AMD: R9xx Speculation

They may call it whatever they like (for instance Z999999) but it will remain the same speedwise. According to me Barts with 6800 series name is a stupid manouver if Cayman is the highest end chip. :LOL: :rolleyes:
 
x7xx were upper-mainstream GPUs with 128bit interface (X700, HD4700, HD5700)
x8xx were lower high-end GPUs with 256bit interface (X1800, HD3800, HD4800, HD5800)
x9xx were large monolitic enthusiast GPUs (X1900, HD2900)

Applying this scheme to Barts and Cayman I think it is more logical to expect, that Barts will be HD6800 and Cayman HD6900. x7xx name was never used for anything "better" than a 180mm² GPU with 128bit memory bus, so it doesn't make sense to call a 2xx mm²+ 256bit GPU HD6700...
 
Then again, perhaps Barts is "on average" 50% faster than Cypress. If so I'll certainly be surprised, and will recant everything I've said against naming it 68xx. I find this situation extremely doubtful however.

Regards,
SB

What if it's 25% faster, would that be good enough for it to be named 6870 or not?

I can see how this could be an issue for some and maybe not for others. If AMD has landed Barts XT 25% faster than Cypress (which I personally doubt, but you never know), it must be a difficult decision to decide whether it gets 7 or 8. Maybe that's why there has been so much discussion over it.
 
People who argue about gpu nameing schemes are silly. :!: Who cares :?:
Lets hope that for majority of people the cards name is the last thing to consider when buying a new graphic card.
 
ОМG!!! It confuses and people have to recheck the reviews. Why should I waste my time in reading some new articles if I'm used to know that 3800-4800-5800 is the highest end and now, all of a sudden, things change. WTF? Are they crazy? :devilish:
 
I buy my hardware base on performance and power use, not by whatever they decide to name it.

I think some people would be better off buying a game console, if this confuses them that much.
 
ОМG!!! It confuses and people have to recheck the reviews. Why should I waste my time in reading some new articles if I'm used to know that 3800-4800-5800 is the highest end and now, all of a sudden, things change. WTF? Are they crazy? :devilish:
Check performance "jumps" of 8800GTX->9800GTX->GTS250->GTS450 ;)
 
That's exactly what I try to point out. It confuses to such extent that the only right choice may be done after consulting with the reviews. ;) If you ask me which one of those cards, I will face some serious difficulties to decide. ;)
I don't want AMD following NVidia. :devilish:
 
I
And to Neliz for needing more room on the product naming scheme, there's far more room below 68xx then there is above it for products. x1xx and x2xx haven't been used much. x3xx and x4xx haven't even been used every product generation since the whole naming scheme reorganization started with 3xxx.

SB


No, no you missed the point. They need less room for discrete graphic product.. so high performing parts. And their naming scheme for discrete graphic must match the one of Nvidia (x9x x8x x7x x6x). They need more room for integrated graphic, in Llano and Ontario etc.
Actually there may be no Turks, as a discrete chip. Turks may be a Llano part of N.I.



I'm sorry but i must ask you a question: Will you ban a 6870@Bart with the performance of Cypress XT for 199$? :D
 
Here's some numbers to think about..

Redwood: 627 MT, 104mm², 400 ALUs, 20 TMUs, 8 ROPs, 128 bit, GDDR5
Juniper:, 1040 MT, 170mm², 800 ALUs, 40 TMUs, 16 ROPs, 128 bit, GDDR5

(Juniper - Redwood) = (170 - 104 = 66mm²) and (1040 - 627 = 413 MT) for 400 ALUs, 20 TMUs, and 8 ROPs.
(627 - 413) = 214 MT for Command Processor, Setup Engine, Tesselator, UVD, I/O, ...
So Juniper use 826 MT for 800 ALUs, 40 TMUs, and 16 ROPs.

24 SIMD = 384 SP 5D = 1920 ALUs, 96 TMUs, "38 ROPs" = (4.8 x 66mm² = 317mm²) + (38mm² x 1.5) for the "dual-engine" uncore = ~374mm².
30 SIMD = 480 SP 5D = 2400 ALUs, 120 TMUs, 48 ROPs = (6 x 66mm² = 396mm²) + (38mm² x 2) for the "tri-engine" uncore = ~472mm².

24 SIMD or 1920 ALUs is 20% more than Cypress, so this chip needs 20% more bandwidth with the same 256 bit bus.
If we look around we can find 6 GHz GDDR5: Elpida EDW1032BABG60F, Hynix H5GQ2H24MFR-R0C, Samsung K4G20325FC-HC03.
6 GHz is 20% more than 5 GHz, so a 1920 ALUs chip with 6 GHz GDDDR5 on a 256 bit bus seems viable.

A 30 SIMD chip at 800 MHz will deliver 3840 GFLOPs (41% more than Cypress) of peak performance with a ~260W TDP.
This chip may have a 384 bit bus (48 ROPs) with 4.8 GHz GDDR5, and 472mm² is sufficient for that. (R600: 512 bit, 420 mm²)
And that's without any architecture optimizations vs Evergreen. (Who's said Fermi killer ?!) ;)

24 SIMD for Bart, and 30 SIMD for Cayman ?
These new GPUs could supplement the Evergreen family.

PS: Antilles = 2xBart = 48 SIMD = 3840 ALUs = 5000 GFLOPS @ 651 MHz.
 
Last edited by a moderator:
I'm sorry but i must ask you a question: Will you ban a 6870@Bart with the performance of Cypress XT for 199$? :D

I'm afraid most people here are too infuriated with renaming rage to notice that this isn't going to be a typical round of renaming.
 
I'm afraid most people here are too infuriated with renaming rage to notice that this isn't going to be a typical round of renaming.

I blame NVIDIA! :D Seriously though, I think we should put this whole naming issue aside and get back to it when Barts is released and we actually know what its name is, and how fast it is.
 
Back
Top