[gamegpu.ru] Game data comparing high end GPUs

[BrightCandle]

Gamegpu.ru tests a lot of games over a long period of time with decent high end hardware. They are testing a lot of different graphics configurations on a much wider range of games than any other site. Unfortunately they don't do roundups and when they do they stick to the same games as the main review sites. However they also test less well played games like Divinity Dragon Commander and Payday 2 and Saints Row IV, games you wont see in an Anandtech etc review but which surprisingly often struggle with performance. Just like with the CPU data I have started going through their reviews and capturing the frame rates for some of the high end cards (7950, 7970, 7970GE, the Asus dual 7970 card, 680, 690, Titan, 780s, 670 etc) and putting it into a spreadsheet.

https://dl.dropboxusercontent.com/u/3638175/GameGPU GPU performance.ods

What I have then done with the data is looked at some comparisons which are often contentious (different competing brands but similar price point, or SLI verses not or fastest single verses SLI etc) at the two resolution points that they capture and done some simple analysis where the card listed on the left is compared to the card on the right. A performance of 109% tells us the card on the left is on average 9% faster than the card on the right.

1080p

[font=courier]
		Titan v 690	690 v 680	Titan v 680	680 v 7970	7970GE v 7970	690 v 2x7970
Maximum		160%		200%		190%		131%		111%		230%
75.00%		137%		170%		140%		114%		110%		108%
[b][i]Average		107%		131%		133%		106%		109%		106%[/i][/b]
25.00%		85%		97%		128%		100%		110%		94%
Minimum		69%		91%		100%		80%		100%		49%
[/font]

1600p

[font=courier]
		Titan v 690   	690 v 680	Titan v 680	680 v 7970	7970GE v 7970	690 v 2x7970
Maximum		156%		197%		182%		135%		112%		197%
75.00%		142%		180%		147%		104%		110%		98%
[i][b]Average		109%		138%		141%		99%		109%		92%[/b][/i]
25.00%		81%		98%		135%		91%		109%		79%
Minimum		72%		95%		100%		76%		100%		50%
[/font]

I have a lot more data to capture, I haven't even finish the reviews they did this year let alone last year. But already I thought there is some interesting data shown.

 

[ocre]

I didnt expect titan to be faster than the 690 when averaged out. Thats interesting for sure. Then the complete 180 the 690 v 2x7970 does from 1080p to 1600p.

 

[RussianSensation]

I didnt expect titan to be faster than the 690 when averaged out. Thats interesting for sure.

That makes total sense. If you look at GameGPU reviews, they review games the same day they come out, or max 2 days. If you look at many of their reviews, HD6990/590/690/7990 have no scaling in the benches since the SLI/CF driver/profile is not in place yet. Therefore, the comparison of their benchmarks for Titan vs. dual-GPU is a total waste of time. You need to look at other reviews for those benchmarks or otherwise you are adding no performance advantage in many instances based on how GameGPU runs their testing. This becomes evident when you compare 690 vs. 680. To add to that, for proper dual-GPU review, frame times/latency is a must which automatically makes GameGPU comparison of FPS of HD7990/690 vs. Titan largely irrelevant to begin with.

As far as other benchmarks go, I do not understand why 7970 is being compared to 680 as HD7970GE has been a direct competitor to the 680 since June 2012. The comparison of 7970 vs. 680 is also irrelevant then.

 

[Majcric]

Perhaps, because many view the 7970 as still the 7970 whether it be a GHZ version or Vanilla. You could easily replace that "GE" with "OC" and still be looking at the same cards. Long story short, there is not enough performance difference between the GTX 680, 7970, 7970GHZ worth talking about.

 

[-Slacker-]

"max/75%/avg/25%/min" .... that's referring to PFS, right? As in max/avg/min fps?

 

[RussianSensation]

Perhaps, because many view the 7970 as still the 7970 whether it be a GHZ version or Vanilla. You could easily replace that "GE" with "OC" and still be looking at the same cards. Long story short, there is not enough performance difference between the GTX 680, 7970, 7970GHZ worth talking about.

That's not the point. HD7970GE competed with 680 and 7970 with 670. Since performance data was available since June 2012, why is a 925mhz 7970 being compared to the 680? Those were not competitors after June 2012. Since the comparison is being made of stock cards, stating that 925mhz is identical to 7970 OC isn't fair either since 680s can overclock as well. The OP should have compared 670 vs. 7970 and 680 vs. 7970GE if he wanted to be consistent wrt to stock GPU comparisons.

 

[BrightCandle]

Where I am the price of the 7970 is slightly lower than the 680 and the 7970 GE is higher than the 680 but the 7970 is closer in price to the 680 than the 7970. I see however from checking newegg that in the USA the Ghz edition is actually priced similar to the 680 and not significantly above it like here. That is an interesting difference, so I'll add that comparison as well for the Americans in the audience. But its something to be aware of in the future, its not true in all markets that the Ghz edition competes with the 680, its actually the 7970 v 680 in Europe as far as I can tell from my check today.

I think the comparison on or near day of release is actually what most of us want. I don't know about you guys and gals but I play most games right after they come out, only a few games get played weeks/months/years. If the drivers aren't there on the day of release I'll never experience the improvement. But I think its also that AAA games do come with SLI/Xfire support and the rest largely don't ever get it because they aren't popular enough. When we look more widely than the AAA games that hardware review sites use we see that a lot of games don't perform very well and which card ends up being the best changes.

The Maximum/75%/Average/25%/Minimum are referring to the games comparing average FPS. Gamegpu.ru only provides average FPS data so I am looking at the percentage differences of each game between the cards and then providing a few key points of that comparison. IE the maximum is the best advantage the card on the left has over the card on the right. The average is the average advantage it has. The 25% point is like taking all the comparisons, sorting them in order and then counting through 25% of them and looking at the advantage (or disadvantage) found. So no it does not refer to the minimum FPS, gamegpu.ru doesn't provide that information.

As I look through the numbers rather than the summaries I come to two conclusions. You are definitely better off having 1 single fast as possible card when playing anything less than the AAA games, of which there are probably about 10 a year maximum. But dual cards is really good for the top games according to the data. AMD v NVidia isn't very interesting at the same price points frankly.

Something else pops out to me as well - look at Titan at 1080p verses 1600p. I notice that it doesn't seem to gain any additional advantage over the 680. Seeing as how we all think (thought?) that the 7970 was doing better there due to memory bandwidth the Titan should be stretching out just like the 7970 does by about 8-9%, but it isn't. All the advice being given to people about choosing Titan/780's for high resolution doesn't look to be right based on this information. Again on an individual game basis its true, but overall it doesn't appear to be.

 

[Black Octagon]

Where I am the price of the 7970 is slightly lower than the 680 and the 7970 GE is higher than the 680 but the 7970 is closer in price to the 680 than the 7970. I see however from checking newegg that in the USA the Ghz edition is actually priced similar to the 680 and not significantly above it like here. That is an interesting difference, so I'll add that comparison as well for the Americans in the audience. But its something to be aware of in the future, its not true in all markets that the Ghz edition competes with the 680, its actually the 7970 v 680 in Europe as far as I can tell from my check today..

I would agree with this and would say it applies to Australia in addition to Europe.

I must also admit I am surprised to hear RS say that the 7970 was not a competitor to the 680 after the Ghz edition arrived. I don't see why we should treat the 7970 Ghz edition as different to any other factory OC'd 7970...also, the 7970 came out before the 680 and both were flagship products. I always saw 7970 as competing with 680, 7950 with 670 and so on. Yes, what was available on the market changed a little in June 2012, but by that logic the issue of "what competed with what" changed again with the 12.11 Catalyst drivers...I don't see these changes as 'dramatic enough' to warrant a change in which GPU competed with which, personally.