Page 1 of 11 123 ... LastLast
Results 1 to 20 of 201
  1. #1
    Secondary Hivemind Nexus squirrel's Avatar
    Join Date
    Jun 2011
    Posts
    1,933

    Just When We Think We Can Have a Great Valued GPU, Nvidia Plays Dirty Trick

    http://www.lazygamer.net/general-new...llocation-bug/

    Screw you Nvidia, what's wrong in offering GPUs with great value-performance ratios?! You screw up a great GPU so that people will be forced to buy price inflated GPUs for excessively high computing power we cannot utilize?

    Conduct business honestly, Nvidia, or you will be severely punished by the market.

    BTW I think it's great time to upgrade because MS has promised to offer free upgrade to Windows 10 for current 8 and 7 users. Don't need to pay extra on OS when Wins 10 is out.
    Last edited by squirrel; 24-01-2015 at 10:26 AM.

  2. #2
    Well, this is a downer..
    I bought myself a 970 recently after thinking about upgrading for quite some time now and pushed my budget for that somewhat because i thought this would be the best value for money if i wanted some shiny graphics.

    Can't really do anything about it now because i exceeded the return policy time of the shop by a few days and wouldn't have the extra paper thingies to upgrade to a 980.

    I feel cheated by Nvidia because i trusted them with my hard earned money and from skimming trough the linked thread it looks like they knew all along and won't be able to fix it via software changes.

    ----
    Hello world, hello hivemind! After years of silent reading and being mind controlled by the overlords i finally got around registering. :)

  3. #3
    Network Hub
    Join Date
    Nov 2014
    Posts
    256
    That has to be the worst writeup of this thing I've seen so far.

    Clever people, using VRAM benchmarking software, have discovered that when the last 500`700MB of VRAM gets accessed, memory performance drops significantly. Some users have even found that their cards go belly up when hitting 3GB.
    AFAIK anything showing results from below ~3600 is from people running it on the display GPU, rather than headless from a second gpu. Not that you would pick up that this is what you're meant to do from lazygamer either.

    The problems seems to be widespread – and Nvidia’s admitted that problem affects every single 970 (to varying degrees). They’re looking in to the issue – but unless it’s something that can be resolved via driver or firmware update, a recall may be on the cards.
    That's reading a post someone from nvidia quoted in a 4 post train long multiquote as an nvidia statement.

  4. #4
    Network Hub
    Join Date
    Nov 2014
    Posts
    256
    Hmm, doing some deeper digging I'm wondering if this isn't an API (or some weird 32bittedness) limitation, or maybe down to the game. FC4 and SOM seem to be the two people go-to to find an issue with it, which I don't have installed anymore. I can't get Rise of Flight to eat more than ~2600mb (even at 4x4 supersampling from 4k) and DCS World tops out at about 3575 at the max AA I can make it eat without falling over, and it's a 64 bit exe, however it's still DX9 and seems to have its own ideas about how to stream data.

    Someone on overclock.net suggested MSI Kombustor's opengl memory burner as another memory test to run -



    that's me just now. Hmm..

    and someone else on overclock.net posted a couple of caps from SOM and DAI showing 3.8-4gb in use at playable frames too

    http://www.overclock.net/t/1514085/o...#post_23362386

    Waiting prior to panic rather than outright panicking about this, so far at least. Have to wonder what the larger hardware sites are doing to try and test this out.
    Last edited by sicemma; 24-01-2015 at 01:09 PM.

  5. #5
    Secondary Hivemind Nexus neema_t's Avatar
    Join Date
    Nov 2011
    Posts
    1,463
    You Won't Believe What They Did Next! Number 8 Will Shock You.

  6. #6
    Network Hub
    Join Date
    Aug 2012
    Posts
    127
    its already debunked and closed at guru3d.

  7. #7
    Network Hub Xzi's Avatar
    Join Date
    Aug 2014
    Location
    Colorado
    Posts
    206
    Quote Originally Posted by peschiNL View Post
    its already debunked and closed at guru3d.
    Good to know. I've been very happy with the performance of my GTX 970, and I've never run in to performance issues, even while using two monitors and maxing out a number of games.
    ~

  8. #8
    Lesser Hivemind Node EPICTHEFAIL's Avatar
    Join Date
    Jan 2013
    Location
    Serbia
    Posts
    875
    Quote Originally Posted by neema_t View Post
    You Won't Believe What They Did Next! Number 8 Will Shock You.
    But Only if You`re a 90`s Kid.
    Guild Wars 2 - Failbaddon, Mars Baneblade, Vulkan Hellhammer
    Elite: Dangerous - EpicTheTrumble
    Steam - EPICTHEFAIL

    Quote Originally Posted by Jams O'Donnell View Post
    I would like to see your mod bits.

  9. #9
    Network Hub
    Join Date
    Nov 2014
    Posts
    256
    Quote Originally Posted by Xzi View Post
    Good to know. I've been very happy with the performance of my GTX 970, and I've never run in to performance issues, even while using two monitors and maxing out a number of games.
    http://www.pcper.com/news/Graphics-C...B-Memory-Issue

    Quote Originally Posted by Nvidia
    The GeForce GTX 970 is equipped with 4GB of dedicated graphics memory. However the 970 has a different configuration of SMs than the 980, and fewer crossbar resources to the memory system. To optimally manage memory traffic in this configuration, we segment graphics memory into a 3.5GB section and a 0.5GB section. The GPU has higher priority access to the 3.5GB section. When a game needs less than 3.5GB of video memory per draw command then it will only access the first partition, and 3rd party applications that measure memory usage will report 3.5GB of memory in use on GTX 970, but may report more for GTX 980 if there is more memory used by other commands. When a game requires more than 3.5GB of memory then we use both segments.

    We understand there have been some questions about how the GTX 970 will perform when it accesses the 0.5GB memory segment. The best way to test that is to look at game performance. Compare a GTX 980 to a 970 on a game that uses less than 3.5GB. Then turn up the settings so the game needs more than 3.5GB and compare 980 and 970 performance again.

    Here’s an example of some performance data:
    GTX 980 GTX 970
    Shadow of Mordor
    <3.5GB setting = 2688x1512 Very High 72 FPS 60 FPS
    >3.5GB setting = 3456x1944 55 FPS (-24%) 45 FPS (-25%)
    Battlefield 4
    <3.5GB setting = 3840x2160 2xMSAA 36 FPS 30 FPS
    >3.5GB setting = 3840x2160 135% res 19 FPS (-47%) 15 FPS (-50%)
    Call of Duty: Advanced Warfare
    <3.5GB setting = 3840x2160 FSMAA T2x, Supersampling off 82 FPS 71 FPS
    >3.5GB setting = 3840x2160 FSMAA T2x, Supersampling on 48 FPS (-41%) 40 FPS (-44%)
    On GTX 980, Shadows of Mordor drops about 24% on GTX 980 and 25% on GTX 970, a 1% difference. On Battlefield 4, the drop is 47% on GTX 980 and 50% on GTX 970, a 3% difference. On CoD: AW, the drop is 41% on GTX 980 and 44% on GTX 970, a 3% difference. As you can see, there is very little change in the performance of the GTX 970 relative to GTX 980 on these games when it is using the 0.5GB segment.
    It was definitely "a thing" people could see and bump up against but it really seemed to depend on how you got there - eg people were reporting that depending on whether you kept a game open or closed while changing settings, or whether you used dsr/other downsampling or hit it with textures, you could get it to jump up above ~3600 usage. Sometimes. And GL furmark seemed to have little problem just cranking shit into memory straight off.

  10. #10
    Network Hub
    Join Date
    Aug 2012
    Posts
    127
    There is no reference to the nvidia quote wich raises red flags for me.

  11. #11
    Network Hub Xzi's Avatar
    Join Date
    Aug 2014
    Location
    Colorado
    Posts
    206
    Quote Originally Posted by sicemma View Post
    It was definitely "a thing" people could see and bump up against but it really seemed to depend on how you got there - eg people were reporting that depending on whether you kept a game open or closed while changing settings, or whether you used dsr/other downsampling or hit it with textures, you could get it to jump up above ~3600 usage. Sometimes. And GL furmark seemed to have little problem just cranking shit into memory straight off.
    Glad I'm not worried about trying to pull down 60 frames at 3456x1944, lol. More than happy enough with how things look on my big 1080p screen.

    As long as VRAM requirements don't keep skyrocketing, the GTX 970 should perform well for a good long while. Shadow of Mordor HD texture pack asks for 6GB VRAM after all, but it's clearly not using that much since I can run it at 60+ FPS with no hiccups.

    Really it's just the poor console ports you have to worry about, but everybody tends to have performance issues with those.
    Last edited by Xzi; 24-01-2015 at 09:04 PM.
    ~

  12. #12
    Network Hub
    Join Date
    Nov 2014
    Posts
    256
    Yeah the main thing I ran into trying to even test it was discovering - holy shit it's actually quite hard to even get to that amount of usage even with insane AA+downsampling before you run into other performance bottlenecks first.

    Just going by other people's grabs from practical use though - DAI, AC:U, SOM etc - and the GL memory test I ran here that seemed to go fine at 3.8 - it seems like this may be the kind of thing the synthetic CUDA benchmark runs into but whatever they do through the card (firmware/driver/geforce experience) is smart enough allocating it not to run into the same issue. Remember when it was turned into a huge scandal that you could rename quake to quack.exe and get different numbers out of a timedemo? That seems very long ago now.

  13. #13
    Why was this not posted in the tech forum instead?

    Anyhow, techreport's report theorizes that this may be due to the 970's ROP issue:

    http://techreport.com/news/27721/nvi...location-issue

    We explored a similar datapath issue related to the GTX 970's disabled SMs in this blog post. In that case, we looked at why the GTX 970 can't make full use of its 64 ROP partitions at once when drawing pixels. Sounds to me like this issue is pretty closely related.

  14. #14
    Lesser Hivemind Node
    Join Date
    Aug 2013
    Location
    Germany
    Posts
    749
    Quote Originally Posted by Xzi View Post
    As long as VRAM requirements don't keep skyrocketing, the GTX 970 should perform well for a good long while. Shadow of Mordor HD texture pack asks for 6GB VRAM after all, but it's clearly not using that much since I can run it at 60+ FPS with no hiccups.
    It doesn't even make sense to use it at 1080p, since you won't be able to see any difference. The HD pack is meant for much higher resolutions.
    Resident graphics snob.

    And remember: Bad practices often become a trend, if not dealt with.

  15. #15
    Network Hub Xzi's Avatar
    Join Date
    Aug 2014
    Location
    Colorado
    Posts
    206
    Quote Originally Posted by DanMan View Post
    It doesn't even make sense to use it at 1080p, since you won't be able to see any difference. The HD pack is meant for much higher resolutions.
    Oh it definitely makes a visible difference.
    ~

  16. #16
    Lesser Hivemind Node
    Join Date
    Aug 2013
    Location
    Germany
    Posts
    749
    Quote Originally Posted by Xzi View Post
    Oh it definitely makes a visible difference.
    I'm not gonna argue about it. There's plenty of analysis all over the net. I've also written about it elsewhere on this forum already.
    Last edited by DanMan; 25-01-2015 at 03:01 AM.
    Resident graphics snob.

    And remember: Bad practices often become a trend, if not dealt with.

  17. #17
    Activated Node
    Join Date
    Jan 2015
    Posts
    33
    In response to the OP - the tool is broken not the video card.



    http://www.reddit.com/r/pcmasterrace...70_vram_issue/

  18. #18
    Secondary Hivemind Nexus squirrel's Avatar
    Join Date
    Jun 2011
    Posts
    1,933
    To put simple, VRAM be divided into 3.5GB and 0.5GB groups. If the half GB VRam be accessed, this section may be not fully utilized, hence this testing result.

    So you guys' opinion is that it is users' fault for not knowing this prior. You don't figure it out yourself, then you are a fool and don't blame on Nvidia for hiding essential technical details. After all, it is cheaper than 980, of course you deserve to be delivered a product with such handicap.

    What kinda business conduct is that, westerners?!

  19. #19
    Network Hub
    Join Date
    Nov 2014
    Posts
    256
    I guess the issue is more, what should have happened?

    They don't bin the chips to enable a 970 product, and just release a more expensive, downclocked 980 instead?

    They don't partition the memory, and lose a relative 1-3% performance at all times regardless of the memory load of the next frame displayed?

    Prioritise a synthetic benchmark over actual games? Create a profile for all conceivable CUDA apps in geforce experience to stop people shooting their own foot off with it?

  20. #20
    Secondary Hivemind Nexus
    Join Date
    Feb 2013
    Posts
    1,639
    It seems quite reasonable really, I see no inherent problem with it. I wonder if this problem persists in older GPU's as well.
    PS2/NS2/Mumble: SirWigglyBottom
    Steam: The Almighty Snark

    Let's play flightsimulators: http://www.rockpapershotgun.com/foru...793#post659793

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
  •