|
Bad Parenting posted:Also whats the deal with the new power connectors, from what I've read they're only needed for the founder's edition cards? I usually go with EVGA GPUs so do I not need to worry about that? if the card you're getting has the 12-pin connector, it'll come with an adapter to convert two 8-pin connectors, so it shouldn't be an issue as long as you already have the two 8-pin connectors Black Griffon posted:Also curious about the reference model and availability and how it differs from AIB. Price? Extra doodads? Overclocking? AIBs tend to be pre-overclocked, and can (read: not always) have a more robust cooling solution than reference models, but whether they actually do and whether that matters can't really be known until reviews start coming in from people who've tested both
|
# ? Sep 2, 2020 10:44 |
|
|
# ? May 31, 2024 03:13 |
|
my case has 315mm of space for the GPU will that be enough for a 3080? It seems the founders' edition is 285mm, but AIBs are 300/320mm apparently... I hope someone's going to do a short but thick version, I don't mind occupying 3 slots and I am not sure how widely available the founder Nvidia card will be
|
# ? Sep 2, 2020 10:47 |
|
Heliosicle posted:Are the founders editions only ever purchased from Nvidia? Any European goons know what the availability here is usually like? I could have sworn I've seen Best Buy sell founders editions before, like the actual Nvidia-branded ones. edit: Oh yeah they still do: https://www.bestbuy.com/site/nvidia...p?skuId=6291646
|
# ? Sep 2, 2020 11:01 |
|
Crunchy Black posted:...if by ML market in this instance you mean folks who are doing it under their desks? Sure?
|
# ? Sep 2, 2020 11:02 |
gradenko_2000 posted:AIBs tend to be pre-overclocked, and can (read: not always) have a more robust cooling solution than reference models, but whether they actually do and whether that matters can't really be known until reviews start coming in from people who've tested both Right. Founders looks absolutely fantastic, but I guess AIB might be a better buy if the reviews pan out. These are the ones up on my most trusted site. No EVGA up, but that might change.
|
|
# ? Sep 2, 2020 11:18 |
|
Carecat posted:Please, you think they haven't thought of that? Oh god, that thing. I have the iChill x4 variant of the 1070 and it's already bloody enormous. Quiet, good performer and cool, but hyuuge. I recently upgraded to a h550 case with a 360 cooler and had to get a riser to jam the drat thing in. Having a 3080 in a tidy box will be a breath of fresh air.
|
# ? Sep 2, 2020 11:23 |
|
NoDamage posted:Um, yeah? Putting together a deep learning rig using a couple of consumer-grade GPUs is not exactly uncommon these days. Well, no poo poo you can build one yourself. I'm talking about actual, recognized channel partners where you can buy a SKU that shows up in a box and you power it on.
|
# ? Sep 2, 2020 11:27 |
|
Happy Noodle Boy posted:Something about 2077 in general feels off and I just can’t tell if it’s the gameplay not being good enough to carry everything else or if everything else will carry the gameplay. Glad I'm not the only one The Blade Runner glory shots of the city are sure to please, but as soon as they show real gameplay it looks like a somewhat lower rent Deus Ex clone.
|
# ? Sep 2, 2020 11:31 |
But all I want is a Deus Ex clone.
|
|
# ? Sep 2, 2020 11:36 |
|
Honestly i haven't even cared about Cyberpunk (never cared for anything CDPR did), I'm just excited about Flight Sim 2020 and how we finally have a game that goes balls to the wall with visual quality
|
# ? Sep 2, 2020 11:36 |
|
lol FS DLSS/RTX when?
|
# ? Sep 2, 2020 11:36 |
I ordered Cyberpunk when that favored store upthread had a price error two years ago. Loved Witcher 3, have absolutely no expectations for Cyberpunk, probably gonna have fun because I paid less than 40 bucks for it. W3, as far as I'm concerned, was a case of the planets aligning.
|
|
# ? Sep 2, 2020 11:42 |
|
Black Griffon posted:W3, as far as I'm concerned, was a case of the planets aligning.
|
# ? Sep 2, 2020 11:49 |
|
Crunchy Black posted:Well, no poo poo you can build one yourself. I'm talking about actual, recognized channel partners where you can buy a SKU that shows up in a box and you power it on. Like these? https://bizon-tech.com/bizon-g3000.html https://lambdalabs.com/deep-learning/workstations/4-gpu/basic/customize https://orbitalcomputers.com/deep-learning.php I'm not sure what your point is? From what I've seen the type of people that work on this stuff tend to be the same type of people that like messing with computers and also like to build their own poo poo.
|
# ? Sep 2, 2020 11:54 |
|
Lenovo confirmed the existence of a 16GB 3070 Ti https://www.pcgameshardware.de/Nvidia-Ampere-Grafikkarte-276295/News/RTX-3070-Ti-mit-16GB-VRAM-gesichtet-1357337/
|
# ? Sep 2, 2020 12:09 |
If using that rtx voice means that people never have to hear my box jade keys in discord then I will happily pay for that.
|
|
# ? Sep 2, 2020 12:15 |
|
shrike82 posted:Lenovo confirmed the existence of a 16GB 3070 Ti Still only using GDDR6 tho. e: English source https://videocardz.com/newz/nvidia-geforce-rtx-3070-ti-spotted-with-16gb-gddr6-memory
|
# ? Sep 2, 2020 12:20 |
|
AirRaid posted:Still only using GDDR6 tho. Is there a real world, noticeable difference between 6 and 6x, or is this an on paper its better thing ?
|
# ? Sep 2, 2020 12:22 |
|
6x has essentially double the bandwidth since it uses the PAM4 encoding which can encode 2 bits of data per waveform. from the Presser last night - AirRaid fucked around with this message at 12:26 on Sep 2, 2020 |
# ? Sep 2, 2020 12:23 |
|
Whiskey A Go Go! posted:If using that rtx voice means that people never have to hear my box jade keys in discord then I will happily pay for that. 100% confirming that RTX Voice totally kills all noise from my box navies. It's pure magic. (You can also edit the installer and it works on a 1080ti, albeit with occasional silent crashes that make me look like a dingus by talking in discord for a while without realizing my mic isn't working.)
|
# ? Sep 2, 2020 12:30 |
|
CactusWeasle posted:Is there a real world, noticeable difference between 6 and 6x, or is this an on paper its better thing ? We don't know yet (other than AirRaid's point on what Nvidia have said), but the assumption that there isn't a real world difference between 6 and 6x is to assume that NVIDIA deliberately made a bad flagship card.
|
# ? Sep 2, 2020 12:37 |
|
shrike82 posted:Commercial outfits have been using ML servers with consumer x80/x80 Ti/Titans forever. There seems to be a general misunderstanding among layfolk about what hardware ML shops/academic research groups use - it's common for outfits to have staff use desktop/laptops with consumer cards to prototype, followed by local x4/x8 consumer cards on servers for prototyping/training, then scaling up to multi-RTX8000/V100 big iron and 3rd party cloud instances if necessary for production/specialized training tasks. Are they still humping the top covers on server chassis? (Ha ha, no, really: https://www.servethehome.com/avert-your-eyes-from-the-server-humping-trend-in-the-data-center/)
|
# ? Sep 2, 2020 13:07 |
|
Radeon and Navi2X feels very dead in the water RN. If the XSX is using full RDNA2, not some hybrid of IP blocks, we already can gauge a best case estimation of "Big Navi". XSX is 52DCU @ 1.825Ghz, and according to developers they have around 2080/2080 Super performance to play around with. The difference between a RX 5700 XT and RTX 2070 Super is about 8% in the 2070's favor, and both have an equal number of ALUs. Unless there is some regression elsewhere, the XSX has about 8% more ALUs than the 2080 Super, so it looks like per ALU nothing has changed between RDNA1 and RDNA2. So assuming perfect scaling, 80CU (5120 ALUs), Big Navi should be equal to a 4726 ALU Turing card. A 2080 Ti has 4352 ALUs, which would peg Big Navi @ ~9% better. So RTX 3070 territory, with overall fewer features. At optimistically 275W. loving lmao.
|
# ? Sep 2, 2020 13:13 |
|
I guess I’m not ready to write them off until they announce their cards and we get independent benchmarks for nvidia 3000. They certainly have a uphill battle but if they can get between 3070 and 3080 performance they’ll be fine. I wonder if they will have software features to match what nvidia has or will be launching with the new cards.
|
# ? Sep 2, 2020 13:19 |
|
Do we know who's making the memory for the NVidia 3xxx series? Is that Micron, Samsung, or both? Does it depend on card vendor?
|
# ? Sep 2, 2020 13:20 |
|
it's more likely Microsoft develops a DLSS analogue for their consoles using DirectML and then opens it up for PC use - their hot chip presentation mentioned ML use cases on their XSX GPU hardware including upscaling
|
# ? Sep 2, 2020 13:22 |
|
MikeC posted:To be fair, he did have sources that allowed him to napkin math a "3080 Ti" at 2x 2080 back in late March. Exactly where Jensen's own presentation today said it would be. who could have guessed that a new xx80ti card would be roughly twice as fast as the previous gen xx80 card, that's never happened before https://www.techpowerup.com/review/nvidia-geforce-rtx-2080-ti-founders-edition/33.html https://www.techpowerup.com/review/nvidia-geforce-gtx-1080-ti/30.html https://www.techpowerup.com/review/nvidia-geforce-gtx-980-ti/31.html
|
# ? Sep 2, 2020 13:27 |
|
pmchem posted:Do we know who's making the memory for the NVidia 3xxx series? Is that Micron, Samsung, or both? Does it depend on card vendor? As far as I can tell Micron is the only company company making GDDR6X right now, which is on the 3080 and 3090. The 3070 uses GDDR6 which I believe is made by Samsung, micron and maybe another company.
|
# ? Sep 2, 2020 13:27 |
|
pmchem posted:Do we know who's making the memory for the NVidia 3xxx series? Is that Micron, Samsung, or both? Does it depend on card vendor? I think only Micron. They officially announced GDDR6X yesterday. https://www.micron.com/products/ultra-bandwidth-solutions/gddr6x
|
# ? Sep 2, 2020 13:28 |
|
AirRaid posted:Still only using GDDR6 tho. Good, so there's a chance it won't pull 300w and can fit in smaller cases...
|
# ? Sep 2, 2020 13:37 |
|
MikeC posted:Leakers have gotten Ampere mostly right. While the details might have been off, we have largely known what their performance was going to be like and that they will be power hogs. Performance numbers for the 3070 and 3080 were reported accurately as early as May and June of this year. The fact that they were stuck on a power-hungry node was well known by early July. Sorry to necro this from pages ago, but I wanted to clarify I never said rdna2 definitely won't match the 3080. What I said was I think they won't beat ampere's efficiency. If they match the 3080 I expect they will need as much or more power consumption to do it, even with a node advantage.
|
# ? Sep 2, 2020 13:46 |
|
https://twitter.com/HardwareUnboxed/status/1301048043232358400?s=20 lol
|
# ? Sep 2, 2020 13:48 |
|
DF sometimes isn't great at navigating the thin line between exclusive coverage and being a PR mouthpiece. e:perhaps worth noting they've done content explicitly sponsored by nvidia as recently as a few weeks ago Rinkles fucked around with this message at 14:08 on Sep 2, 2020 |
# ? Sep 2, 2020 14:04 |
|
Rinkles posted:DF sometimes isn't great at navigating the thin line between exclusive coverage and being a PR mouthpiece. i think they just like to be nice to everyone and let The Almighty Framerate Line do the heavy criticism, which is fine by me
|
# ? Sep 2, 2020 14:13 |
|
Looks like MSI fixed up pages for some of their cards. The Gaming Trios are looking to be 323mm long and the Ventus 3X 305mm. The dual fan 3070 setups are still TBD on specs but should be much shorter of course. So if I'm gonna be sticking with MSI, it'll be the Ventus that will fit my Meshify C without much issue.
8-bit Miniboss fucked around with this message at 14:20 on Sep 2, 2020 |
# ? Sep 2, 2020 14:18 |
|
Encrypted posted:Wanna bet the 3090 isn't that much faster than the 3080 besides running things at 8k due to having 24GB VRAM. Not sure that VRAM usage at 8k matters considering this:
|
# ? Sep 2, 2020 14:20 |
|
What an unnecessary low shot at DF. Not the first time HU have slung poo poo at better reviewers without solid basis. He could've framed this in a way that is less accusatory and made a more compelling point
|
# ? Sep 2, 2020 14:26 |
|
Generic Monk posted:i think they just like to be nice to everyone and let The Almighty Framerate Line do the heavy criticism, which is fine by me The Almighty Framerate Line tends to be pretty favorable when you get to cherry pick the numbers.
|
# ? Sep 2, 2020 14:27 |
|
So 30x0 supers when?
|
# ? Sep 2, 2020 14:35 |
|
|
# ? May 31, 2024 03:13 |
|
OhFunny posted:Funny enough this happened: Exact opposite. VSync is on in Borderlands 3 so it's a locked 60 at a completely worthless benchmark.
|
# ? Sep 2, 2020 14:38 |