Search:

Type: Posts; User: mczak; Keyword(s):

Page 1 of 2 1 2

Search: Search took 0.01 seconds.

  1. Replies
    8
    Views
    1,073

    It is quite notable that as far as Haswell Xeons...

    It is quite notable that as far as Haswell Xeons go the E5-2620v3 is a very cheap option - in fact two of them are still cheaper than a single 5960X...
    For a dual socket solution, this seems...
  2. There's actually quite a good reason why more...

    There's actually quite a good reason why more squarish screens are better, at least in the mobile space. Personally, I don't really care all that much about the exact ratio, but I DO care about the...
  3. Maybe, but I wouldn't expect a OGL driver shared...

    Maybe, but I wouldn't expect a OGL driver shared by everyone soon. Some vendors sell their workstation-class cards mostly by their super optimized drivers, and I wouldn't expect them to just give...
  4. SPIR-V should easily support everything d3d10/11...

    SPIR-V should easily support everything d3d10/11 needs. At least I can't see why it wouldn't. Can't say though how this is going to turn out, so far tgsi has survived all attempts to replace it :-).
  5. I think that's not too surprising - I didn't...

    I think that's not too surprising - I didn't really check this but I suspect sample_c refers to the d3d10 opcode of that name.
    This means it should affect texture sampling with shadow comparisons....
  6. I believe the ycbcr subsampling support was...

    I believe the ycbcr subsampling support was optional (starting with hdmi 1.2) whereas it is now mandatory with 2.0 (2.0 also has a couple other features noone really cares about). So, I don't know if...
  7. That's not quite true. MST doesn't increase...

    That's not quite true. MST doesn't increase bandwidth over a single cable one bit. All it allows is, as it says, drive multiple monitors from the same cable (hence multi stream). But the total...
  8. This change has nothing to do with the microcode....

    This change has nothing to do with the microcode. I don't expect AMD to open that one up... It is already shared between open and closed driver already however.
  9. Carrizo is supposed to be Vulcanic Islands (not...

    Carrizo is supposed to be Vulcanic Islands (not that this would be confirmed).
    So same as Tonga (r9 285). Which fits neither the category of unreleased new hardware nor the one of supported with...
  10. I doubt it would run very well. The single-thread...

    I doubt it would run very well. The single-thread bottleneck for geometry processing would likely get very serious. Plus there's zero optimization done for the vector instruction set of this cpu -...
  11. llvmpipe can use up to 16 threads (actually you...

    llvmpipe can use up to 16 threads (actually you could change it to use more pretty easily though it might not scale all that well past 8, in particular because everything pre-rasterization is single...
  12. I never said I have a problem - the kernel...

    I never said I have a problem - the kernel usually blacklists the affected models so it doesn't use queued trim (just the old non queued version) so you don't see a problem (I believe just about the...
  13. Would be nice to know if this SSD also claims to...

    Would be nice to know if this SSD also claims to support queued trim but fails rather horribly killing your data when you actually try it like the M500/M550 series. (Worse it seems Crucial claimed...
  14. swrast fallbacks are highly overrated. They are...

    swrast fallbacks are highly overrated. They are ok for running conformance tests but not really anything else.
    1) The huge (can easily be ~100x) performance hit is usually simply unacceptable, thus...
  15. Replies
    25
    Views
    7,197

    I highly doubt that given the quite low wattage....

    I highly doubt that given the quite low wattage.
    Pretty sure it's just the reference clock which limits because everything depends on it. Wondering what the "true" clock limit of the cpu cores...
  16. Replies
    31
    Views
    7,882

    So these results are way better than the earlier...

    So these results are way better than the earlier ones for the 5350 and much more like they should be. So what was the problem with the earlier numbers? While I understand the issues with DPM for the...
  17. That mirrors quite well what I've seen elsewhere...

    That mirrors quite well what I've seen elsewhere on the cpu side of things at least. Silvermont core seems to be about 20% or so slower per clock for general integer performance, and worse on the...
  18. No that's definitely not the case -if it would be...

    No that's definitely not the case -if it would be could still just provide a knob to forcibly fake it. In fact I think performance might not be all that bad (that is, not all that much slower than it...
  19. I think people are forgetting this nvidia board...

    I think people are forgetting this nvidia board is a board for _development_. Admittedly, a nice one at that with its quite extensive and useful ports (I am actually sort of surprised the chip...
  20. That doesn't sound quite right. It's been a while...

    That doesn't sound quite right. It's been a while since I've worked on r200, but IIRC r200 is missing the functionality of 1.4 for depth textures, that is ARB_depth_texture and ARB_shadow. It should...
Results 1 to 20 of 36
Page 1 of 2 1 2