AMD: Sea Islands R1100 (8*** series) Speculation/ Rumour Thread

[lanek]

If this is true, the next XBOX is going to quite expensive in relation to its competition. Powerful, yes, but also costly.


Right. I agree that there would be additional benefit from going to 48 ROPs -- architectural design is a massive game of diminishing returns -- but they way I understand it is that it would negate their decision to go asymmetrical. If the point really was to feed the ROPs with more bandwidth, going to 48 ROPs while keeping a 384 bit bus would be a reversion to their older philosophy.

I wonder if AMD or Nvidia will ever hit the holy grail of having a 7 Gbps bus. I was rather surprised by Nvidia's accomplishment of being the first of the two to hit 6 Gbps, considering the historical precedent. It seems to be increasingly difficult to bump speeds higher, but it would be a great thing for them to achieve; provided that it wasn't too resource intensive, of course.

I will not say 6Gbps was an achievement, it was just to choose the speed of the GDDR5, who if i am correct are exactly the same chips used by both AMD and Nvidia ( And both use memory rated at default 6Gbs )... and all retail overclocked AMD 7970 have 6gbs.. ( both of my cards are at 6GBs ( 1500mhz )...

If you compare the bandwith of the 7970 vs 680.. Its another story.-.. I just think AMD was believe 384bits bus + the memory speed choosed by them was enough for the reference version.( + it was safe in case Hynix cant supply anymore the same GDDR5 ) In reality, Nvidia have not really the choice with a 256bits memory controller to push the memory speed for dont be too much down on high resolution.

 

[CarstenS]

I was rather surprised by Nvidia's accomplishment of being the first of the two to hit 6 Gbps, considering the historical precedent.

This historical precedent is less pronounced than you think.

SDR: Ati won
DDR: basically a tie
gDDR2: kind of one-of-a-kind, so Nvidia
GDDR3: Nvidia won
GDDR4: see gDDR2, but ATi/AMD was the only one this time around
GDDR5: At first clearly AMD... which probably is the assumed historical precedent.

 

[lanek]

This historical precedent is less pronounced than you think.

SDR: Ati won
DDR: basically a tie
gDDR2: kind of one-of-a-kind, so Nvidia
GDDR3: Nvidia won
GDDR4: see gDDR2, but ATi/AMD was the only one this time around
GDDR5: At first clearly AMD... which probably is the assumed historical precedent.

Well for GDDR4, AMD had a lot of advance lol ... X1950XTX GDDR4 .. ( I have own 1 + a X1900XTX in CFX ( what a dream for overclocking at this time )

 

[Kaotik]

This historical precedent is less pronounced than you think.

SDR: Ati won
DDR: basically a tie
gDDR2: kind of one-of-a-kind, so Nvidia
GDDR3: Nvidia won
GDDR4: see gDDR2, but ATi/AMD was the only one this time around
GDDR5: At first clearly AMD... which probably is the assumed historical precedent.

Take into account that ATI/AMD moved away from GDDR3 on highend far earlier than nVidia did, giving nVidia "more time to win"

 

[silent_guy]

I think it's fanboy ignorance to assume that either side has or doesn't have the technical competency to design something when the other doesn't. (Or to be fundamentally unable to attain good yields if you will. ) Time to market, cost considerations, mispredicted marketing decision etc are much often a factor than not having the smarts to do so.

Well, at least for companies that have survived/thrived in such a competitive space for so long. Nvidia fans may look down on AMD for one reason and AMD fans may do the opposite, but I'm 100% convinced that they both respect each other for what they are technically capable of. (Only the paranoid survive etc.) The 2 have been leapfrogging each other time and again for a reason. And if they really fall behind in one or the other field, they can simply organize a hiring raid to correct...

 

[gkar1]

And a OEM would not be too happy if they launch a new line-up featring the mighty 8***, if AMD launches new retail 9*** shortly after. This rebranding only makes sense, when the real refresh is either cancelled or not out before 3Q13.

Do people ever bother to read news anymore before waxing self righteousness?

It's not AMD doing this to deceive anyone. It's a new calendar year OEM's demand new product names because they are refreshing their product lines. AMD obliges them.

http://www.anandtech.com/show/6570/amds-annual-gpu-rebadge-radeon-hd-8000-series-for-oems

And before you claim it's only AMD doing this to deceive the defenseless OEMs and consumers by extension, Nvidia just announced their batch of rebrands with more sure to come:

http://www.anandtech.com/show/6579/...730m-and-geforce-710m-partial-specs-published

OEMs want to do yearly updates (regardless of where the technical product cycle really is), so when the calendar doesn’t line up with the technology this is achieved through rebadges of existing products. In turn these OEMs put pressure on component suppliers to rebadge too, so that when consumers compare the specs of this year’s “new” model to last year’s model the former look newer. The end result is that both AMD and NVIDIA need to play this game or find themselves locked out of the OEM market.

 

[CarstenS]

Take into account that ATI/AMD moved away from GDDR3 on highend far earlier than nVidia did, giving nVidia "more time to win"

Depends on what and how you want to count, doesn't it? But it's not changing the fact, that there's no such thing as a fundamental historical inability from either side to design fast memory controllers.

 

[RedVi]

Given that the bus width increased by 50% over Cayman, I don't think increasing the DRAM frequency was worth possibly; risking supply constraints, increasing power and heat and leaving less headroom for improvement with future revisions. Sure, launching with the 7970 GE in January would have seen things go quite differently for AMD, but I don't see the increased DRAM frequency as being the main reason for that products superiority over GTX 680. There are a few reviews out there of cards with 1GHz core and 5.5GHz mem that offer a performance delta over the vanilla 7970 very close to what the 7970GE does.

 

[Homeles]

Depends on what and how you want to count, doesn't it? But it's not changing the fact, that there's no such thing as a fundamental historical inability from either side to design fast memory controllers.

That wasn't what I was suggesting at all. From what I understand, AMD (or really, the former ATI) has a big hand in the past few and upcoming GDDR specifications. They've had a bit of an advantage there.

And before you claim it's only AMD doing this to deceive the defenseless OEMs and consumers by extension, Nvidia just announced their batch of rebrands with more sure to come

It's just how the industry is. The people complaining about it as if it were something new is rather laughable. Then there's the fact that it doesn't even affect them. It's OEM only, for crying out loud.

 

[CarstenS]

From what I understand, AMD (or really, the former ATI) has a big hand in the past few and upcoming GDDR specifications. They've had a bit of an advantage there..

Yep, that's true. They've been very active in the committees and standardization organizations, possibly giving them a headstart in timeframe.

 

[seahawk]

Do people ever bother to read news anymore before waxing self righteousness?

It's not AMD doing this to deceive anyone. It's a new calendar year OEM's demand new product names because they are refreshing their product lines. AMD obliges them.

http://www.anandtech.com/show/6570/amds-annual-gpu-rebadge-radeon-hd-8000-series-for-oems

And before you claim it's only AMD doing this to deceive the defenseless OEMs and consumers by extension, Nvidia just announced their batch of rebrands with more sure to come:

http://www.anandtech.com/show/6579/...730m-and-geforce-710m-partial-specs-published

These new OEM products need to have a shelf life. An OEM interested in the 8970 would be hardly pleased to launch this now and have AMD come out with the 9970 in early April.

 

[Helmore]

These new OEM products need to have a shelf life. An OEM interested in the 8970 would be hardly pleased to launch this now and have AMD come out with the 9970 in early April.

That's why we won't see a 9970 in April, but a (proper) 8970. Or in May or June or something....

 

[Malo]

You can't have an OEM 8970 and a consumer 8970 completely different cards and generations.

 

[Kaotik]

You can't have an OEM 8970 and a consumer 8970 completely different cards and generations.

Why not? nVidia certainly haven't had issues with same name products being different based on wether it's OEM or retail.

 

[Homeles]

Or having one model number for completely different products in general, and not necessarily OEM vs. retail. Recent case in point: GTX 555m.

 

[CarstenS]

Why not? nVidia certainly haven't had issues with same name products being different based on wether it's OEM or retail.

They should have had though.

 

[Kaotik]

They should have had though.

Obviously, and no renames/rebrands ever should have taken place anywhere, but that's not the reality we live in sadly, and in this reality the companies get away with these.

 

[RedVi]

At the very least they should move every model up by +10. They made allowances for this with the 7700 series, at least.

 

[Alexko]

I don't know what different models of GT 555M may exist, but as far as I'm aware, OEM and non-OEM versions of a product have always been at least in the same ballpark.

 

[Kaotik]

I don't know what different models of GT 555M may exist, but as far as I'm aware, OEM and non-OEM versions of a product have always been at least in the same ballpark.

At least these are different GT 555M's
•144 cores 709MHz (GF106), 128Bit GDDR5
•144 cores 590MHz (GF106), 192Bit DDR3
•144 cores 590MHz (GF106), 128Bit DDR3
•96 cores 753MHz (GF108), 128Bit GDDR5
•144 cores 525 MHz (GF116), 128 Bit DDR3

Wether performance of card X's retail and OEMs is in the same ballpark is hard to say, since OEM versions benchmarks are hard to come by, but just from the latest gen at least GTX660 and GTX660 OEM use different GPUs and different configurations