|
Anime Schoolgirl posted:lol if you read D&D and not C-SPAM LOL if you read D&D or C-SPAM.
|
#
?
Feb 14, 2018 05:47
|
|
|
|
| # ? May 19, 2024 12:12 |
|
|
lol just lol
|
|
#
?
Feb 14, 2018 06:04
|
|
|
I just sent in a 980ti Hybrid back to EVGA. Any bets as to what they'll give me in return? I did send them a note saying that while yes I'm running 2 980ti's in SLI I'd actually prefer a 1080 back or I'll pay the difference to a 1080ti. Fingers crossed. A friend sent back a non-hybrid 980ti and they sent him a 1080 so I'm just hoping they aren't bothering with refurbing the hybrids anymore.
|
|
#
?
Feb 14, 2018 07:02
|
|
|
redeyes posted:They named the compute card Ampere and the gaming card Turing. Makes perfect sense. Turing because a sufficiently advanced AI would enjoy pwning noobs
|
|
#
?
Feb 14, 2018 07:14
|
|
|
lol if you're not in YOSPOS
|
|
#
?
Feb 14, 2018 07:48
|
|
|
Hoping this means the card doesn't fail to work with others, is castrated, and then commits suicide.
|
|
#
?
Feb 14, 2018 09:46
|
|
|
so no SLI, removed mining ability and running unsigned bios bricks the card.
Fauxtool fucked around with this message at 10:16 on Feb 14, 2018 |
|
#
?
Feb 14, 2018 10:14
|
|
|
Man things got dark pretty fast.
|
|
#
?
Feb 14, 2018 10:21
|
|
|
Lockback posted:You'll probably be fine if 60 is your target, though a 1070 would be a better choice. My brother has basically your exact setup though in 1080p and usually gets in the 80s on high settings.
|
|
#
?
Feb 14, 2018 10:36
|
|
|
gently caress it I am just going to keep the 1070Ti. Honestly I don't see them launching cards any cards less than $600 anyways. If I am wrong and a Turing 2070 launches at $399 with 30% more performance than my 1070Ti, oh well. Should be a big step up from my 7870TE. Actually the GTX 1060 6GB I had before was mildly disappointing, about 60-80% faster than my TE which is surprising considering it's basically an overclocked, gimped 7950 from 6 years ago. At least the 1070Ti should be a full 100% faster.
|
|
#
?
Feb 14, 2018 11:42
|
|
|
I not putting it past nVidia to sandbag this generation and just improve power consumption and performance per watt or something. What's the competition going to do? Mine them to 1st place?
|
|
#
?
Feb 14, 2018 14:05
|
|
|
There are worse things in life than improving power consumption and performance-per-watt. I mean, just look at Intel's 5 - 10% perf bump YOY for the past goddamn who knows how long.
|
|
#
?
Feb 14, 2018 14:17
|
|
|
I'd love to gently caress around with Tensor cores, but their consumer cards probably won't even have them. I wonder if there are private Tensor mining algos out there using Amazon p3 instances to mine.
|
|
#
?
Feb 14, 2018 15:07
|
|
|
SwissArmyDruid posted:There are worse things in life than improving power consumption and performance-per-watt. I mean, just look at Intel's 5 - 10% perf bump YOY for the past goddamn who knows how long. I think it's more like 1-5%
|
|
#
?
Feb 14, 2018 15:22
|
|
|
I mean we're long overdue for a refresh, and I've been convinced that Pascal limited hard by Nvidia and has plenty of headroom for a total rebrand. I do trust them quite a bit more to do rebrands right though, and its hardly a bad thing if thats the case. Just for the past ...2-3+ years Nvidia hasn't been rebranding anything and AMD has been rebranding everything in the worst way possible so in case anybody doesnt know it was customary there for a little while where if you rebrand a lineup you move the previous generations model numbers down the line. Though there were plenty of exceptions or "almost kinda" to that, but a rough example would be like a GTX 670 would be similar to the GTX 760. Then AMD threw all that poo poo out the window when they rebranded most of the 200 series into the 300 series and then never stopped being lame about that. Maybe AMD has re-set this precedence, to be fair even longer ago this was nothing new, so even a bumping of clockspeeds for straight model to model rebrands is possible. But that is the worst case scenario and there really isn't any indication that will happen.
|
|
#
?
Feb 14, 2018 15:56
|
|
|
NV can just take Pascal and do a dumb shrink on 10nm with more coars + undervolt treatment and still come out with scary perf/W numbers.
|
|
#
?
Feb 14, 2018 16:59
|
|
|
Can someone explain to me, In pleb terms, why you can't mix Nvidia GPUs with AMD GPus in the same computer (Not running SLI). I had a friend ask me but I honestly don't know enough about windows drivers and backend to explain why I think it would never work.
|
|
#
?
Feb 14, 2018 17:22
|
|
|
TipsyMcStagger posted:Can someone explain to me, In pleb terms, why you can't mix Nvidia GPUs with AMD GPus in the same computer (Not running SLI). There's nothing technically stopping you from doing it, it's just mostly pointless because barely any software can do anything useful with an AMD+NV configuration. https://www.youtube.com/watch?v=XrpTwUJTVCQ ...unless you really like playing Ashes of the Singularity I guess.
|
|
#
?
Feb 14, 2018 17:34
|
|
|
TipsyMcStagger posted:Can someone explain to me, In pleb terms, why you can't mix Nvidia GPUs with AMD GPus in the same computer (Not running SLI). I'd assume you'd have driver headaches since usually those aren't designed to work together in the OS, but what is your use case for that? Crypto mining would be the only thing that might work and I wouldn't be surprised if the mining software wasn't well designed to handle that.
|
|
#
?
Feb 14, 2018 18:15
|
|
|
It actually does work, on windows 10 anyway. I could be playing a game on one card while mining on the other.
|
|
#
?
Feb 14, 2018 19:02
|
|
|
Zero VGS posted:Finally, a graphics card as gay as me. But will it be as hot?
|
|
#
?
Feb 14, 2018 19:07
|
|
|
Wistful of Dollars posted:But will it be as hot? Now I wonder if the cards will run Turing Hot or Benedict Cumberbatch as Turing Hot.
|
|
#
?
Feb 14, 2018 19:52
|
|
|
TipsyMcStagger posted:Can someone explain to me, In pleb terms, why you can't mix Nvidia GPUs with AMD GPus in the same computer (Not running SLI). because if nvidia driver detects an amd gpu in your pc, it will purposefully gut some functionality.
|
|
#
?
Feb 14, 2018 19:55
|
|
|
Windows 10 does indeed handle GPU switching wat better than the past. I run into this fairly often at work. Of course the most common thing is switching from igpu and dgpu but AMD and Nvidia seemingly works just as well. There used to be no point at all but as Alpha Mayo said there are now scenarios where its desirable. As far as conflicts I have no idea my experiences with it are all not gaming related
|
|
#
?
Feb 14, 2018 20:04
|
|
|
Alpha Mayo posted:I'd love to gently caress around with Tensor cores, but their consumer cards probably won't even have them. I wonder if there are private Tensor mining algos out there using Amazon p3 instances to mine. there are no algorithms that need mixed-precision matrix math
|
|
#
?
Feb 14, 2018 20:06
|
|
|
Truga posted:because if nvidia driver detects an amd gpu in your pc, it will purposefully gut some functionality. You mean PhysX right? They removed that arbitrary limitation around 2016, you can render on an AMD GPU while physics-ing on an Nvidia GPU if you really want to.
|
|
#
?
Feb 14, 2018 20:07
|
|
|
Paul MaudDib posted:there are no algorithms that need mixed-precision matrix math I get the feeling like tensor cores are the new FP64, where no home user actually uses it, they just post about using it to look smart.
|
|
#
?
Feb 14, 2018 20:09
|
|
|
Having a really weird issue where the Nvidia drivers keep changing the "Content type reported to the display" setting to Auto-Select after every restart. I need it kept at Desktop Programs because auto-select causes black flashing when loading videos and then the desktop looks ugly and fuzzy forever after (this is probably because I use an old HDTV as a display instead of a real monitor, but anyhow.) This wasn't happening until the "Spectre fix" drivers.
|
|
#
?
Feb 14, 2018 20:28
|
|
|
"leak" circulating around the net today: for reference, a Titan Xp has 3840 CUDA cores, 1080 Ti has 3584, and Volta/Titan V has 5120 e: if we assume that 18Gbps ram is on a 384bit bus, we are looking at 864GB/s of bandwidth. Titan Xp has 547, Titan V has 652 (but its HBM) to clarify this is probably horseshit, but fun to theorycraft Cygni fucked around with this message at 20:42 on Feb 14, 2018 |
|
#
?
Feb 14, 2018 20:35
|
|
|
Irresponsible amounts of repeated taglines.
|
|
#
?
Feb 14, 2018 20:38
|
|
|
Don't drink and game irresponsibly (tm)
|
|
#
?
Feb 14, 2018 20:42
|
|
|
Cygni posted:"leak" circulating around the net today: FE 2080 for $999 then?
|
|
#
?
Feb 14, 2018 20:43
|
|
|
fabricated and homosexual 4480 CUDA cores at 2.0ghz would be nearly 18 TFLOPS, not 15.
|
|
#
?
Feb 14, 2018 20:43
|
|
|
Refers to Ampere, not Turing
|
|
#
?
Feb 14, 2018 20:43
|
|
|
Those are some Titan XYZ numbers there. No way it's real. If it is no gamer will ever be able to purchase one.
|
|
#
?
Feb 14, 2018 20:47
|
|
|
PerrineClostermann posted:Refers to Ampere, not Turing to be fair, wasnt that leak was a dude posting on reddit? our desperation for graphics cards has reached considerable lows. unrelated, but I still think its unlikely that Nvidia will launch a 18Gbps GDDR6+384bit bus off the bat
|
|
#
?
Feb 14, 2018 20:48
|
|
|
Cygni posted:to be fair, wasnt that leak was a dude posting on reddit? our desperation for graphics cards has reached considerable lows. The Turing rumour originated from Reuters: https://www.reuters.com/article/us-...f-idUSKBN1FT2AW
|
|
#
?
Feb 14, 2018 20:57
|
|
|
Apparently there's thought that Turing will replace Volta in the compute market, and Ampere will be the mass market.
|
|
#
?
Feb 14, 2018 20:59
|
|
|
Zero VGS posted:Now I wonder if the cards will run Turing Hot or Benedict Cumberbatch as Turing Hot. guessing it won't support Apple
|
|
#
?
Feb 14, 2018 21:01
|
|
|
|
| # ? May 19, 2024 12:12 |
|
|
PerrineClostermann posted:Apparently there's thought that Turing will replace Volta in the compute market, and Ampere will be the mass market. that one is even more fabricated and homosexual. Volta is only 8 months old.
|
|
#
?
Feb 14, 2018 21:02
|
|