Page 12 of 432 FirstFirst ... 210111213142262112 ... LastLast
Results 111 to 120 of 4311

Thread: Polaris 10 size / performance estimation

  1. #111
    Senior Member
    Join Date
    Dec 2010
    Location
    Switzerland Geneva Cern
    Posts
    2,951
    Quote Originally Posted by JeeBee View Post
    Okay, let's do this again.

    TSMC 28nm SRAM Cell: 0.16 ?m^2 in 2012 (so this may have got smaller over time)
    Samsung 14nm SRAM Cell: 0.064 ?m^2 in 2014 (LPE)

    LPP is smaller, btw.

    0.160 / 0.064 = 2.5

    Oh man, I didn't expect it to come out so precisely.

    Clocks can increase more than 20%, that's my point, it's merely 20% from 20nm, not 28nm. 28-20nm gives a 25-25% clock speed increase on top of that. I think I read 1.65x clocks /possible/ on 14nm versus 28nm at the same TDP, but cannot find the darned image now.
    So then what do you expect the transistor count of the supposed 232mm2 chip is supposed to be?

    Assuming it has 5.1 billion transistors we get a transistor density of 22 million transistors / mm2 which is 2x that of tahiti. Which is exactly what i believe its going to be give or take 10%.

    22*232 = 5100

    Assuming it has 6.3 billion transistors we get ~27 million transistors / mm2 which is ~2.5 times tahitis density. That will make the card a direct competitor of 390 / 390x.
    Current rig: i5 3570 @3.8GHz, Asus 7750, Dell 24'' 2560X1440, Corsair Vengence 16GB Ram DDR3 @1600MHz, Chieftec 550W PSU, 480GB OCZ SSD + 500GB Seagate + 1TB WD hd, Windows 10 64-bit

  2. #112
    Senior Member
    Join Date
    Dec 2010
    Location
    Switzerland Geneva Cern
    Posts
    2,951
    Quote Originally Posted by testbug00 View Post
    Use pitcairn. for 232mm^2 double that density you get about 6.1b transistors.

    Pitcairn launched a few months after Tahiti with higher density. Polaris 11 will likely launch a few months after Polaris 10.

    Polaris 10:, 2.5-3.5 billion transistors. Tahiti*100*2 ranging to Tahiti*130*2*1.08
    Polaris 11: 6.1-6.6 billion transistors. Pitcairn*232*2 ranging to Pitcairn*232*2*1.08


    [1.08 comes from density increase that Apple A9 from Samsung over TSMC]
    I thought they were going to be released in the same month or sth?
    Current rig: i5 3570 @3.8GHz, Asus 7750, Dell 24'' 2560X1440, Corsair Vengence 16GB Ram DDR3 @1600MHz, Chieftec 550W PSU, 480GB OCZ SSD + 500GB Seagate + 1TB WD hd, Windows 10 64-bit

  3. #113
    Senior Member
    Join Date
    Dec 2012
    Posts
    7,552
    Quote Originally Posted by Nerdmaster View Post
    I thought they were going to be released in the same month or sth?
    Well, we heard the small one taping out before the bigger one, right? And I imagine AMD would prefer to ramp up a smaller chip to work out any issues before going larger. So why not increase the density a bit?


    Also replacing Hawaii needs to be done. It's the oldest relevant GPU besides Pitacairn which we already know is being replaced. I think AMD design targeted for Polaris 11 to be able to match Hawaii if they were clocked equally in the worst case.

    EDIT: also it turns out that Pitcairn is about 8% denser than Tahiti. So given that Samsung's process advantage holds, than Polaris 11 based on Tahiti would end up around 6.1b transistors.

    (4313/352) * 2 * 1.08 * 232 = 6140
    https://duckduckgo.com/?q=4313%2F352...*232&ia=answer
    -Q

  4. #114
    Senior Member
    Join Date
    Feb 2010
    Location
    Planet Earth, Solar System, Milky Way
    Posts
    3,521
    Quote Originally Posted by Nerdmaster View Post
    Assuming it has 6.3 billion transistors we get ~27 million transistors / mm2 which is ~2.5 times tahitis density. That will make the card a direct competitor of 390 / 390x.
    You continue to ignore the fact that 390/390X is quite old version of the architecture and you cannot expect the same performance from the same number of transistors. You will 100% have higher!

    Also, Polaris 10 and Polaris 11 can very well be launched simultaneously this summer months.

  5. #115
    Banned
    Join Date
    Oct 2010
    Posts
    490
    I am going to say that based on how die sizes of past GPUs like HD 7770 (Cape Verde) and HD 7870 (Pitcairn) were related that a 232 sq mm Polaris GPU if it exists will have atleast 2.5 times the shader count of a 110 sq mm Polaris GPU( the chip demoed at CES was smaller than GK107 and HD 7770 which are 118 and 123 sq mm in die size respectively). Pitcairn at 212 sq mm doubled Cape Verde at 123 sq mm in every resource (sp,tmu,rops,memory bus) with a 75% increase in die size.

    My guess is AMD has improved memory compression (bandwidth efficiency) and shader efficiency to a level such that a 256 bit GDDR5 memory bus running at 7- 8 Ghz can feed the cores without bottlenecks. I think the comments by AMD's Roy Talylor that they would launch Polaris products which are faster than R9 290X and which will run faster, cheaper and consume lower power hint at such a GPU. If I had to guess the 232 sq mm GPU will have 2560 sp which is exactly double the Pitcairn GPU, but with significant shading efficiency improvements which would improve perf/sp, perf/ sq mm and perf/watt. Due to FINFET which would allow higher clocks I think we might actually see such a GPU come close to Fury X. The key here is AMD needs a high volume GPU to power the next gen mid-range and a GDDR5 based GPU seems the only logical candidate.

  6. #116
    Banned
    Join Date
    Nov 2011
    Posts
    576
    Quote Originally Posted by testbug00 View Post
    Well, we heard the small one taping out before the bigger one, right? And I imagine AMD would prefer to ramp up a smaller chip to work out any issues before going larger. So why not increase the density a bit.
    Once both chips are taped out, you can't make changes after the fact based on how one behaves during mass production.

    That's something people here have a very, very hard time understanding.

    The only thing that you could do, is tape out one, wait until it comes back, test if all is well, then tape out the next one. The problem with that is he delay between the 2 tape outs.

  7. #117
    Senior Member
    Join Date
    Dec 2012
    Posts
    7,552
    I don't think that this new chip will be cheaper. Maybe if they can cut down RAM to 4/5/6GB that could offset a small increase in the cost of the die? Still seems suspicious to me. Unless maybe the cut down part is cheaper, and the full part sudisizes it hard like the kind of pricing the 970 and 980 had at launch.
    -Q

  8. #118
    Senior Member
    Join Date
    Jan 2010
    Posts
    3,623
    Yeah, I also think that AMD will be able to sell these for a good chunky price, especially the 232mm^2 version (if that is P11), $399 to $499 wouldn't shock me for a fully-enabled SKU.

  9. #119
    Senior Member
    Join Date
    Dec 2010
    Location
    Switzerland Geneva Cern
    Posts
    2,951
    Quote Originally Posted by JeeBee View Post
    Yeah, I also think that AMD will be able to sell these for a good chunky price, especially the 232mm^2 version (if that is P11), $399 to $499 wouldn't shock me for a fully-enabled SKU.
    If its on par with 390x / 980 you can expect 499$+ price.
    Current rig: i5 3570 @3.8GHz, Asus 7750, Dell 24'' 2560X1440, Corsair Vengence 16GB Ram DDR3 @1600MHz, Chieftec 550W PSU, 480GB OCZ SSD + 500GB Seagate + 1TB WD hd, Windows 10 64-bit

  10. #120
    Senior Member
    Join Date
    Sep 2009
    Posts
    2,115
    Quote Originally Posted by Nerdmaster View Post
    If its on par with 390x / 980 you can expect 499$+ price.
    Then it would be a downgrade from Nano. If AMD could sell 390X performance at $499 Nano sales would be sky high now. My guess is $399 for the fastest Polaris 11 model.

    Primary System: Core i7-6700K - GTX 1060 6GB
    Secondary System: AMD FX-6300 - Radeon RX 470
    Gaming-Only System: Core i5-4670K - Radeon RX 480 8GB

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  
WordPress Appliance - Powered by TurnKey Linux