While you're here, please consider supporting GamingOnLinux on:
Reward Tiers: Patreon. Plain Donations: PayPal.
This ensures all of our main content remains totally free for everyone! Patreon supporters can also remove all adverts and sponsors! Supporting us helps bring good, fresh content. Without your continued support, we simply could not continue!
You can find even more ways to support us on this dedicated page any time. If you already are, thank you!
Reward Tiers: Patreon. Plain Donations: PayPal.
This ensures all of our main content remains totally free for everyone! Patreon supporters can also remove all adverts and sponsors! Supporting us helps bring good, fresh content. Without your continued support, we simply could not continue!
You can find even more ways to support us on this dedicated page any time. If you already are, thank you!
Login / Register
- Fedora KDE gets approval to be upgraded to sit alongside Fedora Workstation
- Steam gets new tools for game devs to offer players version switching in-game
- Palworld dev details the patents Nintendo and The Pokemon Company are suing for
- GOG launch their Preservation Program to make games live forever with a hundred classics being 're-released'
- Sony say their PSN account requirement on PC is so you can enjoy their games 'safely'
- > See more over 30 days here
-
Classic Unreal Tournament and Unreal now easier to down…
- emphy -
Minecraft-like free and open source game VoxeLibre (for…
- kneekoo -
Mesa 24.2.7 out now and Mesa 24.3 may come sooner than …
- nnohonsjnhtsylay -
Civilization VI, Civilization V, lots of DLC and other …
- Pompesdesky -
Mesa 24.2.7 out now and Mesa 24.3 may come sooner than …
- KROM - > See more comments
- Steam and offline gaming
- Snak30 - Does Sinden Lightgun work?
- helloCLD - No more posting on X / Twitter
- Liam Dawe - Weekend Players' Club 10/11/2024
- Pengling - Upped the limit on article titles
- eldaking - See more posts
View PC info
Pretty much ALL GTX 970s are having an issue where the memory speed/bandwidth basically tanks when VRAM fills past 3.5gb, which it usually won't unless forced artificially.. I know I paid for a 4gb card and not a 3.5gb one. It is not an issue on the 980's so some are speculating that it could be due to the cut down 980 die used in the 970 which would make this a hardware problem that would not be solvable by any driver/firmware update. Which would mean a massive GTX 970 recall
ManuelG of Nvidia has acknowledged the problem and says they are looking into it.
ManuelG Nvidia Post
Scary stuff, I myself wondered why FarCry4 wouldn't load up my VRAM past 3.5gb on my 970's playing @ 5760x1080 with much of the settings cranked to the max when I had seen screenshots of it using every bit of VRAM a 980 had.
I did some quick searching but came up empty handed, so does anyone know of a tool for Linux to test the VRAM in linux in a similar manner as the windows users have been testing (see screenshot below). I very much doubt the issue is OS dependent, but it certainly cannot hurt to find out.
The internet is not going to allow Nvidia to sweep this under the rug, and hope everyone forgets. I have not seen this news posted to any of the Linux sites I frequent so I thought I would make everyone here aware. If you are considering picking up a 970, I suggest either waiting for a response/fix from Nvidia or grabbing a 980 instead.
Some Links:
Lazygamer
Reddit
Guru3d
Overclock.net
Nvidia GeForce Forum Post
View cookie preferences.
Accept & Show Accept All & Don't show this again Direct Link
View PC info
I was hopeful that would be the case, but they went the better route (for them), and just deny and "explain" away the issue. It's suppose to be that way lol. Will be interesting to watch this for a few more days, it doesn't seem like many are buying Nvidia's response.
Guru3D: Does the GeForce GTX 970 have a memory allocation bug ? (updated + NV answer)
View PC info
View PC info
http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-970/specifications
To me the problem is that Nvidia lied to their customers.
View PC info
View PC info
All of the memory can be allocated.
However, all memory above 3.5 GB has a MUCH slower memory bandwidth.
According to the reputable German news site heise.de*, the crossbar
configuration does not allow those last 512 MB to be accessed equally fast.
See the numbers in EKRboi's image for channel no. 25 to 29:
~22 GB/s instead of ~150 GB/s!
By design this cannot be fixed in a driver or firmware upgrade.
When more than 3.5 GB are allocated strange effects are to be expected,
e.g. noticeable micro jitter due to the partially slow access to the problematic
memory resources.
However, a driver limiting the card to 3.5 GB seems to be the best solution.
[*] http://www.heise.de/newsticker/meldung/Nvidia-fuehrt-Kaeufer-der-GeForce-GTX-970-hinters-Licht-Nur-3-5-statt-4-GByte-RAM-schnell-angebunden-2528588.html
View PC info
Article from WCCFtech (basically same as above)
Anandtech article ("simpler" explaination)
So, they have now admitted that there was some miscommunication between Nvidia's engineers and the PR team. So what everyone thought they were buying is NOT what they got. That last bit of ram CAN be used but unless it is used perfectly (it normally isn't) then the memory speed of that last bit of ram is severely degraded. The big issue here is the "false advertisement" whether by accident or on purpose it doesn't matter. As a consumer I purchased a card with 4bg of VRAM that runs at 224 GB/s* with 64 ROPs and 2mb of L2 Cache and that is not what I received. I imagine Nvidia is going to be in some hot water in the UK with their consumer protection laws.. here in the US we have "corporate protection from the consumer laws" so who knows what is going to happen here.
*"To those wondering how peak bandwidth would remain at 224 GB/s despite the division of memory controllers on the GTX 970, Alben stated that it can reach that speed only when memory is being accessed in both pools." (lol.. so in other words.. not all the time)
what we were told:
GTX970
64 ROPS
2048kb L2 cache
what we actually got:
GTX970
56 ROPs
1792kb L2 cache
View PC info
I find that extremely unlikely.. I would want to know what people thought of my labor of love if I were a designer/maker of these chips.. What I do think is likely is that someone noticed and Nvidia knew they screwed up and I'm guessing they just hoped nobody would ever put 2 and 2 together. After all, at the moment it is a small fraction of people running multiple GPU's and high resolutions who are actually running up against the memory issues. It could be a year or 2 before many games are running up against that VRAM limit on single card, single 1080p monitor setups.
View PC info