Page 255 of 372 FirstFirst ... 155205245253254255256257265305355 ... LastLast
Results 2,541 to 2,550 of 3718

Thread: Vega 10, Vega 11. What we know

  1. #2541
    Senior Member
    Join Date
    Sep 2011
    Posts
    871
    Quote Originally Posted by Bluefoot View Post
    **** FE is confirmed to be using Fiji drivers for gaming ****

    What part of this is so difficult to understand?
    The whole point of my comment was If the correct Vega drivers will increase performance by at least 20-30% or not.

  2. #2542
    Senior Member
    Join Date
    May 2012
    Posts
    4,673
    Quote Originally Posted by TESKATLIPOKA View Post
    The whole point of my comment was If the correct Vega drivers will increase performance by at least 20-30% or not.
    Could be ? I'd hazard a guess that if you tried to use Fiji game specific optimizations on Vega you could actually end up making things worse ?
    To find the right answers you must ask the right questions.

  3. #2543
    Senior Member
    Join Date
    Dec 2010
    Location
    Switzerland Geneva Cern
    Posts
    2,904
    Quote Originally Posted by TESKATLIPOKA View Post
    The whole point of my comment was If the correct Vega drivers will increase performance by at least 20-30% or not.
    At this point vega FE has worse performance per CU per MHz than fury X. Which itself had one of the lowest performance per CU from all amd cards.

    I guess a 10 - 15% to match fury X perf per CU is guaranteed with proper drivers. This is enough to bring vega on par with reference 1080 cards.
    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

  4. #2544
    Senior Member
    Join Date
    Dec 2010
    Location
    Switzerland Geneva Cern
    Posts
    2,904
    Quote Originally Posted by Stuckey View Post
    Could be ? I'd hazard a guess that if you tried to use Fiji game specific optimizations on Vega you could actually end up making things worse ?
    Well for starters vega has narrower bus and probably different latency. Also lower theoretical bandwidth...
    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

  5. #2545
    8-bit overflow
    Join Date
    May 2010
    Posts
    591
    Quote Originally Posted by TESKATLIPOKA View Post
    The whole point of my comment was If the correct Vega drivers will increase performance by at least 20-30% or not.
    I think the answer you're looking for is "I don't know, 20-30% seemed reasonable so I just through it out there".

    Should definitely see at least 10% from drivers and maybe another 10% from a decent cooler / stable clocks but nobody really knows. There could be 100% more performance and there could be none. RX will perform 0-100% better, I am the most right

    I think it may actually be safe to say that Vega will be faster than the 1080ti by the time Navi lands

  6. #2546
    Senior Member
    Join Date
    Jun 2010
    Posts
    4,228
    Quote Originally Posted by os2wiz View Post
    Are you another flatliner? It has clearly been determined that the drivers on the F.E. Vega are primitive. They are actually FIJI drivers with little exception. They are not using the new hardware in Vega. The final drivers should be some 25% better than the current state of F.E. drivers. Plus the clock speed on RX Vega wil undoubtedly be at least 100mhz higher than F.E. That would mean at least 30% better performance.
    I'm only defending we should wait to see the final product, I don't know how it will end, but things should be really bad to gain the worst release ever prize.

    Quote Originally Posted by TESKATLIPOKA View Post
    As I already said, I was talking about something else. I wasn't talking about his deleted posts by the moderators, I don't even know what he wrote to cause them to be deleted, so I can hardly comment them. I was talking about a situation which happened days before he had those deleted posts.

    If someone says It's a winner or disaster, I think It's better If he also writes his reasoning. On the other hand, If he doesn't, what does It change? It's just his own opinion. You can agree or simply disagree with his opinion.

    Yes, yes, in this thread with ~253 pages, there is not a single hype post, because you didn't see It.
    If you focus only on the last ~10-20 pages then I am not surprised you only see posts about disaster.
    I'm talking about the same thing.
    He claimed AMD doomed, Vega a hot mess, TDP 375w, power consumption easily above 300w, under performing... Now we know FE's temperature is ok with a normal heatsink, power consumption around 280w, performance far better than Titan XP for the market it is after. So he ended wrong and people that asked for proof was right doing it. When asked posts changed to workstation cards with certified drivers when this card is not focused on people who buys those kind of cards, or to RX Vega as if he knew anything about it while the rest continue waiting to see the final product. For example lately he was also asking what made RX Vega a good buy, logically people continued asking what he knew that we didn't about the unreleased card.

    You can add awful remarks and baits like "what would you expect after Polaris?", "RX will be faster with nitrogen liquid?... nonsense about HBM2 and consequently more proof requests from other people, more baits, memes, removed posts for trolling...and a ban, the attitude was clear for some people for some time, more focused on trolling than a discusion, but curiously that's what you are defending.

    Oh, and he was one of those people expecting Vega 11 because the gap was going to be very big, not sure if that was hype for you or not. When your post read as if the thread was full of hype and people "attacked" him because he was one of the few not hyped it looks disingenuous.


    Hype for me is created excitement.
    Reasoning explains why someone reaches some conclusion, an opinion must be written as an opinion and some are not. It matters to me, at some point I have to ask myself if someone is trolling or shilling.
    Yes, but I'm not good evaluating if someone is excited and tries to excite people towards some product with their posts.

  7. #2547
    Senior Member
    Join Date
    Apr 2010
    Posts
    110
    Quote Originally Posted by Stuckey View Post
    ...
    Yeah, not really sure why a game optimization for Fiji would work on Vega ? About the only things they share in common are 4096 shaders and an interposer. Literally every other aspect of them is completely different. Geometry engine, pixel engine, cache, memory system, the shaders themselves, everything.
    ...
    The main difference is that Vega CU can accept fragmented work, but CU width remains 4x16.
    One can reuse old driver packing job magic - no problem, but new abilities aren't exploited ...

  8. #2548
    Senior Member
    Join Date
    Jan 2010
    Location
    London
    Posts
    3,213
    Quote Originally Posted by TESKATLIPOKA View Post
    Don't you think If the problem were the drivers, then AMD would already say something? I didn't hear anything from AMD about how the correct drivers will increase performance by ~20-30%. Did you hear anything? Or are they waiting for something what we are not privy to?
    Well AMD counts on us not to be stupid and assume things. Why do you say there is a problem with drivers? How hard is it for one to understand the meaning of generic drivers?
    AMD released FE as pro card. They showed pro apps, where it excelled. No problem with drivers there, right? Then AMD being a good guy decides to enable basic gaming mode for devs who actually want to test the code they created right away in the same system. Since none of the nvidia Pro cards can even run games at decent performance, AMD with basic gaming drivers already is ahead of nvidia. Now since gaming RX card is month away, and knowing AMD, who are so secretive (always!) about their unreleased products, why would they go over their head and show up full gaming potential of their unreleased card?
    Obviously we would love to know how RX will perform and we would love for AMD to 'leak' performance by releasing full blown desktop drivers for FE cards, but its not their style, regardless if its good or bad for them.

    So to sum up: FE is pro card. OK?
    FE gaming mode driver is basic driver to run games, not to run them great(whoever thought that running games without any optimizations at same performance as 1080 is bad, need to check their heads). OK?
    There is nothing wrong with game mode in the driver, its how it was meant to run and how AMD intended it to run. OK?

    Now it is completely possible that once RX is released and there is nothing to hide anymore and all the optimizations are made, AMD might update pro drivers and its gaming mode to stably boost gaming performance further. No one will cry for that, since people do not buy FE to run 3dmarks.

    On the other hand, this is just my logical opinion, and I might be wrong. But I'm not gonna jump to conclusions from some benchmarks run on pro card.
    Who of you ever based what gaming card to buy from some Pro card gaming tests?

  9. #2549
    Senior Member
    Join Date
    Apr 2010
    Posts
    110
    Quote Originally Posted by muziqaz View Post
    ...
    Obviously we would love to know how RX will perform and we would love for AMD to 'leak' performance by releasing full blown desktop drivers for FE cards, but its not their style, regardless if its good or bad for them.
    ...
    With CUs being able to execute multiple jobs, it can potentially make the GPU to behave like being double or quadruple (or more) in it's CU count.

    That would be a very good secret to keep until launch day ...

  10. #2550
    Quote Originally Posted by Z O X View Post
    With CUs being able to execute multiple jobs, it can potentially make the GPU to behave like being double or quadruple (or more) in it's CU count.

    That would be a very good secret to keep until launch day ...
    GCN CUs have always been able to run multiple workgroups of varying sizes from different dispatches simultaneously.

Tags for this Thread

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