AMD Radeon HD 9970 Specifications Leaked – Twice as fast as GTX 780 (ChipLoco rumor)

[BallaTheFeared]

It happens a lot, because it's a fair assessment.

So long as you list price difference, which in the case of many aftermarket 780s there is none, no harm no foul.

It's a reviewer's job to give an idea of the current market.

 

[BallaTheFeared]

Just like there was no reference 7970 GHz, and yet you didn't hear any of that foolishness when they were comparing it to the reference 680s.

Let's try to keep things if anything, consistent.

 

[Saylick]

Just like there was no reference 7970 GHz, and yet you didn't hear any of that foolishness when they were comparing it to the reference 680s.

Let's try to keep things if anything, consistent.

A great way to do that is to keep reviews reference vs. reference and aftermarket/OC'd vs. aftermarket/OC'd.

In my opinion, you're justifying the use of aftermarket cards in place of reference cards so long as their pricing is the same. From the perspective of a consumer, I find this mentality to be very understandable - "Why buy the reference, non-factory OC'd card when I can get the same thing at the same price, but faster?" is the thought that runs through the consumer's mind. However, from a reviewers perspective, using a consistent basis is the key to a good review. Prices change, some cards have better coolers and may thus OC better, some cards are known to have binned chips, etc. Point is, the multitude of variables is vast, but if there is one thing that can be certain, it's that there is typically a reference model which is accepted by nVidia/AMD that can be used as a baseline. Any variances beyond this baseline are additional variables which may introduce bias into the review. It is this basis which should be used in a review for a launched card.

As for the existence of OC'd 780's which are relatively the same price as a normal 780, I do not believe it is fair to include these cards in an unveiling review. These cards may be factory OC'd but remember that nVidia/AMD set specific stock clock speeds that these cards should be run at. You might think that if all 780's are capable of hitting 10% higher than their stock clocks then 780+10% OC should be the new baseline for all 780s, but that isn't true. For nVidia to have mandated such a scenario in the first place would have lowered yields and/or raised the launch price. What you're looking at, in reality, is third party companies buying stock chips and OC'ing them manually so they may gain a leg up on their competition at the expense of taking on that additional risk of buying lemon chips that do not OC as high as they'd like. In other words, if Company X buys 1000 780 chips from nVidia, all of which are capable of meeting the minimum stock speed requirements, you're looking at <1000 chips that are suitable for 780 ACX cards. This is where the premium comes in. Sure, the 780 ACX isn't all that more expensive than the stock 780. This could be because EVGA is selling the card with lower margins, OR that there is a chance that the 780 chip really does guarantee an +X% OC, enough so that X amount of stock chips can more or less be used to produce X amount of 780 ACXs. Again, this is all unknown in the eyes of anyone outside of said company. Therefore, we stick to what is known to all and what is guaranteed: the stock clocks.

 

[Osjur]

So if its fair to compare overclocked models to a stock as some of you are saying then why didn't every site test highest factory clocked 7970 against 680 when it launched...

Stock vs Stock and OC vs OC is the way to go. Otherwise the review is skewed or biased.

EDIT: or they have to take the price and power / heat into consideration as well

Just like there was no reference 7970 GHz, and yet you didn't hear any of that foolishness when they were comparing it to the reference 680s.

Let's try to keep things if anything, consistent.

But many sites tested with reference 7970ghz clocks even if they had factory overclocked version in their hands.
And AMD actually did send reference cards to some sites even if that model never came into retail market.

 

[wand3r3r]

Just like there was no reference 7970 GHz, and yet you didn't hear any of that foolishness when they were comparing it to the reference 680s.

Let's try to keep things if anything, consistent.

Well, if there's one thing consistent, it's the love for a manufacturer from some (I guess).

 

[RussianSensation]

Hawaii is flying in COH2.

More benches: http://videocardz.com/45753/amd-radeon-r9-290x-slightly-faster-gtx-titan

 

[krumme]

Like the "current market" is an objective assessment. Lol.

If amd marketing personel - the good nice guys - reads threads like this it should be pretty obvious letting reference cards to the market in this situation is plain stupid and yet another waste of shareholders money. It takes a huge bang to change a fan cognitive system. Thats what defines brand value. How many years does it take to understant that?

 

[Tempered81]

The power is obv 6 for VDDC + 2 for MVDDC - how can these numbskulls write incorrect specs when there is already a picture of the PCB? lol. Check out the reference capacitor count too - more than any other ref card... 4870, 285, 5870, 7970, titan, etc. Power delivery looks good, but hawaii die probably has higher transistor density & it is AMD's largest ever, so it likely has cramped ring bus & IMC to squeeze in shader count + quad front end + 512bit worth of mem controllers, so it's gonna be harder to keep higher clock - would be my guess. Advancements in lithography will probably make up for it. Looks like a doubled up Pitcairn, but did they reduce the DP? I'm still expecting similar to 7970/780/Titan OC headroom. The PWM & VRM looks strong, depends voltage control is easy for us and not blocked. If there is a boost software in the bios, can prob bypass via custom bios for high OC. Perhaps clock-for-clock it will equal GK110 Titan, and probably lose to full GK110, depending on how much a high memory clock boosts Hawaii performance numbers, and also how high we can get the memory without mods.

 

[BallaTheFeared]

Well, if there's one thing consistent, it's the love for a manufacturer from some (I guess).

Indeed.

Just look at how many people were upset over comparing it to aftermarket 780s, but none of them cried foul when AMD failed to produce a reference 7970 GHz.

 

[Black Octagon]

Indeed.

Just look at how many people were upset over comparing it to aftermarket 780s, but none of them cried foul when AMD failed to produce a reference 7970 GHz.

I did. I found both to be upsetting. Happy?

 

[BallaTheFeared]

Ecstatic

 

[Black Octagon]

Hahaha. Anyway, I hope for all our sakes that these 'benchmarks' are real. Lots of people on these and other boards seem to have already decided to sell their existing cards to jump on this new generation (!)

 

[wand3r3r]

I've been thinking the same thing. There's been both excited and bitter tears (oddly) and we don't even know if the information is real much less the correctness of it. It's fun to speculate but some are too attached to a brand.

 

[brandonmatic]

According to the Videocardz link someone posted earlier, the leaked benchmarks may be based on higher-clocked, non-reference speeds. In which case these benchmarks should perhaps be compared to the non-reference overclocked 780s.

The leaked review you&#8217;ve seen yesterday shows the performance with non-reference clocks. It&#8217;s worth noting that R9 290X will have a Turbo Mode. That said we believe 1020 MHz (1GHz) clock was obtained with this mode enabled. The real clock is lower (around 900 MHz). Without Turbo R9 290X will be slower than TITAN.

http://videocardz.com/45837/amd-hawaii-gpu-2816-stream-processors

 

[Saylick]

According to the Videocardz link someone posted earlier, the leaked benchmarks may be based on higher-clocked, non-reference speeds. In which case these benchmarks should perhaps be compared to the non-reference overclocked 780s.

http://videocardz.com/45837/amd-hawaii-gpu-2816-stream-processors

Something about those specs leads me to think that 290X is not a fully utilized die. 44 ROPs is uncommon for a fully enabled die; if those numbers are true, I wouldn't be surprised if the full die is actually 3072 SP, 192 TMU, 48 ROP.

 

[rgallant]

According to the Videocardz link someone posted earlier, the leaked benchmarks may be based on higher-clocked, non-reference speeds. In which case these benchmarks should perhaps be compared to the non-reference overclocked 780s.

http://videocardz.com/45837/amd-hawaii-gpu-2816-stream-processors

too bad nv didn't make the "non-reference overclocked 780s." small detail , but I could sell the same ,custom cooler and bios.

 

[Erenhardt]

According to the Videocardz link someone posted earlier, the leaked benchmarks may be based on higher-clocked, non-reference speeds. In which case these benchmarks should perhaps be compared to the non-reference overclocked 780s.

http://videocardz.com/45837/amd-hawaii-gpu-2816-stream-processors

Yep. Reviewers should disable Turbo in Hawaii during benchmarking to avoid giving AMD card unfair advantage of available core clock headroom over Nvidia products.

Or if they don't that should be compensated with LN cooling + heavy OC for GTX780.

Anyway, GK110 is bigger chip than hawaii, so buy buying 780 you get more!
On top of that, AMD doesn't have cooler with transparent section above the heatsink, which is a huge bonus for any creative designer - such an inspiration comes from looking the collecting dust.

And last, but not least, AMD drivers for hawaii are terrible! Nothing to see here. Go, better buy Titan (eventually gtx780) while supply lasts

 

[ocre]

According to the Videocardz link someone posted earlier, the leaked benchmarks may be based on higher-clocked, non-reference speeds. In which case these benchmarks should perhaps be compared to the non-reference overclocked 780s.
"According to the Videocardz link someone posted earlier, the leaked benchmarks may be based on higher-clocked, non-reference speeds. In which case these benchmarks should perhaps be compared to the non-reference overclocked 780s."

http://videocardz.com/45837/amd-hawaii-gpu-2816-stream-processors

Every bit of this is suspect. If it was out of the box faster than titan then the remarks in the forbes article would have been totally different. The results dont add up at all. They are no correct, i am sorry. A lot of people are getting hyped up and could really really be let down. Videocardz says the results "may" be from a higher clocked, non reference speeds because why?
Also if this was results from pushing higher clocks then the power consumption would be much higher. I am not trying to start a fire here but please dont let your wild speculations over hype the situation like this. All arrows point to null and void. There is no reason for videovcardz to have such remarks if the results where realistic. There is no reasoning behind for the forbes comments if the 290X was out of the box faster than titan. And there is no way these results are from an overclocked 290x. Think about it for a minute. The cap on performance is TDP and AMD wants to get up there as high as they can and it will use at least as much power as titan. If this is an overclocked card with such fantastic power consumption then at stock it would use less than the 7970ghz and be less powerful than the titan.

Logic tells us that AMD is gunning as high as they can and their performance will be capped by power consumption. It will at least use as much power as titan. The card will be very powerful and I am sure i expect 780 class performance. But do not over-hype this as it can only do bad things. Having your expectations already sky high gives a larger chance at dulling the entire refresh if it doesnt meet those limits. Dont hype this up so high. Just by AMDs own comments, its clear that this is not aiming at titan performance so lets come back down on our projections.
Besides if it does come as an out of a box titan killer, not having expectations that high would only fuel the excitement. It will be all the better....trust me

 

[Erenhardt]

Every bit of this is suspect. If it was out of the box faster than titan then the remarks in the forbes article would have been totally different.

There is no reasoning behind for the forbes comments if the 290X was out of the box faster than titan.

And where does Forbes interview state that Hawaii will be faster/slower than Titan?

Besides if it does come as an out of a box titan killer, not having expectations that high would only fuel the excitement. It will be all the better....trust me

True. Better be positively shocked than overhype and be let down by the fact that hawaii is not alien technology.

 

[Black Octagon]

^ hear hear. ALWAYS wait for real reviews

 

[3DVagabond]

FWIW 9-22-13

Hawaii GPU Specifications AMD Hawaii GPU

It is more or less confirmed that Hawaii will have 2816 cores. The exact configuration is:

44 GCN Compute Units
2816 Stream Processors
176 Texture Mapping Units
44 Raster Operating Units
So as you can see there are less ROPs than we thought. Sources are also saying that the GPU is around 420 mm2. The leaked review you’ve seen yesterday shows the performance with non-reference clocks. It’s worth noting that R9 290X will have a Turbo Mode. That said we believe 1020 MHz (1GHz) clock was obtained with this mode enabled. The real clock is lower (around 900 MHz). Without Turbo R9 290X will be slower than TITAN.

One other thing. The fastest Radeon R9 290X might cost $650 USD, that’s the latest information on the price.

http://videocardz.com/45837/amd-hawaii-gpu-2816-stream-processors

Why are they wanting to compare without turbo? "Without Turbo R9 290X will be slower than TITAN." Is the faster speed they're referring to of Titan without boost?

 

[3DVagabond]

One thing: If Hawaii needs 2816/2880Sps clocked at 1020Mhz to beat Titan then it was not a good deal. Hope it was 2560Sps clocked at 1Ghz to do this(Architecture efficiency and overclocking potential capabilities matter more than ever this time).

The real 290X enemy is not the terribly overpriced Titan but upcoming Maxwell GTX 880. Will not worth if 290X defeat Titan today(as we seeing it gonna do) but lose to 880 3-4 months after.

I haven't read anything about new nVidia cards in 3-4 months. Where did you?

 

[skipsneeky2]

Maybe nvidia can come up with something like a gtx785,price it accordingly with the 290x and just have some major aggressive boost 3.0 function.I would love if nvidia got their balls off the floor,enabled some beast of a turbo mode that dramatically increases tdp on the fly depending on temperatures of course and apply higher then ever before boost clocks.

Perhaps my dream of this happening is way out there,but the folks spending $500+ on a card most likely will be buying two,or have more then amps on their psu to supply such wattage to a beast of a card but i feel the 290x might honestly be to much for a gtx785/titan ultra.

 

[Fastx]

Why are they wanting to compare without turbo? "Without Turbo R9 290X will be slower than TITAN." Is the faster speed they're referring to of Titan without boost?

IMO this a stock reference card out of the box stock which includes turbo mode in a stock reference card and is how it should be tested against other stock reference cards out of the box imo. To me NOT to test it with its factory out of the box stock turbo would NOT be the correct/right way, but its just opinion.

 

[RaistlinZ]

Maybe nvidia can come up with something like a gtx785,price it accordingly with the 290x and just have some major aggressive boost 3.0 function.I would love if nvidia got their balls off the floor,enabled some beast of a turbo mode that dramatically increases tdp on the fly depending on temperatures of course and apply higher then ever before boost clocks.

Perhaps my dream of this happening is way out there,but the folks spending $500+ on a card most likely will be buying two,or have more then amps on their psu to supply such wattage to a beast of a card but i feel the 290x might honestly be to much for a gtx785/titan ultra.

Shouldn't be too hard. Slap some 7,000Mhz RAM on the 780, bump up the clock to 1,050MHz or so and it should perform similar to the 290x.