HD 6990 Preview/Review

[-Boris-]

Hot, expensive and noisy. Actually I would like AMD to use their performance per watt and performance per mm² leadership to make one single larger chip to compete with nVidias high end.
At GTX 580s power consumption and die size AMD would have a winner. The dual GPU approach is getting out of hand.

[-Boris-]

how can you claim "scientific" and "recording to be played back on user's system with different audio config and volume" at the same time ?

He claimed "None Scientific"

[SKYMTL]

Bring me Eyefinity reviews...

580 SLI is a lot more expensive, though.

Where are the Eyefinity reviews?!? Testing these cards at 25x16 is a waste of time.

Read!

Page 18 & 19 have a load of games tested with Eyefinity & NVIDIA Surround, with and without AA Enabled.

[AAbenson]

...actually its not getting out of hand :
certain part of the population doesnt care about noise,power consumption,a $600 price tag as long as they get the "fastest".
...see the nVidia 480 and all the Intel Extreme editions in each processor generation - hot,power hungry and costly,neither product sold millions of units.
btw AMD knows the dual gpu market well enough since this is not their first dual gpu solution.
if you rely on your entire product line to consist of multi gpu solutions like 3DFX did at the time then youre in trouble.,if you have some "extreme" product which just gives you bragging rights then its completely a different story.

[RacingTurtle]

EK's cooling configurator have the full-cover block @ ETA 2/3 weeks for the Asus HD 6990.

[onethreehill]

In stock in the UK

HIS £474.99 ex VAT (£569.99 inc VAT)
http://www.overclockers.co.uk/showpr...odid=GX-033-HS

Sapphire £491.66 ex VAT (£589.99 inc VAT)
http://www.overclockers.co.uk/showpr...odid=GX-257-SP

[WaterFlex]

PowerColor Radeon HD 6990 CrossFire

[Lanek]

Read!

Page 18 & 19 have a load of games tested with Eyefinity & NVIDIA Surround, with and without AA Enabled.

Nice i was wait on your review.. going to read that now as i have a bit of time .

[Iconyu]

@Waterflex: That scaling is pretty good.

[SKYMTL]

Considering the prices other cards have retailed for, I don't think $700 is an extreme amount for the performance it delivers.

Remember:

8800 Ultra launched for $850
The X850XT for $675 in 2003

And so on.

[-Sweeper_]

u$600 buys you two HD6950, after the infamous HD6970 mod you get the same performance, less noise and less heat for u$100 less (relative to HD6990)

heck, u$20 more buys you two HD6970

HD6990 could be a hit for u$600, but U$700 sounds... well, lets see

[Oztopher]

This made me lol.

[Lanek]

One question SKYMTL, maybe i have miss it, the games benchmark are so done on 830 ? it's not a problem i was just not find the info at first maybe just have miss it.

[Mad Pistol]

I like the card's performance, but I would never buy one. It's just too expensive.

Also, I don't think the 6990 on a 16x pcie slot is a problem at all, as mentioned in earlier posts. And you definitely cannot consider each GPU as using a single 8x slot. That's just absurd. The data is definitely being sent to the interconnect chip. Because of that, it looks like the data is being partitioned out (or doubled, not sure which) to each GPU. It's still the same data from the CPU, but the PLX chip is routing the data each direction. I mean, it's just getting double the processing power, not double the data load.

I can see 2 of these cards working fine in crossfire (quad crossfire I guess) as long as you've got dual x16 lanes. If not, don't bother. I'm pretty sure x8/x8 will bottleneck this card.

[SKYMTL]

One question SKYMTL, maybe i have miss it, the games benchmark are so done on 830 ? it's not a problem i was just not find the info at first maybe just have miss it.

All game benchmarks are done at stock speed (830 / 5000) while in the overclocking section we go into the very, very minimal difference between 830 and the "OC" profile of 880.

There are some other things we go into there as well...like the pathetic overclocks one can achieve without moving the BIOS switch.

[SKYMTL]

The data is definitely being sent to the interconnect chip. Because of that, it looks like the data is being partitioned out (or doubled, not sure which) to each GPU. It's still the same data from the CPU, but the PLX chip is routing the data each direction. I mean, it's just getting double the processing power, not double the data load.

Basically, the layout is something like this:



The PLX chip incorporates 48 Gen 2.0 lanes. There are 32 lanes dedicated to the GPU interconnect (16 to each) while the other 16 are used for pathway communications with the PCI-E bus.

[hokiealumnus]

i like seeing temp and power consumption in gaming. furmark is nice and reliable and all that, but there are reasons why more roadblocks are being built to prevent such tests from working.

i also think the should post fps in furmark, aswell as temps and load to show efficiency.

With a sandy bridge system at stock and idling, Furmark power draw from the wall peaked around 496W. With the system loaded and running Vantage graphics tests I didn't see above 450W.

It remains to be seen if all cards have the same problems with coil whine during memory load. And it seems to be isolated to memory load and increase with memory frequencies.

My sample had that. All ATI cards I've tested have to some extent though. It was most audible in 3DMark03. Everything else I could tell it was there but not near as loud. Considering it happened mostly when pushing 1000-3000FPS (never heard it in game tests that I recall), I don't think it will be a common complaint from most users.

Considering the prices other cards have retailed for, I don't think $700 is an extreme amount for the performance it delivers.

Remember:

8800 Ultra launched for $850
The X850XT for $675 in 2003

And so on.

Agreed, plus it will most likely drop as time goes on. It's definitely not cheap, but those that want the absolute best card residing on a single PCB should already know they're going to pay a pretty penny for it.

In case anyone wants yet another perspective, feel free to check out my review at Overclockers. No Eyefinity unfortunately...just no monitors to use for it. You bench with what you have, not what you wish you had.

[Lanek]

All game benchmarks are done at stock speed (830 / 5000) while in the overclocking section we go into the very, very minimal difference between 830 and the "OC" profile of 880.



There are some other things we go into there as well...like the pathetic overclocks one can achieve without moving the BIOS switch.

Ok thanks,

Yes i have see, really interessant, cause this confirm what i was think the voltage on core are set for just under the 880mhz (870), they wanted really set the lower Vcore possible. ... anyway there still the possibility to don't use the switch and OC the voltage manually with 3th party tool if im not wrong. ( i think actual Afterburner is not compatible yet ).

On the other hand, in position 2, 975mhz is not so bad ( with default voltage )

About the "noisy pwm"; i have this too with my CFX when i OC the cards for bench, and in 3Dmark03 it's incredibly loud in the 2 first bench, remind me old 56K modems... But nothing in games.

[[NH]Woxxman]

how can you claim "scientific" and "recording to be played back on user's system with different audio config and volume" at the same time ?

I think it states pretty clear that it’s a "none scientific" test.

And it would be interested to hear hearing your own experiences with the card? I havent seen anyone else mentioning it yet and AMD hasnt gotten back to us yet about the issue.

My sample had that. All ATI cards I've tested have to some extent though. It was most audible in 3DMark03. Everything else I could tell it was there but not near as loud. Considering it happened mostly when pushing 1000-3000FPS (never heard it in game tests that I recall), I don't think it will be a common complaint from most users.

Thanks for the input, I don’t know if it’s good to hear that we aren’t the only one's having these problems or if we should get more worried...: P
I find it very strange that high profiled review sites haven’t brought it up though, guess it doesn’t happen on every card at least.

The thing is that I first thought that a water cooled HD 6990 would be pretty sweet. But if the coil whine is troubling with the fan drowning the sounds I can’t imagine how it would feel in a more silent system.
But hopefully it’s just our card that’s a bit more sensitive and that it won’t be that noticeable in normal usage. Guess we’ll find out in the long run.

[SKYMTL]

My sample had that. All ATI cards I've tested have to some extent though. It was most audible in 3DMark03. Everything else I could tell it was there but not near as loud. Considering it happened mostly when pushing 1000-3000FPS (never heard it in game tests that I recall), I don't think it will be a common complaint from most users.

There were many cases of this and some games where it was VERY bad. AvP and F1 2010 made the whine / hiss particularly audible over the fan noise.

[hokiealumnus]

There were many cases of this and some games where it was VERY bad. AvP and F1 2010 made the whine / hiss particularly audible over the fan noise.

Strange, I ran AvP too and didn't notice. I've got three Panaflo monsters on the radiator next to the bench station though. They were turned down to barely audible but maybe they overrode it. Perhaps it was so much quieter than high-FPS benches that I didn't notice, which is also entirely possible

EDIT - FWIW, I was a little surprised in hearing how loud it was when benching, but I distinctly remember being surprised by the 6970 too. Considering this is essentially two 6970's (with slower memory speed of course), the difference made sense. It doesn't sound great when it happens, but wasn't entirely unexpected after its single-GPU little brother.

[SKYMTL]

Or maybe it is a sample to sample difference.

[Lanek]

Or maybe it is a sample to sample difference.

Can be too PSU difference related ... i don't have it on some models of PSU and i have it on some other.

[ChinStrap]

The most beautiful PCB I have ever seen

[hokiealumnus]

Or maybe it is a sample to sample difference.

Touché.