HD 6970 & HD 6950 Reviews

Discussion in '3D Hardware, Software & Output Devices' started by onethreehill, Dec 15, 2010.

  1. liolio

    liolio Aquoiboniste
    Legend

    Joined:
    Jun 28, 2005
    Messages:
    5,724
    Likes Received:
    194
    Location:
    Stateless
    Most review are a mixed bag but I can't help but think that the HD6950 should offer an incredible value after some drivers updates say Q1-2011 :)
     
  2. psolord

    Regular

    Joined:
    Jun 22, 2008
    Messages:
    444
    Likes Received:
    55
    Can someone explain why the single 570 has only 3W difference in idle, compared with the single 6970, while when in dual gpu mode of these cards, the difference jumps to 22W? I'd guess it should have been 6W! :S

    [​IMG]
     
  3. Dave Baumann

    Dave Baumann Gamerscore Wh...
    Moderator Legend

    Joined:
    Jan 29, 2002
    Messages:
    14,081
    Likes Received:
    651
    Location:
    O Canada!
    Crossfire ULPS is in place by default. Our power management capabilities allows us to put the GPU's that do not have displays attached into an ACPI state that saves more power than the normal static-screen states (which still need high enough clocks to refresh the screen). This means that our Crossfire static power is not the same multiple as each card, often moe like a 0.5x adder per addional card.
     
  4. psolord

    Regular

    Joined:
    Jun 22, 2008
    Messages:
    444
    Likes Received:
    55
    Yikes!

    And to think that I was disabling ULPS through the registry, in order for the MSI Afterburner's unofficial overclocking not to hang! :p

    So in an essence, I guess I will be looking at the same idle power usage as before.

    Maybe now that I have flashed my 5850s with the unlocked Asus BIOS and the ULPS disabling is not neede any more, I will be using my 5850s machine more, for daily tasks.

    Thank you good sir.:grin:

    PS On second thought though, the second 6970 added 18W compared to the single one.

    Now the second 570 added 37W, which can't be right, since we know that the 6970 has an idle power consumption of 20W which is consistent with the added 18W, while the second 570 added 37W which is not consistent with the 25W of the 570! :S
     
  5. Sxotty

    Legend Veteran

    Joined:
    Dec 11, 2002
    Messages:
    5,046
    Likes Received:
    413
    Location:
    PA USA
    Techreport's graph on the bottom of the second to last page showing system price and performance misses out on an opportunity. The idea is a nice one, but given the implementation it is pretty much pointless. I do realize they don't have the time to do the idea jusitce, but what they should have done is have a mid range system with the lower end cards.

    You have to have two pci-e slots for sli and xfire. So you they could get a cheaper mobo for the single cards. Obviously there are other such things such as CPU that could change. I realize this would get fairly complex quickly, but as it is the graph is misleading in its results. If they could do it justice then it would be quite interesting to see what the outcome was and if anything changed. For example the advantage of dual gpu in terms of $/performance might change. It might change more for one than another as well do the licensing fees or the lack thereof.
     
  6. gongo

    Regular

    Joined:
    Jan 26, 2008
    Messages:
    582
    Likes Received:
    12
    Why are Barts and Cayman overclocking potential so poor (HDXX70 maxes at around 940mhz..binned for HDXX50 struggle to hit 900mhz) and yet...Fermi can go up to 900mhz from the mid 700mhz?
     
  7. AnarchX

    Veteran

    Joined:
    Apr 19, 2007
    Messages:
    1,559
    Likes Received:
    34
  8. Silent_Buddha

    Legend

    Joined:
    Mar 13, 2007
    Messages:
    16,987
    Likes Received:
    6,237
    Again, I can't help looking at this and thinking 6870 would have been a quite apropo name for the card, ah well, that's a horse I've beaten to death.

    Cudo's again to AMD for not holding back the industry by implementing DP 1.2. Hopefully Nvidia gets off its lazy arse and implements DP in its next round of products.

    Powertune is also far more interesting than I expected. With it AMD no longer has to set default clocks for all cards (leaky ones and non-leaky ones) to the same value. Nor does it have to take into account things like user case ventilation or ambient room temperature when determining the lowest common denominator for determining clock speeds.

    Of course, that means the potential for some variable benchmarking results (leaky core versus less leaky core). And people living without air conditioning with high ambient temperatures and inadequate case ventilation will have slower performance and less risk of frying their card. :p

    Conversly, it also means that potential performance in outlier programs (shader heavy situations as Tech Report mentions for Perlin noise test, or high AA levels combined with high GPU load game) may increase with later releases of a product as the process matures. Not likely a factor with Cayman (assuming 40 nm is fairly mature now) but may be a factor with whatever hits on 28 nm.

    As well, there might be performance benefits to gaming in an air conditioned (or windows open if in Alaska or Siberia :p) room with a well ventilated case with certain programs.

    Or if you wanted lower power consumption and aren't averse to high noise levels, you could lower the TDP clamp by 20% and increase fan speed. And perhaps maintain the same level of game performance. Probably not useful to people living in a quite environment, but someone with high ambient noise levels (living next to a major arterial in a city for example) may not even notice noise from a higher fan speed.

    Here, I'm even more wishing some site would include load temperatures with each game tested. As well, going forward ambient room temperature and case ventilation will be important for reviews. But unlikely that the majority of review sites will report one this. Knowing how things work, a "pro-Nvidia" site could benchmark games in a not very well ventilated case with high room temperatures for example.

    Also interesting to note that the Cayman scales better with resolution and AA than the GF110 series. So 570 appears to be a slightly better proposition at 1920x1200 w/4xAA and below, while 6970 appears to be better at 1920x1200 w/8xAA and above. Although depending on game that could go either way.

    As well, I never noticed before how CrossfireX actually has less problems in games than SLI. Starcraft min FPS is borked on SLI. Civ V doesn't seem to do much with a second Nvidia card. Still, that's not relevant to me, I still don't particularly like AFR even though it's gotten slightly better.

    And load power of 6970 compared to 570 is a bit disappointing. Idle power is nicely improved however (compare 5870 2GB model to 6970, over 10 watts less).

    Either way can't say I'm sad that I won't be buying an AMD card for a while. Some interesting improvements. And somewhat compelling at the resolution I game at (2400x1500) but not enough to make me regret swearing off AMD for a while.

    Regards,
    SB
     
  9. Dave Baumann

    Dave Baumann Gamerscore Wh...
    Moderator Legend

    Joined:
    Jan 29, 2002
    Messages:
    14,081
    Likes Received:
    651
    Location:
    O Canada!
    No, the default algorthim has been implemented to minic the worst case still, so the performance is consistent across all ASIC's within a product. The PowerTune Overdrive slider allows that variablility to be exposed.
     
  10. Silent_Buddha

    Legend

    Joined:
    Mar 13, 2007
    Messages:
    16,987
    Likes Received:
    6,237
    Hmmm. So does this mean it adjusts TDP/power use based off a baseline established for each particular machine in each operating condition?

    So, for example, in a hot ambient temperature location with low case ventilation and high leakage core, it would establish the (just as an example) bootup conditions as a baseline to then use for determining when to throttle?

    Which would in essence then be different from the baseline it establishes for a cool ambient temperate, well ventilated case with a low leakage core?

    So overall performance between those two situations would be similar, but the baseline it uses to determine when to throttle would be different? And thus temperature/power use would also be different but still under rated TDP?

    Regards,
    SB
     
  11. entity279

    Veteran Regular Subscriber

    Joined:
    May 12, 2008
    Messages:
    1,259
    Likes Received:
    441
    Location:
    Romania
    How exactly? If the equation is the same and only one constant (TDP) changes, won't that variability still be hidden?
     
  12. ben6

    Regular

    Joined:
    Feb 8, 2002
    Messages:
    864
    Likes Received:
    3
    regarding drivers used in reviews

    I actually want to defend some of the sites that reviewed the product as to the drivers we/they used. With all due respect to AMD, some sites that were not briefed by them received cards (like mine). My site received several cards, with Driver CDs with the 10.11 drivers, and boxes. In fact for all the information that we got I almost put a "We have no official information on this architecture, the Features section of the review will entirely be based upon conjecture, slides that were NDA slides posted on websites and personal experience." I received 10.12 betas from a undisclosed source along with the information I needed on the review but not from AMD...
     
  13. Alexko

    Veteran Subscriber

    Joined:
    Aug 31, 2009
    Messages:
    4,514
    Likes Received:
    934
    Why isn't it possible to entirely turn off the second card, and just turn it on when a game is launched? (I assume that if it were possible, it would already be the case.)
     
  14. eastmen

    Legend Subscriber

    Joined:
    Mar 17, 2008
    Messages:
    10,316
    Likes Received:
    1,748
    Do you still need a DP adapter when using 3 dvi monitors or 2 dvi and a hdmi monitor ?
     
  15. ben6

    Regular

    Joined:
    Feb 8, 2002
    Messages:
    864
    Likes Received:
    3
    I am unsure what you mean. Do you mean an active adapter?
     
  16. Dave Baumann

    Dave Baumann Gamerscore Wh...
    Moderator Legend

    Joined:
    Jan 29, 2002
    Messages:
    14,081
    Likes Received:
    651
    Location:
    O Canada!
    Desktops are not always happy when the PCIe device turns itself off while the system is on. Addionally, completely powering down a device also means both reloading the BIOS and the driver (for that device) which takes time.
     
  17. Dave Baumann

    Dave Baumann Gamerscore Wh...
    Moderator Legend

    Joined:
    Jan 29, 2002
    Messages:
    14,081
    Likes Received:
    651
    Location:
    O Canada!
    This is an algorithm that calculates an infered power draw based on execution unit activity. Voltage and Leakage could be a variable to that algorithm, but for the sake of providing a consistent default product performance they are input as a constant per SKU, based of the worst case for that SKU. Tempurature variablity presents more challenges, though it is something we are looking at in terms of enhancing the capabilties for the future.
     
  18. swaaye

    swaaye Entirely Suboptimal
    Legend

    Joined:
    Mar 15, 2003
    Messages:
    8,567
    Likes Received:
    652
    Location:
    WI, USA
    I want a review that looks at all of these cards with SSAA. My new favorite mode for everything.
     
  19. Florin

    Florin Merrily dodgy
    Veteran Subscriber

    Joined:
    Aug 27, 2003
    Messages:
    1,658
    Likes Received:
    242
    Location:
    The colonies
    Wait, so 6870 is actually slower than 5870, and 6970 is slower than 5970? Keep up the good work lads and by the time the 9870 rolls around we might be down to a card which is as fast as an S3 Virge, but at least it'll be DX11.
     
  20. Alexko

    Veteran Subscriber

    Joined:
    Aug 31, 2009
    Messages:
    4,514
    Likes Received:
    934
    OK, thanks. Is that something you think might change with future versions of PCI-E?

    Would that be the reason PowerTune started throttling Damien's card even though it was only drawing about 214W in Furmark?
     
Loading...

Share This Page

  • About Us

    Beyond3D has been around for over a decade and prides itself on being the best place on the web for in-depth, technically-driven discussion and analysis of 3D graphics hardware. If you love pixels and transistors, you've come to the right place!

    Beyond3D is proudly published by GPU Tools Ltd.
Loading...