AMD: R9xx Speculation
- Thread starter Lukfi
- Start date
That was years ago. Maybe during the R520 era?
Subtlesnake
Regular
I believe the first Vista drivers are where we saw the OpenGL rewrite.
Anyone else seen this?
http://techreport.com/discussions.x/20195
I think it's the first time I've seen SP & transistor count for Turks/Caicos (Whistler/Seymour).
Blackcomb: 960SP, 1.7B, 680Mhz
Whistler: 480SP, 715M, 725Mhz
Seymour: 160SP, 370M, 800Mhz
(Mhz calculated from quoted gflops - like usual I don't really expect them to quite hit these in actual notebooks.)
The lineup imho has too many chips however. In particular, Whistler looks like it's a Redwood with Barts tweaks and an additional simd. There's not really that many more transistors compared to that predecessor (+14%) and hence probably die size is quite similar too, so I'm not sure why AMD is promoting that old one to HD6500M series instead of just dropping it - could still sell partially disabled Whistler parts with 400 (or even 320) SPs.
Blackcomb is Barts with 2 SIMDs disabled - interesting that AMD chose not to release a fully enabled part. In any case, it should be quite a beast, beating all currently released chips (including that monster GTX480M with its stellar TDP) easily (no doubt nvidia will counter with a less crippled GF104/GF114 than they currently are offering, or even GF110).
Seymour is the most interesting wrt speculation - the transistor count increase over Cedar is way too big for a simple increase of the simds to 16-wide. So either it just adds 2 more 8-wide simds for a total of four, or there's other changes gobbling up the transistors (could for instance have the faster Redwood/Juniper style frontend, don't forget Cedar was simpler there).
Also, I think it would have been a good opportunity to drop that anemic cedar-based HD6300M. Contrary to the past, I don't think it will be too popular since when equipped with ddr3 it will get beaten by Sandy Bridge IGP. Granted it has dx11 support, better drivers, probably much faster MSAA, better quality AF filter and so on, but that's probably not enough to really be an option instead of the IGP (considering cost and power). With gddr5 it might be faster but I still doubt it's worth it.
Speaking of gddr5, AMD is saying all chips support gddr5 or ddr3. I can only hope noone takes that literally and will offer totally stupid things like HD 6800M with ddr3 (which would likely be slower than HD 6700M with gddr5). That's also something I really wish AMD would finally enforce - products with different ram type (if you REALLY want to offer them) just NEED a different name (the table is only using series names so can't judge that yet but considering the past fearing the worst). Slightly different clocks using the same name might be ok (if that's "slightly"), but not something which changes performance by ~30%.
Edit: why guess if you can look it up...
http://www.amd.com/us/products/note...md-radeon-6400m/Pages/amd-radeon-6400m.aspx#2
So, it looks like Seymour simply extended the 8-wide simds to 16-wide to get twice the SPs compared to Cedar/Park (it still has 8 tmus). Thus I guess it must have faster frontend to account for the huge transistor count increase.
http://techreport.com/discussions.x/20195
I think it's the first time I've seen SP & transistor count for Turks/Caicos (Whistler/Seymour).
Blackcomb: 960SP, 1.7B, 680Mhz
Whistler: 480SP, 715M, 725Mhz
Seymour: 160SP, 370M, 800Mhz
(Mhz calculated from quoted gflops - like usual I don't really expect them to quite hit these in actual notebooks.)
The lineup imho has too many chips however. In particular, Whistler looks like it's a Redwood with Barts tweaks and an additional simd. There's not really that many more transistors compared to that predecessor (+14%) and hence probably die size is quite similar too, so I'm not sure why AMD is promoting that old one to HD6500M series instead of just dropping it - could still sell partially disabled Whistler parts with 400 (or even 320) SPs.
Blackcomb is Barts with 2 SIMDs disabled - interesting that AMD chose not to release a fully enabled part. In any case, it should be quite a beast, beating all currently released chips (including that monster GTX480M with its stellar TDP) easily (no doubt nvidia will counter with a less crippled GF104/GF114 than they currently are offering, or even GF110).
Seymour is the most interesting wrt speculation - the transistor count increase over Cedar is way too big for a simple increase of the simds to 16-wide. So either it just adds 2 more 8-wide simds for a total of four, or there's other changes gobbling up the transistors (could for instance have the faster Redwood/Juniper style frontend, don't forget Cedar was simpler there).
Also, I think it would have been a good opportunity to drop that anemic cedar-based HD6300M. Contrary to the past, I don't think it will be too popular since when equipped with ddr3 it will get beaten by Sandy Bridge IGP. Granted it has dx11 support, better drivers, probably much faster MSAA, better quality AF filter and so on, but that's probably not enough to really be an option instead of the IGP (considering cost and power). With gddr5 it might be faster but I still doubt it's worth it.
Speaking of gddr5, AMD is saying all chips support gddr5 or ddr3. I can only hope noone takes that literally and will offer totally stupid things like HD 6800M with ddr3 (which would likely be slower than HD 6700M with gddr5). That's also something I really wish AMD would finally enforce - products with different ram type (if you REALLY want to offer them) just NEED a different name (the table is only using series names so can't judge that yet but considering the past fearing the worst). Slightly different clocks using the same name might be ok (if that's "slightly"), but not something which changes performance by ~30%.
Edit: why guess if you can look it up...
http://www.amd.com/us/products/note...md-radeon-6400m/Pages/amd-radeon-6400m.aspx#2
So, it looks like Seymour simply extended the 8-wide simds to 16-wide to get twice the SPs compared to Cedar/Park (it still has 8 tmus). Thus I guess it must have faster frontend to account for the huge transistor count increase.
Last edited by a moderator:
I fully agree. The AMD product pages in your second link gives some slim hope for the future, because of the different 6700M vs. 6600M monikers, where the only possible difference hinted in the spec sheet is indeed ram type.
However, no such difference in naming is shown for any other device, so one is left to wonder what the heck the 6700M vs. 6600M is really about. I wish Dave Baumann would step in and clarify.
This situation sure does need saving… At this point, buying a laptop with a good graphics card is a daunting task, and AMD/NVIDIA aren't making it any easier.
Broken Hope
Regular
Any reason why AMD has decided to effectively increase the prices on the 69xx series? They were just about competitive at their old prices compared to Nvidia's offerings, at increased prices not so much.
http://forums.overclockers.co.uk/showpost.php?p=18139400&postcount=29
http://forums.overclockers.co.uk/showpost.php?p=18139400&postcount=29
I believe their prices have gone up because the UK increased VAT by 2.5%.
Besides, in terms of raw performance per dollar/pound/euro, the 6950 is still a better deal than the GTX 570, especially if you consider the fact that at least 90% of all rev.1 6950's seemingly can be flashed to a 6970.
Broken Hope
Regular
This isn't the VAT increase, the purchasing manager, in that link superficially says that AMD were giving them rebates which they no longer are.
As to the person who said it's time for AMD to start making money, why would anyone buy the 69xx series if Nvidia have cards at the same or better performance for a cheaper price? The whole point is AMD have smaller chips so should be able to sell them cheaper than Nvidia's much larger chips.
Don't worry too much about it! If the consumers buy less cards than being produced, the prices will start to drop. You know, it isn't like AMD is sitting on a huge pile of those HD6900 cards. If they can sell them for more, they will do it. AMD is a company and not a charity organization after all.
Exchange rate?
It isn't 2008 or anything but in November it was above 1.6 USD to GBP now it's 1.54 USD to GBP. There is an expectation the GBP will fall further due to risk of double dip recession in country(due to new taxes, austerity measures and problems in the economies of the euro trading partners).
On the production side the reverse is happening with all the asian currencies battling inflation(ie Chinese interest rate rise on Christmas day) and their currencies trying to break free of the US Dollar as their economies push ahead.
Yes it looks like you will be paying more
It isn't 2008 or anything but in November it was above 1.6 USD to GBP now it's 1.54 USD to GBP. There is an expectation the GBP will fall further due to risk of double dip recession in country(due to new taxes, austerity measures and problems in the economies of the euro trading partners).
On the production side the reverse is happening with all the asian currencies battling inflation(ie Chinese interest rate rise on Christmas day) and their currencies trying to break free of the US Dollar as their economies push ahead.
Yes it looks like you will be paying more
Similar threads
