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

Page 19 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Status
Not open for further replies.

tolis626

Senior member
Aug 25, 2013
399
0
76
How efficient do you think the reference cooler will be guys?I never owned a reference GPU to know,but I guess that aftermarket ones are quite better,eh?
That said,September 25 can't come soon enough.My exams finish that day (Good riddings biochemistry and anatomy!!) and Hawaii is announced.Patience...
 

BrightCandle

Diamond Member
Mar 15, 2007
4,762
0
76
If it really is faster than Titan this is a major step for AMD in the performance density of their parts, this suggests a major architecture change and not a simple refresh.

In the previous generation the 7970 GE was around about the same performance as the 680 (5% faster with the later clock speed bump) but the 7970 was 365mm^2 verses 294mm^2 of the 680. The Titan is 551mm^2 and only about 40% faster on average (sometimes its 90% faster) in performance than a 680. AMD is saying the new card is 30% smaller than Titan, making the new card have a die size of 385.7mm^2, an increase of just 5.7% over the 7970. Yet the claim is that this will deliver a performance boost of 35% or more in order to equal and then exceed Titan.

That is a significant jump in the performance density of AMD's card, far greater than any jump they have made in their recent history, far greater than any manufacturer of GPUs has made in quite some time. I have always felt the 7970 underperformed for its die size, a mere 5% quicker on average than a 680 with a refresh and clock speed bump and some 24% larger but at the same time it had the compute performance and the memory bandwidth the 680 never had, I can believe that costs 24% die size in the same way that 680 to Titan is an 87% larger die and only brings 40% extra performance. Such a jump for so little die size increase seems unlikely.
 

insertcarehere

Senior member
Jan 17, 2013
639
607
136
512 bit memory bus confirmed, 2560 or 2880 stream processors still up in the air?

Damn, 512bit memory interface, I guess AMD are actually starting to moving away from the "small die" strategy that started with RV670/770, and seem to want to compete for performance with Nvidia in the top end again. I guess they realise after these years that brand recognition is just as, if not more important than price/performance.
 

Skurge

Diamond Member
Aug 17, 2009
5,195
1
71
If it really is faster than Titan this is a major step for AMD in the performance density of their parts, this suggests a major architecture change and not a simple refresh.

In the previous generation the 7970 GE was around about the same performance as the 680 (5% faster with the later clock speed bump) but the 7970 was 365mm^2 verses 294mm^2 of the 680. The Titan is 551mm^2 and only about 40% faster on average (sometimes its 90% faster) in performance than a 680. AMD is saying the new card is 30% smaller than Titan, making the new card have a die size of 385.7mm^2, an increase of just 5.7% over the 7970. Yet the claim is that this will deliver a performance boost of 35% or more in order to equal and then exceed Titan.

That is a significant jump in the performance density of AMD's card, far greater than any jump they have made in their recent history, far greater than any manufacturer of GPUs has made in quite some time. I have always felt the 7970 underperformed for its die size, a mere 5% quicker on average than a 680 with a refresh and clock speed bump and some 24% larger but at the same time it had the compute performance and the memory bandwidth the 680 never had, I can believe that costs 24% die size in the same way that 680 to Titan is an 87% larger die and only brings 40% extra performance. Such a jump for so little die size increase seems unlikely.

Titan is 30% larger, making Tahiti about 23% smaller. So it will be 430mm^2
 

boxleitnerb

Platinum Member
Nov 1, 2011
2,601
2
81
For all you enthusiasts out there, better start praying there aren't more mem chips on the back of the PCB...

Highly unlikely that there are. 6 GB already doesn't make any sense (at least currently), 8 GB would be a total waste and just make the card more expensive.
 

Osjur

Member
Sep 21, 2013
92
19
81
512 bit memory bus confirmed, 2560 or 2880 stream processors still up in the air?

I know its my first post in these forums so you can (and should :biggrin take this with a grain of salt but some of you may know user called The Stilt (who happened to fix the issues with the slowness of x87 on AMD cpu's) says that it will have 2816SP's, (44x64) / 176 TMU's and 44 (+ 4) ROP's (44 normal ROP's and 4 TCC [clusters?]).
He also says that the die size of hawaii is around 420-430mm2 and compared it to die of Tahiti and got these calculations:

Tahiti:

SIMD = 21,6%
ROP = 7,4%
MCT / IO = 21.7%

Hawaii:

SIMD = 21,5%
ROP = 8%
MCT / IO = 12.9%
Memory controller seems to be packed much denser on Hawaii

I myself trust that guy because most of the time he has been right.

Hopefully the specs are true and we have a Titan killer in our hands.
 

Abwx

Lifer
Apr 2, 2011
11,167
3,862
136
If it really is faster than Titan this is a major step for AMD in the performance density of their parts, this suggests a major architecture change and not a simple refresh.


Yet the claim is that this will deliver a performance boost of 35% or more in order to equal and then exceed Titan.

Such a jump for so little die size increase seems unlikely.

Looking at a 7970 die we can see that computing
is devoted 70% of said 365mm2 , that s about 256mm2 ,
the rest is for UVD , memory controler and so on , hence
the added 67mm2 if dedicated only to the compute part
would increase its size by only 26% , unlikely at first glance
to increase perfs by 35% or so , but there s more to the story
according to Hardware.fr.

It is quite possible that Hawai use the same process
as Bonaire wich allow 11.8% higher density than Tahiti
and if it s the case the numbers above will be inflated
well above the 26% figure since it would allow Hawai
to have actualy 31% more transistors than Tahiti ,
wich , using the logic above , would yield 44%
bigger computing area.....


http://www.hardware.fr/news/13356/gpu-hawaii-20-plus-petit-que-gk110.html
 

Fastx

Senior member
Dec 18, 2008
780
0
0
FWIW posted today.

Well apparently the rumor was true, AMD Hawaii GPU has (supposedly) 512-bit interface and 4GB memory according to leaked photos.

AMD’s Radeon R9 290X equipped with Hawaii XT GPU is said to have at least 2560 Stream Processors. Some claim it will have 2816, which is not exactly confirmed. In fact the information we have indicates that the full Hawaii GPU will go straight to FirePro model, whereas the 2560 edition will be used with the flagship Radeon. That’s the rumor of course. I could not verify anything because AMD is very quiet about their new release.

It doesn’t take a lot of effort to count the memory modules. You can easily find sixteen 256MB modules with a total capacity of 4GB, but that’s not all, since the PCB could provide additional space on the back, with dual-sided memory, 8GB does it sound good? Nevertheless, 4GB memory size obviously suggest 512-bit memory bus.

http://videocardz.com/45704/amd-radeon-r9-290x-hawaii-gpu-pictured-512-bit-4gb-memory
 

BallaTheFeared

Diamond Member
Nov 15, 2010
8,115
0
71
Yeesh, I wonder what Nvidia is thinking right about now...

I'm guessing 20nm and Maxwell can't come soon enough for them.
 

Saylick

Diamond Member
Sep 10, 2012
3,385
7,151
136

PPB

Golden Member
Jul 5, 2013
1,118
168
106
You can decouple VRAM ammount from bus width, Nvidia has proven this in the past. So still you cant 100% confirm 512b bus.
 

Osjur

Member
Sep 21, 2013
92
19
81
IIRC random pictures on this internetz do not a truth make.

True, but it's one hell of fake (comparable to the famous "puppy") because there are pictures of the whole card and it looks pretty legitimate in my eyes.

Also the picture was leaked by DICE (who also leaked 7990) so this could be a case of controlled leak.

512bit could still be only 256bit though I really doubt that the would do that
 

raghu78

Diamond Member
Aug 23, 2012
4,093
1,475
136
Yeesh, I wonder what Nvidia is thinking right about now...

I'm guessing 20nm and Maxwell can't come soon enough for them.

TSMC 20nm is not ready. its looking more like Q3 2014 for 20nm GPUs. But if Apple has signed up with TSMC for 20nm then Nvidia and AMD have no other choice but to stand in queue behind Apple and pick the leftovers. Given that AMD is launching Hawaii in Oct 2013, their 20nm Pirate Islands flagship should be out roughly the same time next year. right in time for holiday 2014.
 

BrightCandle

Diamond Member
Mar 15, 2007
4,762
0
76
Even at 425mm^2 its only 16% bigger and the expectation is that will gain the card 35-40% extra performance to meet a Titan, more than that to beat it. That is still a really significant architectural change and performance boost, the likes of which seems unlikely. 425mm^2 means its going to be relatively expensive, not Titan expensive of course but its not going to be a $300 card.
 

Fastx

Senior member
Dec 18, 2008
780
0
0
The below is for those in this thread who haven't followed or seen SW's post.


One thing that comes to mind is Sushiwarror said about 2-3 months ago from specs he said he seen/has is it will be over 2500 sp's (2500-2900), larger than a 384 bit and now we are seeing talk of these specs really hitting lately just like he posted. I'm thinking between Sushiwarror and news hitting lately we are looking at a 512 bit card with around 2800 sp's as looking most likely now like he posted/said months ago.
 
Last edited:
Status
Not open for further replies.
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |