AMD to launch new flagship Radeon HD 7970 desktop graphics card on December 22.

[guskline]

Did you see what AMD used as it's primary test rig? An Intel 3960X SandyBridge E. The 8150FX Bulldozer was used as a "secondary" testbench for Eyefinity. CEO Rory should use the GPU guys for AMDs CPUs.

 

[Arkadrel]

@Guskline

I really hope that isnt true, thats not really fair benching.

 

[Vesku]

Sometimes I really don't get Russian, he can be pretty straightforward but then he starts comparing a 6970 to a 580 in order to say AMD doesn't have any games where it has a distinct advantage.

Shogun 2 has a distinct 20+% advantage to 6970 over the 570 in 1080P and 2560x1600. Sure you can say NVIDIA tends to have a distinct max FPS advantage in several games (TWIMTBP at work) but to say there are 0 games where that is true for AMD (Gaming Evolved at work too) does not seem to hold water.

 

[guskline]

@Guskline

I really hope that isnt true, thats not really fair benching.

Isn't true? Sorry, but I just read it in the AMD "official" (leaked early) PDF file on the 7970 benchmarking specs. Check for yourself!

 

[Arkadrel]

http://www.mediafire.com/?f5k5bfuovbrz8dj

^ that is the only one Ive seen so far, its 61 pages long, looks offcial, and doesnt show any system specs in it.

its called :
AMD Radeon HD 7900 series zonder notes en watermerk.pdf

 

[skycake]

@Guskline

I really hope that isnt true, thats not really fair benching.

What's unfair about that? Shouldn't they use the most powerful CPU they can to ensure it's the graphics card limiting the benchmark and not the CPU?

 

[Arkadrel]

Skycake, the only thing that should be differnt in each tested system, should be the graphics cards.

I assume that when I see a chart.
Everything else should be the same, settings ect, hardware ect.

That said, are you sure those are "real" guskline? or official amd slides? maybe their 3rd party from some hardware review site that didnt have 2 identical pcs or something.

Also in the pdf I looked at, there wasnt any system mentions, or really much/any performance figours.

 

[guskline]

Arkadrel: I'm referring to the @872k PDF file that is not graphical. The 13 meg file has the powerpoints.

 

[blackened23]

And why wouldn't I want to force AA via drivers? Did you forget that MLAA causes horrible texture blurring while conventional MSAA does not? So what makes you think somebody considers MLAA "good enough" when an NV card is perfectly fine with running MSAA in SC2? Another excuse that this scenario shouldn't count because AMD cards suck in deferred AA modes (similar to BF3)?

I'll answer this one for you. I own both 580 sli msi lightnings and 6970 crossfire in 2 PC's, so I feel qualified to state what the pros and cons of each brand are.

First of all since you don't own nvidia, i'll take it that you don't realize the ramifications of this. I'm not sure if you're suffering from grass is greener syndrome, but i've used both brands and here's the deal: Override settings DO NOT WORK a great majority of the time on nvidia cards. How nvidia does things is on a per game basis, if you look inside of nvidia inspector that will tell your card how to behave regarding override settings for that particular game. Most games are flagged with "treat override as use application preference". I've combed through the list of games and roughly 65% of games COMPLETELY IGNORE override settings. This has been mentioned to me before but I discovered it first hand. There are TONS of games that I OWN that completely ignore everything I put in a profile setting. Some examples: Dead Space 1/2, Dead Island. I got so frustrated because Dead Island looks like garbage without AA, with AMD CCC you can simply check SSAA (which runs respectably on xfire) yet with nvidia you CANNOT enable AA. You can use the override setting which is ignored. Load up nvidia inspector: Sure enough, "treat override as use application preference". There are *many* other games as well that nvidia completely ignores override settings for. Sometimes you can input settings into nvidia inspector to get past that, but most of the time you can't.

Second point. If you're testing with driver AA settings you are wrong. You are testing driver functionality which doesn't always enable the AA that you specify, and you are not testing the game. You should ALWAYS USE INGAME AA for benchmarking purposes.
-- Now I love my SLI MSI lightnings. I prefer them for gaming. But i'll be the first to tell you nvidia has caveats that you probably aren't aware of since you don't own it. Override being broken being one, and also not supporting SSAA is another. (AND transparency SS is *not* SSAA) You can use SSAA in dead space and other older games with AMD CCC. With nvidia, you only get TrSS (which is garbage) and you can't enable AA in dead space unless you do some whiz bang stuff in nvidia inspector -- you have to enter a specific AA compatibility bit code into nvidia inspector. And then enabling anything higher than 8x AA will result in a dead space slide show. I don't even consider this because no end user should have to go through these steps to enable AA in a game.

I have to wonder if nvidia gets an unfair advantage in benchmarks done by reviewers because of this. Because driver settings do nothing for 60%+ of games, yet most users are completely oblivious to it. Whatever. So basically, Russian, if you do things on a driver basis you're testing 1 card that is using AA and another that isn't using AA. Because enabling driver AA isn't a 100% deal, and with nvidia CP it is closer to 50% (hit or miss). AMD CCC has better override settings that work more often than not, although it doesn't function in some DX11 games. I've mentioned this to some hardware editors from various websites that have done this practice and they ended up agreeing with me. Because i've provided proof that override doesn't work in many games -- and they only use ingame settings now. Testing using driver settings is not fair because as I said -- one card may ignore the override setting while another doesn't.

anyway, russian, you should get sli 580s as I did and see this stuff for yourself. Obviously you have major grass is greener syndrome. Then you'll start posting with AMD defense posts :shrug:

 

[guskline]

Arkadrel: Here's the link I'm referring to:
http://www.megaupload.com/?d=H86KSABC

Check out Appendix 2 on pg 27.

 

[guskline]

What's unfair about that? Shouldn't they use the most powerful CPU they can to ensure it's the graphics card limiting the benchmark and not the CPU?

Skycake: I don't disagree with you. Just a bit of Irony - that's all, especially with the build up to the release of the Bulldozer.

 

[RavenSEAL]

So according to this, 2 HD 6870s, which are 20~ FPS faster than a GTX 580 in Heaven Benchmark default settings, should be a bit slower than this?

 

[Arkadrel]

@Guskline:

I think your mis-interpreating it.

They have 2 systems, but they used same cpu on both gpu's for tests x-y.

It just means that all the tests wherent run on the intel cpu, some where run on a bulldozer, but for those tests they used the same system (with the bulldozer) for both the Nvidia card and for the AMD one.


^ thats how I read it anyway looking at it.

You make it sound like they put the AMD card into the Intel machine, and run tests, and put the Nvidia one into a bulldozer and run tests, and then compair them. I dont think thats the case.

Thats you jumping the gun on assumptions, because they dont list card uses in the configourations.
Only the systems, thus I assume for some reason they ran maybe for exsample 10 tests with Intel system, and then ran for exsample 5 tests with the bulldozer or so, but they did same tests, with same systems on both cards.

 

[blackened23]

Nvidia CP ignores override settings for 50%+ of games: Just go through the list to see for yourself if you have a nvidia card.

So this is why you don't test with driver settings, Russian.

 

[skycake]

@Guskline:

I think your mis-interpreating it.

They have 2 systems, but they used same cpu on both gpu's for tests x-y.

It just means that all the tests wherent run on the intel cpu, some where run on a bulldozer, but for those tests they used the same system (with the bulldozer) for both the Nvidia card and for the AMD one.


^ thats how I read it anyway looking at it.

You make it sound like they put the AMD card into the Intel machine, and run tests, and put the Nvidia one into a bulldozer and run tests, and then compair them. I dont think thats the case.

Thats you jumping the gun on assumptions, because they dont list card uses in the configourations.
Only the systems, thus I assume for some reason they ran maybe for exsample 10 tests with Intel system, and then ran for exsample 5 tests with the bulldozer or so, but they did same tests, with same systems on both cards.

I'm not sure where you got that impression from. It sounded to me like he was just laughing at the idea of AMD graphics division not having enough faith in AMD CPU division to use their top-end CPU for benchmarking.

I certainly can't blame them for the decision, but it's definitely amusing.

 

[Arkadrel]

Thats because AMD cpus arnt up to pair.... even AMD GPU devision knows it.... sad but true.

 

[AdamK47]

First of all since you don't own nvidia, i'll take it that you don't realize the ramifications of this. I'm not sure if you're suffering from grass is greener syndrome, but i've used both brands and here's the deal: Override settings DO NOT WORK a great majority of the time on nvidia cards. How nvidia does things is on a per game basis, if you look inside of nvidia inspector that will tell your card how to behave regarding override settings for that particular game. Most games are flagged with "treat override as use application preference". I've combed through the list of games and roughly 65% of games COMPLETELY IGNORE override settings. This has been mentioned to me before but I discovered it first hand. There are TONS of games that I OWN that completely ignore everything I put in a profile setting. Some examples: Dead Space 1/2, Dead Island. I got so frustrated because Dead Island looks like garbage without AA, with AMD CCC you can simply check SSAA (which runs respectably on xfire) yet with nvidia you CANNOT enable AA. You can use the override setting which is ignored. Load up nvidia inspector: Sure enough, "treat override as use application preference". There are *many* other games as well that nvidia completely ignores override settings for. Sometimes you can input settings into nvidia inspector to get past that, but most of the time you can't.

I've had luck with the "Enhance the application settings" in the driver profile for Dead Island. I can set it to 4X / 8X FSAA. With the 295.18 drivers I can also apply FXAA over top of it. Looks really good. With Dead Space 1 and 2 you have to manually add an AA compat flag to the profile for AA to work. Even then it only half works. I simply use FXAA Injector (the driver FXAA doesn't yet work for this game) along with the games built in AA. Not as good as MSAA+FXAA, but it still looks nice.

 

[blackened23]

I've had luck with the "Enhance the application settings" in the driver profile for Dead Island. I can set it to 4X / 8X FSAA. With the 295.18 drivers I can also apply FXAA over top of it. Looks really good. With Dead Space 1 and 2 you have to manually add an AA compat flag to the profile for AA to work. Even then it only half works. I simply use FXAA Injector (the driver FXAA doesn't yet work for this game) along with the games built in AA. Not as good as MSAA+FXAA, but it still looks nice.

This type of stuff shouldn't be used in benchmarking. You're testing the cards strength, so if you're testing 1 card that might obey AA override and another that doesn't. OBVIOUS CONCLUSION.

Like I said, I love my 580s to death but the fact that many of my games can't use override IQ settings is extremely annoying. AMD CCC is without a doubt better at obeying override specifications, nvidia seems
to outright ignore it more often than not.

 

[guskline]

AMD's words, not mine. Primary Benchmarking system was the Intel 3960X

 

[SHAQ]

Nvidia CP ignores override settings for 50%+ of games: Just go through the list to see for yourself if you have a nvidia card.

So this is why you don't test with driver settings, Russian.

So you enabled ambient occlusion in Dead Space and expected it to work in Oblivion? Everything I have enabled in the last year has worked. I see the results on screen and by GPU loading/vram usage. I don't even mess with the CP most of the time.

 

[blackened23]

So you enabled ambient occlusion in Dead Space and expected it to work in Oblivion? Everything I have enabled in the last year has worked. I see the results on screen and by GPU loading/vram usage. I don't even mess with the CP most of the time.

LMAO. So you enable driver forced AA without messing with CP most of the time? Very interdasting. Obviously you haven't used nvidia inspector becuase you don't know how from the wording of your post. I only posted the nvidia CP pic to substantiate using a 580. Secondly, nvidia inspector, if you look in the "common" section, I don't have SSAO enabled. Everything in that profile is default value.

Also: ignorance / placebo effect is bliss Am I using AA? Hey I think so.

 

[guskline]

@Guskline:

I think your mis-interpreating it.

They have 2 systems, but they used same cpu on both gpu's for tests x-y.

It just means that all the tests wherent run on the intel cpu, some where run on a bulldozer, but for those tests they used the same system (with the bulldozer) for both the Nvidia card and for the AMD one.


^ thats how I read it anyway looking at it.

You make it sound like they put the AMD card into the Intel machine, and run tests, and put the Nvidia one into a bulldozer and run tests, and then compair them. I dont think thats the case.

Thats you jumping the gun on assumptions, because they dont list card uses in the configourations.
Only the systems, thus I assume for some reason they ran maybe for exsample 10 tests with Intel system, and then ran for exsample 5 tests with the bulldozer or so, but they did same tests, with same systems on both cards.

I know they used the same machine for both the AMD 7970, 6970 AND the Nvidia GTX 580. What I find funny is despite the big hoopla about the release of the Bulldozer, the AMD test guys opted to use the Intel 3960X as the primary testbed and listed the Bulldozer 8150FX as the secondary testbed. OUCH!

 

[Arkadrel]

the AMD test guys opted to use the Intel 3960X as the primary testbed and listed the Bulldozer 8150FX as the secondary testbed. OUCH!

Admitting their CPUs arnt upto snuff yeah, that Intel 3960X > bulldozer.

 

[skipsneeky2]

Did you see what AMD used as it's primary test rig? An Intel 3960X SandyBridge E. The 8150FX Bulldozer was used as a "secondary" testbench for Eyefinity. CEO Rory should use the GPU guys for AMDs CPUs.

Pulls a 3960x out of my ass and hoping for the best performance with a radeon 7000 series gpu.

Mainstream gaming cpu is a 2500k why the hell are they testing it with a cpu 90% of gamers can't afford to buy or are not stupid enough to buy cause a 2500k is just as good?

Will the 2500k bottleneck the 7970?

 

[SirPauly]

LMAO. So you enable driver forced AA without messing with CP most of the time? Very interdasting. Obviously you haven't used nvidia inspector becuase you don't know how from the wording of your post. I only posted the nvidia CP pic to substantiate using a 580. Secondly, nvidia inspector, if you look in the "common" section, I don't have SSAO enabled. Everything in that profile is default value.

Also: ignorance / placebo effect is bliss Am I using AA? Hey I think so.

One may change the behavior flags with nVinspector -- I like hybrid/mixed modes at times and still enjoy to play Supreme Commander 2 skirmish modes -- I have to change the behavior flag to do this.