Accurate ranking of GPUs?

[Bluescreendeath]

Is this ranking a good indication of GPU and gaming performance?

The 9800GTX/GTX+/GTS250 and 4850 serves as the base value of 1.0 - with higher values being better performance. This is with SLi/Xfire enabled games. This is a general indicator of performance - not specifically fps benchmarks, since Xfire/SLi fps improvements can vary widely in different games.
(ie. Crysis gets maybe a 30-40% fps boost while COD4 gets maybe a 70-80% fps boost)


1 8800GT/9800GT = .8
2x 8800GT/9800GT SLi = 1.4

1 9800GX2 = 1.45

1 4830 = .8
2x 4830 Xfire = 1.4

1 9800GTX/GTX+/GTS250 = 1
2x 9800GTX/GTX+/GTX250 SLi = 1.8

1 4850 = 1
2x 4850 Xfire = 1.8

1 4870 512mb = 1.25
1 4870 1GB = 1.3
2x 4870 512mb Xfire = 2.25
2x 4870 1GB Xfire = 2.34

1 GTX260 192core = 1.25
1 GTX260 216core = 1.3
2x GTX260 192core SLi = 2.25
2x GTX260 216core SLi = 2.34

1 GTX295 = 2.45

1 GTX280 = 1.45
1 GTX285 = 1.5
2x GTX280 SLi = 2.61
2x GTX285 SLi = 2.7


*GX2 to 1.5

 

[mactronix]

Quite why you would want to do this is beyond me personally but i guess if it helps you get your head around it or makes it easier then why not.

Ok there are numerous issues which to me at least would make doing such a thing totally invalid.

1. Crossfire and SLI just muddy the water, you cant just say "This is with SLi/Xfire enabled games" because all the differant games and chipsets support the differant games to differant levels. Some blatently favour one chip or another and so from that point the comparison of performance cant be a fair representation.
2. I see you have the 1GB cards out performing the 512mb cards. this would only be true at higher resolutions. and why do the 2x cards differ by 9 and the single cards by 5
3. You have 1 8800gt at .8 and SLI at 1.4 but the GTX260 gets a whole point increase. when infact with the right drivers and right game the 8800gt cna very nearly double its performance in SLI.

So its not that simple im afraid

Mactronix

 

[Bluescreendeath]

Quite why you would want to do this is beyond me personally but i guess if it helps you get your head around it or makes it easier then why not.

Ok there are numerous issues which to me at least would make doing such a thing totally invalid.

1. Crossfire and SLI just muddy the water, you cant just say "This is with SLi/Xfire enabled games" because all the differant games and chipsets support the differant games to differant levels. Some blatently favour one chip or another and so from that point the comparison of performance cant be a fair representation.
2. I see you have the 1GB cards out performing the 512mb cards. this would only be true at higher resolutions. and why do the 2x cards differ by 9 and the single cards by 5
3. You have 1 8800gt at .8 and SLI at 1.4 but the GTX260 gets a whole point increase. when infact with the right drivers and right game the 8800gt cna very nearly double its performance in SLI.

So its not that simple im afraid

Mactronix

1. I see your point. Although it represents performance in general...which may or may not translate directly into fps. Sure, some games don't support SLi/Xfire, but most nowdays do. And the ones that support SLi and crash with Xfire, or vice versa are the ones with bad drivers that get fixed fairly quickly.

2. True, but for the 4870, a bigger memory means you can turn up AA and AF, not just turning up the resolution.
Perhaps .5 was a too much of an increase for a 512mb memory boost?

3. The GTX260 gets an entire point increase because the GTX260 has a higher starting value (1.2) since it's a much better GPU than the 8800GT. So it would have a bigger improvement in SLi than SLing two 8800GT.

A game that doubles in performance with a 2nd 8800GT will probably double in performance with a 2nd GTX260. Sure you have diminishing values for higher end cards, but I don't think the 8800GT has significantly better SLing capabilities than the GTX260.

For any card, you don't get a 100% improvement from SLi, so I used 80% improvement for adding a 2nd card for SLi. 8800GT getting a .8 + .64 ~ 1.44/1.4 when the 2nd card is applied and the GTX260 getting a 1.25 + 1 ~ 2.25

 

[jennyh]

If a 4830 1gb is 1.3, a 280gtx isn't 1.45 if you consider a 285gtx is 1.5

Those would be a lot closer to 1.3, 1.37, 1.5 imo.

Also, you very rarely get 80% improvement from SLI/Crossfire although it's a lot better than it has been.

You are waaay off on how good you think a 295 is also. It is never 1.7x as good as a GX2. Your crossfire/sli scaling is a long way off tbh.

 

[Bluescreendeath]

If a 4830 1gb is 1.3, a 280gtx isn't 1.45 if you consider a 285gtx is 1.5

Those would be a lot closer to 1.3, 1.37, 1.5 imo.

Also, you very rarely get 80% improvement from SLI/Crossfire although it's a lot better than it has been.

You are waaay off on how good you think a 295 is also. It is never twice as good as a GX2. Your crossfire/sli scaling is a long way off tbh.

I'd rank the 4830 1GB as .85, not 1.3
I ranked the two 4830 in crossfire as 1.4

Although I'm not sure what you mean by 1.3, 1.37, 1.5?

Yeh, you're right about 80% - I considered that to be an example of a game that supports SLi/Xfire to a healthy degree. Games that are awesome for SLi (such as Call of Duty 4), gets almost a 100% improvement from SLi.


Although I do think the GTX295 wipes the floor with the GX2. A 9800GX2 is two OCed 8800/9800GT or two 8800GTS cards in a single GPU. A GTX295 is two OCed GTX260 or two weakened GTX280 cards in a single GPU. HardOCP showed that the GX2 struggled in games such as Crysis and Fallout3, and was outperformed by a healthy margin by two 9800GTXs in SLi.

Though you're right that the GX2 is stronger...maybe 1.5, equal to a GTX285.

The GX2 beats the GTX280 at COD4, but gets beaten at Crysis.

 

[king_play334]

LOL, so what's better the GTX295 SLI or 9800GTX SLI?, once in sli with 2 physical cards you are running quad SLI with 2 GPU's on each card. Or am i better off going for GTX 285 SLI?, thanks!

 

[Bluescreendeath]

2xGTX295 SLi (quad SLi)
or
2x4870x2 Xfire (quad Xfire)
or
3xGTX285 (triple SLi)
or
3xATI4890 (triple Xfire)

are currently the best setup you can have.

Because the more GPU you have, the performance gain for each additional GPU decreases, and because Quad-SLi/Xfire is VERY expensive, you're usually better off getting 2 high end cards in SLi...
ie. get 2xGTX285 Sli or 2xGTX275 SLi or 2x4890 Xifre

 

[EXT64]

Be careful though, as going from 2-3 (or 3-4 especially) GPUs doesn't add much (especially compared to cost).

 

[Annisman]

Oh yea, my third 48701GB Core, adds like, 10-30% performance maximum, and that's in games that actually allow multi-gpu scaling, or aren't CPU limited, your BEST bet is one GPU, and you could argue that a second is logical.

 

[gamerk316]

The GX2 is a freaky card, and drivers have made it an attractive option at $150 or so. I upgraded from a 9800GX2 to a 4890, and I LOST FPS in Company of Heroes (I could crank up AA at high res though, but at highest settings/2x AA, the GX2 was faster).

Someone really needs to re-review the GX2 with new drivers, testing both with and without AA. I'd be interested to see the results...

 

[jennyh]

That's because Company of Heroes favours Nvidia cards by a huge margin. CoH is to Nvidia like Grid is to ATI.