top of the range GPU. e.g. why buy a 9800XT for around £300 when a 15% slower 9800Pro can be had for half the price!

 

[digitalwanderer]

Do you mean something like a "A PIII800 is a good match for a GF4, but the 9600 Pro is overkill for it. Matching the 9600 Pro to a 1.68Ghz Tulatin Celeron showed the true power of the card, but it didn't see much improvement being put in a system with a Barton 2500+. (Conversely the 9700 Pro gamed about as well as the 9600 Pro in the 1.68 Celery)" or some such like that?

 

[DW Fan!!!!!]

You just managed to fry my brain completly Dig, its gonna take about 20 mins to unravel what you just said and repair the damage

 

[digitalwanderer]

Sorry, just some bits from a 9600 Pro review I've still not finished. I was comparing it on all my rigs against all my cards and it kind of gave a good feel for how different cards work on different rigs, it's what I am shooting for at least.

 

[DW Fan!!!!!]

Nice, where you gonna be posting the review then, ill take a gander at it, my system is a bit backward at the minute as I have a Rad 9800pro matched with a duron 800MHZ , im gonna upgrade as soon as those A64 939 pin chips come out as UT is running like a stop motion film at the minute and I cant get into it.

 

[Tahir2]

OMG Dig you got a stalker....

As to this topic on an NVIDIA NFORCE2 system with onboard gfx the best performance is gained by having everything running at the same FSB (Processor 333, memory 333 etc) and at the lowest latency you can manage from my limited experience.

 

[DarN]

I think the stalked one has a point.
It would certainly be helpful with such an approach to things. Like: "This card needs a Barton 2500+ to fully utilize its potential. If you have anything less, don't bother upgrading".
However CPU/GPU limitations vary with each game...

 

[Unknown Soldier]

I think Tom's had something like that already, where they tested 32 CPU's??

I seem to remember something like that. I know they also tested alot of GPU's at one time too.

US

 

[digitalwanderer]

However CPU/GPU limitations vary with each game...

I forgot to mention that, but I did take it into consideration. I ranged my games pretty good from old games to the latest just for that reason. And you are quite correct, it varies from game to game/pc to pc...but you really can still get a good idea of how well various games play on the various hardware.

 

[Tim Murray]

CPU/GPU limitations vary per game, yes. And that's exactly why this isn't going to work in a general sense. If you pick a specific program to test, sure, you could probably have a go at it. But remember, you're going to see times in each game (regardless) when you're CPU limited, and times when you are GPU limited (well, in newer games at least). I don't think it's even possible to remove CPU performance as a factor entirely from a benchmark (unless you are the God of Demo Recording).

 

[Ylandro]

Hmm... looking back all these years, I think there's two guide lines:

1. The cpu should be comparable to the video card at that time.
current high end videocard + current high end cpu,
mainstream videocard + mainstream cpu.

Most games are pretty balanced and require both good GPU and GPU. There are notable exceptions. Some are extremely CPU bound, other extremely GPU bound. But game developers generaly take customers complete system into account, and try to optimise for it.

Last two years it seemed that the GPU became more important than the CPU, but newest games seem to take increasing amounts CPU power again. We're getting used to fancy graphics, and want good gameplay again.

2. videocard performance scales more or less lineairly with cost.
CPU most definately do not.

Double the price of your videocard, you will see generally see an equal raise in performance. Even at high end.
However, with the CPU it ridiculous. At high end, double the price will give 5% performance increase. Can't understand anyone pays $500 extra for a few % performance (P4 EE's )
And at low-end CPU's you often can get double performance for 30% cost increase.
With CPU's there's always a very clear optimum bang for buck point. With videocards, that optimum is not so clear, or even non-existant.


Base your system choice on the videocard. Judge from review sites which ones suites your requirements. Then select a CPU for it. And make sure it's not outragely expensive. If you're going over budget, first look at the CPU. Then again at CPU, and then at the videocard.

Choice of memory, motherboard and stuff generally does not matter much to performance. In those case it does, you have to know what you're doing. There certainly aren't good guidelines for it.

 

[Fuz]

Having a cpu bottleneck would mean you could crank up all the goodies on your vid card (AA, AF and Resolution) for free.

I am not saying that pairing a high end vid card to a low end cpu is a good idea, but if you have a mid range machine.... looking at a high end vid card isn't such a bad idea.

Ideally, you want to eliminate all possilble bottlenecks.... but budget doesn't always allow this.

 

[Diplo]

for somebody who's been a member for 2 days, don't you think it's too early to be slagging off helpfulness of the forum?

Awww, c'mon, that was a wry 'smiley' comment, not 'slagging off'. However, as you have more posts than me, I'll take my ticking off like a good little junior.

 

[Randell]

for somebody who's been a member for 2 days, don't you think it's too early to be slagging off helpfulness of the forum?

Awww, c'mon, that was a wry 'smiley' comment, not 'slagging off'. However, as you have more posts than me, I'll take my ticking off like a good little junior.

I may have been a bit harsh, but it's not been good round here lately with lots a new members coming in on the attack against other people.

 

[Delorentis]

Cheers for the responses everyone

I particularly take the point of GPU/ to next gen GPU improvement vs CPU to next gen CPU improvement. This is what I found when upgrading from an Athlon XP 1600+ to an Athlon XP 2400+ - the jump for the buck was pretty good. Similarly when I upgraded from an AIW Radeon 7000 to the AIW Radeon 9700. That was great actually...

So it really depends on the games and the stress they put on each component - and a general guideline would be to max out on all components if you choose one of them to be high end. If not - go mainstream and match the components that way. Hmmm...if this is the case - could you for instance use something like winstone for the CPU performance and get a scale which you can then match to a 3d mark03 (GPU intensive) scale. eg:

Winstone 10000 - 15000 is best matched to a 3dmark03 3000-5000?

 

[Vadi]

There is another factor: link

 

[Diplo]

I may have been a bit harsh, but it's not been good round here lately with lots a new members coming in on the attack against other people.

OK, point taken, and I understand where you are coming from. No offence meant or taken. I actually joined this forum because of the interesting technical discussions and not because I'm monotheistic (if you get my drift). Everybodies a newbie at some point, remember

 

[jolle]

I think this captures the subject rather well..
http://www.anandtech.com/video/showdoc.html?i=1650&p=6

unfortunatly there is nothing newer then Geforce4 and R8500 in there..
such tests should be done more often tho...