Results 1 to 18 of 18

Thread: So, what have I missed!

  1. #1
    Xtreme Addict
    Join Date
    Feb 2006
    Location
    northern ireland
    Posts
    1,006

    So, what have I missed!

    Haven't posted in a decade, are Intel still on top? ATI still making good cards? Extremesystems forum still the place to be? AMD zone still full of crazy fanboys that don't realise that AMD are finished? I mean, real men have FAB's, right? Might fire my prescott up and do some prime runs? Did they ever make a more efficient cpu?

  2. #2
    I am Xtreme zanzabar's Avatar
    Join Date
    Jul 2007
    Location
    SF bay area, CA
    Posts
    15,817
    ATi doesnt exist. AMD is the best CPU again to the point that intel is not competative without liquid. Nvidia just put out a set of 400W cards on the stock OC bios.
    5930k, R5E, samsung 8GBx4 d-die, vega 56, wd gold 8TB, wd 4TB red, 2TB raid1 wd blue 5400
    samsung 840 evo 500GB, HP EX 1TB NVME , CM690II, swiftech h220, corsair 750hxi

  3. #3
    Xtreme Mentor
    Join Date
    Feb 2004
    Location
    The Netherlands
    Posts
    3,176
    Quote Originally Posted by zanzabar View Post
    ATi doesnt exist. AMD is the best CPU again to the point that intel is not competative without liquid. Nvidia just put out a set of 400W cards on the stock OC bios.
    lol i'm getting hd 5870 vibes all over again.

    Ryzen 9 3900X w/ NH-U14s on MSI X570 Unify
    32 GB Patriot Viper Steel 3733 CL14 (1.51v)
    RX 5700 XT w/ 2x 120mm fan mod (2 GHz)
    Tons of NVMe & SATA SSDs
    LG 27GL850 + Asus MG279Q
    Meshify C white

  4. #4
    Xtreme Mentor
    Join Date
    May 2008
    Location
    cleveland ohio
    Posts
    2,836
    Quote Originally Posted by biohead View Post
    lol i'm getting hd 5870 vibes all over again.
    Just difference is that none of these card will link with each other. like GTX 480 and HD 5870 would all go quad link. Since it's now "dead"

    I recall a long time ago someone mentioned that if we ever get stuck with single cards for gaming that prices would of those cards would go up into the $1,200.00 range < irony that's what the RTX 2080 ti ended up at.
    HAVE NO FEAR!
    "AMD fallen angel"
    Quote Originally Posted by Gamekiller View Post
    You didn't get the memo? 1 hour 'Fugger time' is equal to 12 hours of regular time.

  5. #5
    I am Xtreme zanzabar's Avatar
    Join Date
    Jul 2007
    Location
    SF bay area, CA
    Posts
    15,817
    Big navi might do xfire still, maybe even quad. It wont be useful except for open class 3dmark, but it might happen.
    5930k, R5E, samsung 8GBx4 d-die, vega 56, wd gold 8TB, wd 4TB red, 2TB raid1 wd blue 5400
    samsung 840 evo 500GB, HP EX 1TB NVME , CM690II, swiftech h220, corsair 750hxi

  6. #6
    Xtreme Mentor
    Join Date
    May 2008
    Location
    cleveland ohio
    Posts
    2,836
    Quote Originally Posted by zanzabar View Post
    Big navi might do xfire still, maybe even quad. It wont be useful except for open class 3dmark, but it might happen.
    The last cards to support crossfire where the RX 500 series.

    I'm pretty sure that the Radeon VII dropped crossfire, but it does support mGPU in DX12 if the game supports it, same with Navi.

    Which I'm pretty sure even lowest cards RX 5500 should support mGPU on DX12.
    HAVE NO FEAR!
    "AMD fallen angel"
    Quote Originally Posted by Gamekiller View Post
    You didn't get the memo? 1 hour 'Fugger time' is equal to 12 hours of regular time.

  7. #7
    I am Xtreme zanzabar's Avatar
    Join Date
    Jul 2007
    Location
    SF bay area, CA
    Posts
    15,817
    Vega 56/64 do xfire. They were supposed to do quad, but that driver never came in. The 570-590 were the last mid range cards to support it. The radeon 7 supports some multi gpu features, but it uses the workstation driver. They said it was a gaming card, but it really a fire card.



    All cards that support dx12 can be used with explicit multi gpu.
    5930k, R5E, samsung 8GBx4 d-die, vega 56, wd gold 8TB, wd 4TB red, 2TB raid1 wd blue 5400
    samsung 840 evo 500GB, HP EX 1TB NVME , CM690II, swiftech h220, corsair 750hxi

  8. #8
    Xtreme Mentor
    Join Date
    Feb 2004
    Location
    The Netherlands
    Posts
    3,176
    Crossfire/SLI is not a viable scaling solution when it doesn't improve latency and developers have to support it. I'm not surprised it's been phased out.

    The next step in scaling performance is chiplet based packages like Zen. Since wafers are getting more and more expensive the chiplet approach seems like the only way to keep solve for costs, yields, heat density, etc. RDNA3 is supposed to be the first chiplet-based arch. I'm really excited for that. Imagine 4x40CU chiplets with an IHS the size of a threadripper, boom.

    Ryzen 9 3900X w/ NH-U14s on MSI X570 Unify
    32 GB Patriot Viper Steel 3733 CL14 (1.51v)
    RX 5700 XT w/ 2x 120mm fan mod (2 GHz)
    Tons of NVMe & SATA SSDs
    LG 27GL850 + Asus MG279Q
    Meshify C white

  9. #9
    I am Xtreme zanzabar's Avatar
    Join Date
    Jul 2007
    Location
    SF bay area, CA
    Posts
    15,817
    Quote Originally Posted by biohead View Post
    RDNA3 is supposed to be the first chiplet-based arch. I'm really excited for that. Imagine 4x40CU chiplets with an IHS the size of a threadripper, boom.
    That will a big turning point, but it might also have bad latency so high frame rates might not happen. It can be an issue for CPU right now going cross chiplet or even cross CCX if you have to bounce off the IO die.
    5930k, R5E, samsung 8GBx4 d-die, vega 56, wd gold 8TB, wd 4TB red, 2TB raid1 wd blue 5400
    samsung 840 evo 500GB, HP EX 1TB NVME , CM690II, swiftech h220, corsair 750hxi

  10. #10
    Xtreme Addict
    Join Date
    Dec 2004
    Location
    Southern California
    Posts
    1,225
    glad to see you still kickin around here zan
    i9-10900k@5.3ghz//MSI MEG z490 Unify//32GB Gskill TridentZ b.die@DDR4666//RTX 2080ti(+150/+700) kingpin bios//Samsung 970 Pro//Corsair AX1200i
    Custom Loop: Dual DDCs->Dual EK XE360 w/GT's -> HK IV CPU -> HK IV GPU ->EK X3 Res controlled by Aquaero 6

  11. #11
    I am Xtreme zanzabar's Avatar
    Join Date
    Jul 2007
    Location
    SF bay area, CA
    Posts
    15,817
    Quote Originally Posted by Circaflex View Post
    glad to see you still kickin around here zan

    It is good to see people back.
    5930k, R5E, samsung 8GBx4 d-die, vega 56, wd gold 8TB, wd 4TB red, 2TB raid1 wd blue 5400
    samsung 840 evo 500GB, HP EX 1TB NVME , CM690II, swiftech h220, corsair 750hxi

  12. #12
    Xtreme Mentor dengyong's Avatar
    Join Date
    Nov 2006
    Location
    A great place again
    Posts
    2,590
    Anyone know when Nvidia will drop their next big card?

  13. #13
    I am Xtreme zanzabar's Avatar
    Join Date
    Jul 2007
    Location
    SF bay area, CA
    Posts
    15,817
    3090 is thursday the 24th, and that is when the good pcb 3080 should launch. GB leaked a 20GB 3080 a 16GB 3070 and a 3060 in a promo that ends in December.
    5930k, R5E, samsung 8GBx4 d-die, vega 56, wd gold 8TB, wd 4TB red, 2TB raid1 wd blue 5400
    samsung 840 evo 500GB, HP EX 1TB NVME , CM690II, swiftech h220, corsair 750hxi

  14. #14
    Xtreme Addict
    Join Date
    Dec 2004
    Location
    Southern California
    Posts
    1,225
    Quote Originally Posted by dengyong View Post
    Anyone know when Nvidia will drop their next big card?
    well 3090 is coming, i think when big navi drops we see some Ti variants, or the larger 20g monster.
    i9-10900k@5.3ghz//MSI MEG z490 Unify//32GB Gskill TridentZ b.die@DDR4666//RTX 2080ti(+150/+700) kingpin bios//Samsung 970 Pro//Corsair AX1200i
    Custom Loop: Dual DDCs->Dual EK XE360 w/GT's -> HK IV CPU -> HK IV GPU ->EK X3 Res controlled by Aquaero 6

  15. #15
    Xtreme Mentor
    Join Date
    Feb 2004
    Location
    The Netherlands
    Posts
    3,176
    here's the RDNA1 whitepaper: https://www.amd.com/system/files/doc...whitepaper.pdf

    On page 6 you can see each 10 CU module has its own memory controller, L1/L2 cache and render backends. Compared to CPU cores, I don't think these modules really have to talk to each other through an L3 cache etc. Graphics workloads are really all about parallel throughput. (There is some talk of RDNA2 having an L3 cache to reduce memory bandwidth dependencies, but it's all rumours right now.) I think if you split those shader engines into chiplets, yes there would be latency penalties in the render pipeline but you could offset these by increasing CU count. We'd still be talking microseconds. That matters for memory/cache access but not so much in the render pipeline probably. Just a guess.

    Ryzen 9 3900X w/ NH-U14s on MSI X570 Unify
    32 GB Patriot Viper Steel 3733 CL14 (1.51v)
    RX 5700 XT w/ 2x 120mm fan mod (2 GHz)
    Tons of NVMe & SATA SSDs
    LG 27GL850 + Asus MG279Q
    Meshify C white

  16. #16
    Xtreme Mentor
    Join Date
    May 2008
    Location
    cleveland ohio
    Posts
    2,836
    Quote Originally Posted by biohead View Post
    here's the RDNA1 whitepaper: https://www.amd.com/system/files/doc...whitepaper.pdf

    On page 6 you can see each 10 CU module has its own memory controller, L1/L2 cache and render backends. Compared to CPU cores, I don't think these modules really have to talk to each other through an L3 cache etc. Graphics workloads are really all about parallel throughput. (There is some talk of RDNA2 having an L3 cache to reduce memory bandwidth dependencies, but it's all rumours right now.) I think if you split those shader engines into chiplets, yes there would be latency penalties in the render pipeline but you could offset these by increasing CU count. We'd still be talking microseconds. That matters for memory/cache access but not so much in the render pipeline probably. Just a guess.
    chiplets with a central hub that can delete late frames would be fantastic, but that would be a very very fast central hub. No more micro-stutter ever.
    HAVE NO FEAR!
    "AMD fallen angel"
    Quote Originally Posted by Gamekiller View Post
    You didn't get the memo? 1 hour 'Fugger time' is equal to 12 hours of regular time.

  17. #17
    Xtreme Addict
    Join Date
    Oct 2007
    Location
    Chicago,Illinois
    Posts
    1,182
    Quote Originally Posted by zanzabar View Post
    Vega 56/64 do xfire. They were supposed to do quad, but that driver never came in. The 570-590 were the last mid range cards to support it. The radeon 7 supports some multi gpu features, but it uses the workstation driver. They said it was a gaming card, but it really a fire card



    All cards that support dx12 can be used with explicit multi gpu.
    Vega frontier edition is a quad fire card, they dropped eyefinity support when crossfire is in use. Still works though



  18. #18
    Xtreme Addict
    Join Date
    Oct 2007
    Location
    Chicago,Illinois
    Posts
    1,182
    Quote Originally Posted by demonkevy666 View Post
    Just difference is that none of these card will link with each other. like GTX 480 and HD 5870 would all go quad link. Since it's now "dead"

    I recall a long time ago someone mentioned that if we ever get stuck with single cards for gaming that prices would of those cards would go up into the $1,200.00 range < irony that's what the RTX 2080 ti ended up at.
    The 8800 GTX and the 2900XTX were $700 sli/xfire cards.
    A single card will not give performance needed unless everything is limited, like graphics and resolution or Fov,vast areas with nothing going on . I hope Crysis Remastered or cyberpunk can give us that 2007 feel. Back then I was playing united offensive on a no running , sniper only server.k/d was like 24-1, my tag was Hellhound



Bookmarks

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •