AMD Radeon HD 7990 Graphics Card Reviews

[kuroikenshi]

-pb

:x 2

[SKYMTL]

I have no doubt that is what you saw. I'm just saying thats not what I saw with 12.7. Its probably not a universal issue.

Likely not. Issues seem to pop up from one config / driver install to another. AMD's drivers have improved but they are still very, very fickle.

[boxleitnerb]

Perhaps because it runs like total balls on *anything*. At 2560x1600 its a bust on all cards and at 1920x1200 30fps is a struggle in many areas of the game. Few sites test metro with both AA and DOF. If its not realistic realword, whats the point?. May as well ask why don't you test everything with 8x super sampling...

This review was about SLI and CF and the 7990. In 2560x1600, very high (but no DoF and no AA) all cards pulled at least 40fps min and 60fps avg. There is still some headroom there.

This has always been the test and it hasn't changed. I've left it this way simply because Metro 2033 will soon be replaced and re-benchmarking every card on it would be an absolute waste of time. It will likely be replaced by Max Payne while we'll likely also add another off-brand title like Wargame to ensure a balance of flagship titles plus hardly-benchmarked titles.

Fair enough. Maybe with the next Metro 2013 you can bench differently?

[SKYMTL]

Fair enough. Maybe with the next Metro 2013 you can bench differently?

That really depends. I don't want to modify detail settings to a point where the game runs like a slideshow on even the highest end configurations.

[boxleitnerb]

Well at least for SLI/CF then. You're running 60+ fps already with those setups in 1600p at the highest setting in your test.

[BababooeyHTJ]

Likely not. Issues seem to pop up from one config / driver install to another. AMD's drivers have improved but they are still very, very fickle.

I had all sorts of issues with Kepler. Performance issues that I didn't have with Fermi, vsync issues, problems with 3d vision. It was awful. So far Tahiti has been a much smoother experience for me.

Nvidia isn't perfect either. Their recent vsync issues afaik are still ongoing. The TDR issues were very public. They released two drivers that killed video cards a couple of years back. There was the 460 high dpc latency issue that took forever to clear up. Its not like the Nvidia is without issues of its own.

[SKYMTL]

I had all sorts of issues with Kepler. Performance issues that I didn't have with Fermi, vsync issues, problems with 3d vision. It was awful. So far Tahiti has been a much smoother experience for me.

Nvidia isn't perfect either. Their recent vsync issues afaik are still ongoing. The TDR issues were very public. They released two drivers that killed video cards a couple of years back. There was the 460 high dpc latency issue that took forever to clear up. Its not like the Nvidia is without issues of its own.

I never said they were perfect. However, a question was asked. I answered.

[boxleitnerb]

SKYMTL, I have two suggestions for future reviews:

Could you post a video for each of the benchmarks you're doing? So one can clearly see what exactly was benched and how.
Maybe you could change the way you're measuring power consumption. First, these are peak values and thus have little relevance. Secondly, real gameplay should be used, not synthetic benchmarks and not just one game but more. If using all of them is too much work, maybe 3 or so. And finally, would it be possible to measure the power consumption of the card itself, not the whole system? When a card renders faster, the CPU has to work faster feeding the card, using more power. That says little about the card itself and might skew the values.

[DeathReborn]

SKYMTL, I have two suggestions for future reviews:

Could you post a video for each of the benchmarks you're doing? So one can clearly see what exactly was benched and how.
Maybe you could change the way you're measuring power consumption. First, these are peak values and thus have little relevance. Secondly, real gameplay should be used, not synthetic benchmarks and not just one game but more. If using all of them is too much work, maybe 3 or so. And finally, would it be possible to measure the power consumption of the card itself, not the whole system? When a card renders faster, the CPU has to work faster feeding the card, using more power. That says little about the card itself and might skew the values.

I agree on the videos but maybe an article on it's own where you can go into detail about how you benchmark, why you chose what you chose & then when you update your benchmarks just update the article and link it in your reviews for those who want to know the ins & outs of benchmarking. Maybe showing CPU usage when measuring power consumption would help with that side, so we can see which cards are making it work more.

[boxleitnerb]

An extra article is a great idea!
As for the power consumption, I would go all the way. All intermediate steps are just too error-prone and would influence the results. It would require some work or purchase of equipment (measure the power at the PCIe slot and the dedicated power cables and add them up), but I'm sure it would be worth it. It could be done only for reference cards to minimize the amount of work necessary.

[SKYMTL]

SKYMTL, I have two suggestions for future reviews:

Could you post a video for each of the benchmarks you're doing? So one can clearly see what exactly was benched and how.

This is already done. Visit the Benchmark Sequences page of reach review.

Maybe you could change the way you're measuring power consumption. First, these are peak values and thus have little relevance. Secondly, real gameplay should be used, not synthetic benchmarks and not just one game but more. If using all of them is too much work, maybe 3 or so. And finally, would it be possible to measure the power consumption of the card itself, not the whole system? When a card renders faster, the CPU has to work faster feeding the card, using more power. That says little about the card itself and might skew the values.

Real gameplay will never be used due to the fact that CPU power consumption distorts the results. However, maybe in the future average power consumption will be used.

[boxleitnerb]

Cool, very good!

Okay, that is also my concern. But why don't you measure the card directly, then as some other sites do?

[SKYMTL]

Cool, very good!

Okay, that is also my concern. But why don't you measure the card directly, then as some other sites do?

We do not have the tools to do that unfortunately. Also, some sites (not all) use a simple clamp meter over the PCI-E connectors. This approach will give HIGHLY inaccurate results.

[TuKo]

Hi guys!

I've just RMAed my Asus 5970. Asus replied I'll have to wait until 26.10.12 to get a replacement because actually they don't have any product to make an exchange. Is it the sign that Asus is coming with a 7990 too? Time will tell..

[slaveondope]

Hi guys!

I've just RMAed my Asus 5970. Asus replied I'll have to wait until 26.10.12 to get a replacement because actually they don't have any product to make an exchange. Is it the sign that Asus is coming with a 7990 too? Time will tell..

I would not count on it. The PowerColor is a custom designed, limited edition card not likely to be shared with Asus or others.
If you need another 5970 give me a month I'll have a dozen for sale.