AMD Radeon HD 7990 6 GB

[Nihilus]

2x 8 pin plus the PCI is 375 watts in a PERFECT world. This card doubtfully can get all of that and I think AMD screwed the pooch on the end of it.

 

[W1zzard]

Is it really an AMD issue though? Does AMD control which coils vendors use on their cards or just the specifications? I guess Im not certain what control AMD has over the reference board specification and the parts that go on it.

AMD has full control over the reference board. Their engineers design the PCB routing, pick the components to use and are responsible that it works properly.

All these cards are produced for AMD and AMD pays for them. If a board partner wants a 7990 he has to buy the card from AMD.

On other designs the reference design schematics, BOM (bill of materials), PCB layout, advisory documents are released to board partners so they can produce the cards and optionally modify stuff and swap out components to reduce cost or make it "better". That's how non-reference designs are created. I'm not sure if/what requirements AMD puts on these customs designs and what kind of testing AMD requires, but NVIDIA is VERY strict.

Maybe they need to pick up some better parts.

to fix coil noise you need to pick "different" coils, not "better" or "more expensive"

this is just an engineering problem, it's done in virtually every product in your house that consumes significants amount of power and runs different voltage internally

 

[Prima.Vera]

...
to fix coil noise you need to pick "different" coils, not "better" or "more expensive"

this is just an engineering problem, it's done in virtually every product in your house that consumes significants amount of power and runs different voltage internally

I thought those products are rigorously tested before released to the public. This makes me think AMD is not as serious company as it should...

 

[qubit]

I thought those products are rigorously tested before released to the public. This makes me think AMD is not as serious company as it should...

This sounds to me like a quality control issue more than anything else. Still inexcusable, but I can't imagine the engineers testing out the prototypes, hearing this awful noise and going "Yeah, it's just fine!" Nah, can't see that. Someone fucked up somewhere.

 

[radrok]

AMD has full control over the reference board. Their engineers design the PCB routing, pick the components to use and are responsible that it works properly.

All these cards are produced for AMD and AMD pays for them. If a board partner wants a 7990 he has to buy the card from AMD.

On other designs the reference design schematics, BOM (bill of materials), PCB layout, advisory documents are released to board partners so they can produce the cards and optionally modify stuff and swap out components to reduce cost or make it "better". That's how non-reference designs are created. I'm not sure if/what requirements AMD puts on these customs designs and what kind of testing AMD requires, but NVIDIA is VERY strict.



to fix coil noise you need to pick "different" coils, not "better" or "more expensive"

this is just an engineering problem, it's done in virtually every product in your house that consumes significants amount of power and runs different voltage internally

So basically my two Zotac and Gainward cards differ only in serial stickers behind them?

nVidia produced them and AIB put stickers and box on it, right?

Thank you for sharing the knowledge, much appreciated.

 

[qubit]

On other designs the reference design schematics, BOM (bill of materials), PCB layout, advisory documents are released to board partners so they can produce the cards and optionally modify stuff and swap out components to reduce cost or make it "better". That's how non-reference designs are created. I'm not sure if/what requirements AMD puts on these customs designs and what kind of testing AMD requires, but NVIDIA is VERY strict.

I wish nvidia had been stricter with the GTX 580 reference design. Until recently, I was using a Zotac Amp! version, which is basically a reference card with a slightly higher clock speed. Everything worked fine, even in the most demanding games and benchmarks designed to stress it out - until I tried Folding@Home with it.

After a few seconds, it would suddenly start to emit the most abominable and loud coil noise you could imagine. It was so severe in fact, that it broke through the audio circuitry and clearly came out through the speakers!

I was so pissed off about this that I made a thread about it here. Yeah, sure I could have probably used glue to lessen the problem, but I wasn't up for taking the thing apart and voiding the warranty over this, so I dont run F@H on it.

I've now got the EVGA GTX 590 in my specs, but haven't gotten around to trying F@H on it yet. I hope this doesn't have the same problem.

 

[Rahmat Sofyan]

I'm really concerned about performance per dollar score, does it include or considered the 8 game bundle?or only as it is as graphic card?

 

[HumanSmoke]

This sounds to me like a quality control issue more than anything else. Still inexcusable, but I can't imagine the engineers testing out the prototypes, hearing this awful noise and going "Yeah, it's just fine!" Nah, can't see that. Someone fucked up somewhere.

Not an entirely unknown situation. Remember the design problem that the HD 6970 suffered with PCI-E plug placement ?

 

[W1zzard]

I'm really concerned about performance per dollar score, does it include or considered the 8 game bundle?or only as it is as graphic card?

it uses the 1000 USD price and the results from performance summary (not performance summary 2)

 

[Rahmat Sofyan]

it uses the 1000 USD price and the results from performance summary (not performance summary 2)

ahhh see

now I can see clearly, where is performance per dollar score come

thanx W1zz.

 

[mirakul]

IMHO, including those 3 Blizzard games in the test is unfair. All of them favor nVidia card and are CPU-tied. Those results don't reflect the actual power of the cards to be honest, turning the overall performance chart to a laughing stock.
660>7870?
660Ti>7950?
680>7970 GHz?
Sorry, no way in hell

 

[W1zzard]

and are CPU-tied

only star craft ii is cpu limited, the rest scale fine on both manufacturers. amd just fails with their wow crossfire profile

 

[mirakul]

I have some questions for u Wizz
1. Did you test all the card in your overall performance chart with its latest driver at your time(314.22 and 13.5 beta 2)

2. Why were your AA settings so inconsistent through out the tests?

3. According to nVidia, the 314.22 boosts the performance mostly in
"GeForce GTX 680:
Up to 41% faster in BioShock Infinite
Up to 60% faster in Tomb Raider
Up to 23% faster in Sniper Elite V2
Up to 13% faster in Sleeping Dogs"

And they still lost to AMD in your BioShock Infinite, Tomb Raider, Sleeping Dogs tests.
So, why they suddenly triumphed in your overall performance chart?? If not because of the Starcaft 2 and World of Warcaft results?

Please enlighten me

 

[W1zzard]

I have some questions for u Wizz
1. Did you test all the card in your overall performance chart with its latest driver at your time(314.22 and 13.5 beta 2)

2. Why were your AA settings so inconsistent through out the tests?

3. According to nVidia, the 314.22 boosts the performance mostly in
"GeForce GTX 680:
Up to 41% faster in BioShock Infinite
Up to 60% faster in Tomb Raider
Up to 23% faster in Sniper Elite V2
Up to 13% faster in Sleeping Dogs"

And they still lost to AMD in your BioShock Infinite, Tomb Raider, Sleeping Dogs tests.
So, why they suddenly triumphed in your overall performance chart?? If not because of the Starcaft 2 and World of Warcaft results?

Please enlighten me

1) NVIDIA: 314.22 WHQL, ATI: Catalyst 13.3 Beta 3, HD 7990: 13.5 Beta 2
2) because that's what the games support
3) "up to 41%" means 0-41%. highly dependent on settings, card, resolution, aa
i havent done a detailed analysis of why amd turns out to be slower now. we did remove some benchmarks and added new. shouldn't be hard for you to figure it out.

 

[mirakul]

Thanks for your response, Wizz
3. nVidia confirmed that 314.22 driver has the most impact in those listed games, and nVidia's card still lose to AMD in them. So the driver was not the reason why nVidia got ahead overall. It was pointed out in my above post that the boost was from the new games added to your benchmark, which all come from Blizzard and favor nVidia cards (FYI, the recent patch to WoW is a nightmare for AMD users)

So again, I'm asking about the fairness of your bench.

 

[W1zzard]

Thanks for your response, Wizz
3. nVidia confirmed that 314.22 driver has the most impact in those listed games, and nVidia's card still lose to AMD in them. So the driver was not the reason why nVidia got ahead overall. It was pointed out in my above post that the boost was from the new games added to your benchmark, which all come from Blizzard and favor nVidia cards (FYI, the recent patch to WoW is a nightmare for AMD users)

So again, I'm asking about the fairness of your bench.

We added Crysis 3, Tomb Raider and Bioshock Infinite. All games that are bundled by AMD.
StarCraft II was only temporary removed because they released the addon which caused all older scores to get invalidated.

 

[Animalpak]

which one you will get if you have the money ? Look at the price of AMD !! CRAZY !!

 

[W1zzard]

which one you will get if you have the money ? Look at the price of AMD !! CRAZY !!

http://img.techpowerup.org/130429/Immaginefe.jpg

it's just a placeholder for pre-orders. cards won't be available for at least another week

 

[the54thvoid]

Thanks for your response, Wizz
3. nVidia confirmed that 314.22 driver has the most impact in those listed games, and nVidia's card still lose to AMD in them. So the driver was not the reason why nVidia got ahead overall. It was pointed out in my above post that the boost was from the new games added to your benchmark, which all come from Blizzard and favor nVidia cards (FYI, the recent patch to WoW is a nightmare for AMD users)

So again, I'm asking about the fairness of your bench.

I think you need to take a wee seat and think about how worked up you are getting. When you say a game shouldn't be included because it's unfair against AMD, that's just tough. Blizzard games are massively popular and to not include them is worse. If anything it highlighted AMD's poorer developer relations, something they are now rectifying. Also, in a single game, the AMD card is demolished by the GTX 680 which skews the final performance summary. I don't necessarily think that this is statistically accurate but all the graphs are there for you to see and if you choose to look at individual games, as you should, you will see where the cards strengths are.

Furthermore, if you want to truly gauge a cards performance, use lots of different sites to examine a global result. I'm afraid you will find most sites tend heavily towards Nvidia cards still. TPU was one of the first places to put a 7970 GHz above a 680 in the performance charts for BF3 so kudos for that surely?

Finally, it's massively disrespectful to come onto a site as a new person (though I've got a feeling you're not new at all) and effectively criticise one of the more respected technical reviewers and software writers around the tech sites at the moment. It shows a complete lack of understanding and also tremendous naivety regarding the overall purpose of TPU. W1zzard does more than just review gfx cards for a living - that is just a small but unfortunately highly fevered area of tech TPU does.

I am no moderator and have no vested interest in TPU other than believing that you should take a step back from your own highly opinionated stance and breath in the overall impressions and opinions voiced in what is after all just a single review. You have clearly stated and shown your opinions and you are entitled to them but that does not make them valid or worth incessantly smacking W1zz around the head with.

Try, if possible to not see the review as a competition between AMD and Nvidia and instead as an objective review of the 7990 against it's peers. W1zz is not alone in his underwhelmed response to the 7990, many others are too. And for balance, obviously some sites think it is awesome in all regards.

But to finally hammer the point one more time, if the world likes Blizzard games, it would be professionally remiss to neglect them from a review, regardless of how poorly coded they are for AMD cards. And remember, Tomb Raider hobbled Nvidia on release. The code given to Nvidia was not the final release code. So that was very unfair to Nvidia but the game is still in the review, fair or not.

 

[mirakul]

Thanks for your lecture, guys, and a special thanks to W1zz's great patience.

From the statistic perspective, the fact that an observation changed the overall result is still tickling me. However, I get what you guys wanna give here.

I'm looking for your next review, W1zz!

Cheers

 

[Nihilus]

Seems that tri-fire is AMDs strong side. Tweaktown has a review of a 7990 + 7970 in trifire. It had great scaling and matched or beat SLI titans in most cases. Tom's had a trifire 6870 review way back when that had great scaling as well. Any chance Wizzard will follow suit?

 

[Ravenas]

When are the OEMs going to release this card? What is the hold up?

 

[qubit]

it was present out of the box, all the time, no matter which game was running, and didn't change over the course of my testing.

That would totally drive me nuts, even for a minute. What a fail.

 

[HumanSmoke]

When are the OEMs going to release this card? What is the hold up?

Was supposed to be the 7th May for retail availability. Judging by the lack of cards at Newegg, and all of three verified owner reviews, I'm guessing that the cards must be hand made.

 

[Ravenas]

Does anyone know if they have resolved the coil noise? At $599 plus 8 games it's really hard to turn this down.