I have no idea what you mean about '1300Mhz to the 670'. I never once mentioned anything even related to that. I said i was going to give a 5% bump to the 670 because it was a reference model being compared to a non-reference with improved cooling and better PCB components.
Lets look at those numbers shall we, as now there's more data available than when i originally wrote this:
7950: 54.7 -- what do you know, it's almost the same as above despite the newer driver and the fact that 2xMSAA is off. Guess my two assumption in the two links above were correct after all?
So you see it makes no difference.
Do you know that the HD 7950 is available in OC editions. People have been clocking in these forums at 1200+ Mhz speeds. I clearly said BF3 will be similar clock for clock
. BF3 is one of the best games for Nvidia and still AMD is on par at OC's speeds. AMD needs to get to similar clock speeds for similar perf. Also there was a concern in hardocp forums where they questioned why 2x and 4x MSAA were not tested. my guess is the answer would be more in AMD's favour.
BF3 Ultra 4X MSAA 2560 X 1600
HD 7970 (950) - avg 40 min 27 fps
HD 7970 (1280 ) - avg 50.7 min 34 fps
GTX 680 - avg 42.9 min 25 fps ( press samples have been shown to boost to 1100 Mhz i am still giving conservative boost)
I would love to see MSAA 4X numbers for GTX 680 OC at 1.3 Ghz and my guess is GTX 680 will fall slightly behind given that MSAA is a huge bandwidth hog
You are showing a HD 7950 (1050) which is at the very low end of voltage OC. Also understand that Nvidia has not allowed voltage unlocking and has capped max voltage at 1.175 for GTX 680 / 670. Its for reasons which Nvidia knows better. when we have unlocked voltage then we will have another comparison. But for all you know AMD's latest chips seem to hit 1250 Mhz easily. So certainly gets interesting
http://www.techpowerup.com/165559/AMD-Readies-Radeon-HD-7970-GHz-Edition.htmlI think we can clearly agree Deus Ex is better on HD 7950 .
First off, the 7950 is being scaled too high. I said i thought i was scaling it up too much and later on after that new HardOCP review came out today i can easily confirm that now. I was WAY too generous with the numbers here in favor of the 7950. I'm a little disappointed that you pointed out the 7950 was not overclocked enough in the Deus ex example above (and i was in agreement with you even before you posted this) but you don't have a problem with the 670 being too low here? This clearly shows some small degree of bias (or maybe you just overlooked it, i hope it's the later ). Given what i know now, i wouldn't have given the 7950 a 19% boost here, but instead maybe only a 12% boost. So instead of 50fps it should have been around 46fps. So 46fps for the 7950 and 39fps for the 670.
HD 7950 OC (1150 ) - 50
GTX 670(915 / 980) - 38
GTX 680(1006 / 1058) - 39
GTX 680 OC (1137 base / 1267 boost) - 41
Now that you have the data and a top of the line GTX 680 with 1267 boost speeds can't cross 41 fps I think it should be pretty clear that bandwidth is affecting the scaling. There is only a 3 fps increase for 192 extra shaders and 220 Mhz core boost with 290 Mhz turbo boost when you compare stock GTX 670 to 680 OC .
No matter what OC I don't think the GTX 680 can even hit HD 7970 (925 Mhz) at 44 fps. What seems to be your problem about the HD 7950 scoring 50 fps ? Perf is what is shown. HD 7950 scales very nicely because it has bandwidth and is not handicapped in any manner at high clocks like GTX 600 series. Is it that voltage OC was used. I said at the start the maximum performance is what I am discussing. The ability to hit 1200+ Mhz speeds with MSI Afterburner is well known and cannot be disounted.
Do you remember Metro 2033 was a Nvidia TWIMTBP title and Nvidia bragged about its performance in this game with GTX 480 and GTX 580 with tesselation and DOF (which uses compute shaders and is bandwidth intensive). Nvidia did not suck at this game with the GTX 580. It really kicked the crap out of HD 6970. If Nvidia could never get good drivers how come they were dominating this game last generation.
HD 6970 - 28
GTX 580 - 33
GTX 680 - 39
GTX 680 OC - 41
HD 7950 - 40
HD 7970 - 44
HD 7950 (1125) - 50the performance scaling from GTX 580 to GTX 680 is the lowest of all games <20%. With GTX 680 OC its close to 25%. look at the performance scaling from HD 6970 to HD 7950 OC. thats 70% and I don't think you can discount bandwidth being a factor
There's no such thing as inconsequential numbers in a benchmark. This data is being used, like i just said above, to find the general strength of the card in ANY game by extrapolating all of our data. But just for the sake of answer your question, there is a big difference between 90fps and 120fps for someone with a 120hz monitor (like myself). I can easily tell the difference both visually and by the feel of the game.
Thats why the choice of games, their relevance in terms of when the game was released and how they really give a good indication of GPU performance. There are many websites who bench Crysis games and Metro 2033 still. They are doing it because they are the most demanding and can show how these GPUs perform when really all resources of the GPU are really stressed like shader power, bandwidth, ROP etc . They also give the best idea of bottlencks. let me give you a good example
2560 X 1600
HD 7970 (925 Mhz ) - 36 fps
HD 7950 - 31.1 fps
GTX 680 - 30.8 fps
GTX 580 - 26 fps
While Crysis was a strong game for the GTX 580, the same cannot be said of the GTX 680. NVIDIA is off to a very poor start here, with the Radeon HD 7970 easily outperforming the GTX 680, and even the 7950 is tied or nearly tied with the GTX 680 depending on the resolution. On the bright side the GTX 680 does manage to outperform the GTX 580, but only by a relatively meager 17%.
Given the large gap in theoretical performance between the GTX 680 and GTX 580, as it turns out we’ve run into one of the few scenarios where the GTX 680 doesn’t improve on the GTX 580: memory bandwidth. In our overclocking results we discovered that a core overclock had almost no impact on Crysis, whereas a memory overclock improved performance by 8%, almost exactly as much as the memory overclock itself.
When it comes to the latest generation of cards it appears that Crysis loves memory bandwidth, and this is something the Radeon HD 7900 series has in spades but the GTX 680 does not. Thankfully for NVIDIA not every game is like Crysis."Do you still want to argue bandwidth is not important for a USD 400 - 500 card which you want to use for atleast 2 years
As for a 120 hz monitor I have a question for you. Between 100 and 110 fps can you tell the difference. I think you can't . Beyond 60 fps things become very difficult for the human eye to discern. Our eyes are not built like professional cameras with thousands of fps capability and ability to distinguish at those speeds.
Yes, some games run better on AMD than NVidia. It has really nothing to do with the cards themselves but with the drivers.
I think thats the most incorrect statement of all. The GTX 580 excelled in games where there was lots of tesselation, very complex compute shaders and high bandwidth requirements like Metro 2033, Crysis 2, BF3. Drivers matter but the underlying hardware also does matter. Look at the HD 6970 being decimated in all these tests by the GTX 580 in the guru3d charts.
That 670 link is to a weak factory overclocked 670. Again, your confirmation bias is clearly coming to light as you have no problem comparing an light factory-oc'ed 670 to a heavy manual overclocked 7950 but when it was the reverse you had a problem with that (and for the record, so did i, proving that i'm being impartial to the data).
As for anno 2070
HD 7950 (800 ) - 67 45
HD 7950 (1020) - 85 56
HD 7950 (1150) - 93 62
GTX 670 (915 / 980 boost) - 75 48
GTX 670 (1058 / 1137 boost) - 85 54
GTX 680 - 87 55
GTX 680 (1137 / 1267 boost) - 100 64
I think your understanding of turbo boost is not enough. the boost speeds are the speeds at which the card runs unless severely constrained by cooling . turbo will function all the time if your case cooling is good and the GPU does not overheat or crosses its TDP which it won't because the binning assures of it running at turbo speeds across most workloads except maybe furmark or folding .
HD 7950 (1150) is fighting against GTX 680 (1267 BOOST) at 1600p. I think we can agree that GTX 670 is no comparison.
Alan wake runs clearly better on HD 7900 series
"The Radeon HD 7000 quite well appreciate this game in which they easily crushed the GeForce GTX 600 models which are particularly affected with 8x MSAA. "
at 1080p max (8XAA) the
GTX 670 - 42
HD 7950 - 50
The cards are closer at 1080p high but the lead grows at max settings. also this is a stock vs stock comparison the OC comparison will only widen the gap.
from another website
at 1080p and 1600p
GTX 670 - 70.6 and 42.9
HD 7950 - 70.4 and 45.5
Given the HD 7950 at 800 Mhz against GTX 670 (980 turbo) is winning at 1600p and tying at 1080p the HD 7950 will stomp the GTX 670 at OC'd speeds. for just an indication of HD 7900 perf scaling
HD 7970 (1070) - 59.9 fps at 1600p
stock HD 7970 (925 Mhz) is ahead of GTX 680 (1110 Mhz boost) card without ubersampling. with ubersampling its a tie. similarly for HD 7950 v GTX 670. I think you and I can agree the OC headroom of HD 7900 will put it clearly ahead when the cards are OC'd.
phew... I am a bit tired Edited by raghu78 - 5/15/12 at 5:52am