Latest CPU & GPU Discussion (Intel/AMD/Nvidia)

Thread Rating: 5 votes, 5.00 average.
(5 votes)
Results 14,801 to 14,818 of 14818

  1. Post
    I saw that the (16C, 32T) 3950x which costs $750USD destroyed Intel's $2,000USD i9-9980XE (18C, 36T) in multicore workloads in geekbench - that has got to hurt them (and it was only an engineering sample on lower than final base and boost clocks). 61,072 multicore points vs 46,876

  2. Post
    HELL KNIGHT wrote:
    I saw that the (16C, 32T) 3950x which costs $750USD destroyed Intel's $2,000USD i9-9980XE (18C, 36T) in multicore workloads in geekbench - that has got to hurt them (and it was only an engineering sample on lower than final base and boost clocks). 61,072 multicore points vs 46,876
    And then suddenly Intel are suggesting you shouldn't use synthetic benchmarks as a test of a good CPU!

  3. Post
    SL1CKSTA wrote:
    I really really doubt intel will start fabing somewhere else not until they are in real trouble.
    https://www.techpowerup.com/256613/i...-14-nm-process

  4. Post
    Wow so this is quite telling. They've basically admitted they cant figure out how to go from 14nm to 7nm on their own. Yet AMD a company with less R&D resources was able to figure it out in less time too. lol. Oof.

  5. Post
    Valeyard wrote:
    Wow so this is quite telling. They've basically admitted they cant figure out how to go from 14nm to 7nm on their own. Yet AMD a company with less R&D resources was able to figure it out in less time too. lol. Oof.
    Intel's problems is with it's 10nm FAB production, AMD doesn't have it's own FABs all it's production is outsourced to TSMC. Your comparison doesn't make too much sense - you should be comparing Intel vs TSMC RnD resources.

    Unfortunately, Intel likes to design it's architecture in a way that makes it's own in house FAB process central and key to making it work. This mean Intel trying to outsource Sunnycove to TSMC 7nm or Samsung 7nm+ EUV would require an architecture re-design.

    Rumors have it that Intel can't get the clock speeds up high enough, it's got significant IPC improvements from Sunnycove over Coffeelake - unfortunately it's 10nm process is such a failure that they can't produce a CPU that goes over 3.7ghz which negates any IPC gain, plus yields are likely very low.

  6. Post
    Just release it and deal with clockspeed down the track IMO. Just like any new node.

  7. Post
    This all goes in cycles. Intel has obviously reached the limits of their current architecture they obviously can't refine it any more.

    They must be designing a new architecture. It will take time but it will likely trounce AMD when it does come out and so the wheel turns...

  8. Post
    The cycle theory is fine, but you're ignoring the fact Intel got off their bike for a couple years, and are playing catch up. By the time Intel has something new, so will AMD. Intel are on the back foot, no two ways about it, well deserved too. AMD is already developing the next two generations of Zen. Intel is just sitting this round out.

  9. Post
    Fragluton wrote:
    The cycle theory is fine, but you're ignoring the fact Intel got off their bike for a couple years, and are playing catch up. By the time Intel has something new, so will AMD. Intel are on the back foot, no two ways about it, well deserved too. AMD is already developing the next two generations of Zen. Intel is just sitting this round out.
    Yeah the thing is AMD is now on Zen 2 of the chiplet architecture, by the intel hits back with chiptlets of their own, AMD will be on gen 4, and that's 4 iterations of debugging problems/issues and enhancing/optimizing their designs
    While intel will be on their first that is assuming Intel will go chiplets on their next arch

  10. Post
    Don’t forget the Zen mastermind jim keller is now working at Intel. It will take time, but they’ll come out swinging.
    (I’m not an intel fan... just saying)

    All hail Zen2

  11. Post
    3dfx_ownd_joo wrote:
    Don’t forget the Zen mastermind jim keller is now working at Intel. It will take time, but they’ll come out swinging.
    (I’m not an intel fan... just saying)

    All hail Zen2
    Wasnt he employed for the Xe development? I see hes working on foveros at the moment possibly for Xe.

  12. Post
    @cronicash interesting read regarding your question on the pricing logic for Ryzen 3950x

    https://www.pcgamesn.com/amd/ryzen-9...est-gaming-cpu
    Last edited by dickytim; Today at 6:53 pm. Reason: forgot link

  13. Post
    If x4 PCIE4.0 is just as fast as x8 pcie3.0 then navi would run four way pretty well on 16x 4x4x4x4? Wondering if AMD would employ a multi gpu niche like this making the board cheaper having less lanes per pcie and their gpu being the only option for more than 2x? Cant really see this pcie4.0 as a game changer especially since SIG announcing pcie6.0 standard recently. When does NDA drop for zen2? Surley not another where reviews came same day as sale like vega2? Looking at large heatsinks on these new pcie 4 M.2's also wondering if gpus will fit over top, even help a supposedly hot x570 chipset overheat faster?

  14. Post
    1st of July is the NDA IIRC

    Some of the boards like the Gigabyte Extreme use a copper heatpipe that runs off the PCH and into the VRM heatsink, so 1) it's thin and won't interfere and 2) the GPU blocking airflow won't matter since the heat is moved to the open area of the MOBO

    Additionally, if you get a board that has 2 PCIE 4 slots, you can just put the GPU in the bottom PCIE4 slot so that nothing blocks airflow over the PCH.

    The bottom PCIE4 is probably only x8, but that's more than enough for any GPU

  15. Post
    I see most boards interestingly have m.2 now above first slot or two slots away from first ,or even require you to utilize the mobo heatspreaders.

  16. Post
    SirGrim wrote:

    Some of the boards like the Gigabyte Extreme use a copper heatpipe that runs off the PCH and into the VRM heatsink, so 1) it's thin and won't interfere and 2) the GPU blocking airflow won't matter since the heat is moved to the open area of the MOBO

    Additionally, if you get a board that has 2 PCIE 4 slots, you can just put the GPU in the bottom PCIE4 slot so that nothing blocks airflow over the PCH.

    The bottom PCIE4 is probably only x8, but that's more than enough for any GPU
    Was talking about high heatsinks on m.2's soz

  17. Post
    cronicash wrote:
    Surley not another where reviews came same day as sale like vega2?
    It looks that way

  18. Post
    So looks like the intel samsung thing is for chipset ic's.