|
they call it the 3080tie because they've got you tied up and bound at their whim if you want a card maybe that's why people are willing to pay so much
|
# ? Jun 5, 2021 08:57 |
|
|
# ? Jun 10, 2024 11:29 |
|
The 3080ti is loving awful value It gets away with it simply by existing in tyool 2021, in any other year it would be a massive flop See also: 2080/2080ti
|
# ? Jun 5, 2021 09:07 |
|
https://twitter.com/charlielangton/status/1400811059452252162?s=20
|
# ? Jun 5, 2021 09:14 |
|
https://www.pcmag.com/news/meet-the-people-who-camped-out-at-best-buy-for-nvidias-rtx-3080-ti-gpuquote:The limited supplies caused at least a few customers to get desperate. “People offered to buy our seats. They said, ‘We’ll give you whatever you want. $2,500 for a seat,’” Arciga recalled.
|
# ? Jun 5, 2021 09:35 |
|
Oh my god
|
# ? Jun 5, 2021 09:39 |
|
Sphyre posted:this comment didnt square up at all with the near universal confusion caused by the 3080 Thai, so i went back and checked: good job man. thai sayer ppl begone
|
# ? Jun 5, 2021 09:40 |
|
dude who showed up in the AM and didnt make it to finals offered to buy from us and one guy asked 3200 since that was the scalped 3090 price and he agree. Granted he said he'd go to the bank and get it but as far as I could tell he never came back and I can't tell if they met up later
|
# ? Jun 5, 2021 09:41 |
|
lmfao
|
# ? Jun 5, 2021 09:41 |
|
Unreal Engine 5 marks Raytracing as "deprecated", considers the performance cost too high compared to their own "Lumen" solution for global illumination. Lumen does not use RT cores. https://www.youtube.com/watch?v=aXzyTV-KvYo&t=256s
|
# ? Jun 5, 2021 12:36 |
|
mcbexx posted:Unreal Engine 5 marks Raytracing as "deprecated", considers the performance cost too high compared to their own "Lumen" solution for global illumination. Lumen does not use RT cores. quote:Lumen uses Software Ray Tracing through Signed Distance Fields by default, but can achieve higher quality on supporting video cards when Hardware Ray Tracing is enabled. quote:Hardware Ray Tracing supports a larger range of geometry types than Software Ray Tracing, in particular it supports tracing against skinned meshes. Hardware Ray Tracing also scales up better to higher qualities — it intersects against the actual triangles and has the option to evaluate lighting at the ray hit instead of the lower quality Surface Cache. However, Hardware Ray Tracing has significant scene setup cost and currently cannot scale to scenes with more than 100,000 instances. Dynamically deforming meshes, like skinned meshes, also incur a large cost to update the Ray Tracing acceleration structures each frame, proportional to the number of skinned triangles. First quote was the second line in the article.
|
# ? Jun 5, 2021 12:48 |
|
King of Solomon posted:You sign up to get a check mark in a few hours saying you don't get to buy anything. I did it, I got the checkmark
|
# ? Jun 5, 2021 12:52 |
|
Sininu posted:Steve or his team didn't read documentation well enough before doing the video. Seems like the choice between software RT vs hardware RT will largely depend on how your game works in terms of environments and scale.
|
# ? Jun 5, 2021 12:59 |
|
Lumen isn't as good as RTX/hardware raytracing but it doesn't need to be. It does 80% of the results for a fraction of the cost, and does not need specialized hardware. My 1080ti is running lumen demos at 1440p and manages to keep 60fps in a lot of them, meanwhile it struggles to do Quake 2 RTX at anything higher than 10fps at 320x240. Epic really did strike gold with UE5, not just lumen but nanite as well. Zedsdeadbaby fucked around with this message at 16:30 on Jun 5, 2021 |
# ? Jun 5, 2021 13:07 |
|
#heroesofcapitalism
|
# ? Jun 5, 2021 13:21 |
|
I couldn't get Lumens HWRT path to work in the first early access build to compare the performance, apparently it works if you compile the engine from master but it'll wait for the next build Software SDF tracing seems to work pretty well for global illumination and rough reflections but the inaccuracy becomes obvious on sharper reflections, everything looks like it's made of clay Sininu posted:Steve or his team didn't read documentation well enough before doing the video. Yeah what they deprecated is raytracing as it existed in UE4, which was a fairly brute force approach, and replaced it with Lumens more efficient surface caching approach repiv fucked around with this message at 13:52 on Jun 5, 2021 |
# ? Jun 5, 2021 13:38 |
|
Sininu posted:Steve or his team didn't read documentation well enough before doing the video. Yeah that's an insanely wrong, dumb and bad take by Steve. Getting this right does not require much reading. Epic is doing exactly what you would reasonably expect - creating an engine GI and reflection solution that is forward-looking and scalable, able to extract useful performance from limited console RT hardware, scale up to support more powerful RT hardware, while retaining support for hardware that isn't fully obsolete yet. I could understand getting it wrong if Epic had done something surprising here, but they did exactly what one would have expected them to do...
|
# ? Jun 5, 2021 14:21 |
|
Everyone posted:Steve Messed Up! I mean, yeah, he should have re-shot or overdubbed that part, but he did include an in-video correction at the end of the segment, so I don't get what all the fuss is about?
|
# ? Jun 5, 2021 15:02 |
|
mcbexx posted:Wenn ist das Nunstück git und Slotermeyer? Ja! Beiherhund das Oder die Flipperwaldt gersput! Email that to the Nvidia HQ. e: Nvidia really should start to slowly implement performance blockers for GTX1000 and older series in their new drivers. That would be really fitting for these hellworld years. AutismVaccine fucked around with this message at 17:17 on Jun 5, 2021 |
# ? Jun 5, 2021 17:08 |
|
GPU Megat[H]read - The Gee-forccé Arty-ex 3080 Tai-tee-ai-thai-tie-teyneeuhm-minum-mynee-mo is out
|
# ? Jun 5, 2021 17:19 |
|
|
# ? Jun 5, 2021 17:20 |
|
Mr.PayDay posted:https://www.pcmag.com/news/meet-the-people-who-camped-out-at-best-buy-for-nvidias-rtx-3080-ti-gpu Am I missing something? 2500 for the seat plus 1200 for the card is 3700. If you're willing to pay that much couldn't you just ebay a 3090 or something?
|
# ? Jun 5, 2021 17:59 |
|
there is a hell and we live in it
|
# ? Jun 5, 2021 18:02 |
|
VorpalFish posted:Am I missing something? 2500 for the seat plus 1200 for the card is 3700. If you're willing to pay that much couldn't you just ebay a 3090 or something? original receipt + direct purchased on card from authorized retailer grants some additional benefits. some of the higher tier amex/visa will probably blindly replace it if you mine it to death within a year or something
|
# ? Jun 5, 2021 18:59 |
|
Do miners use quadro cards for mining too? How are the prices of quadro cards compared to previous years?
|
# ? Jun 5, 2021 19:33 |
|
I haven't noticed Quadro prices going up significantly, I assume the price/performance of those for mining just isn't worth it.
|
# ? Jun 5, 2021 19:56 |
|
KinkyJohn posted:Do miners use quadro cards for mining too? How are the prices of quadro cards compared to previous years? There's no reason. The A6000 is like $5000 US and it's main draw in 48GB of VRAM and it's FP32/64 precision over the 3090 (someone that does ML correct me on this) No miner goes for 3090s due to the cost-to-hash ratio so they sure as hell wouldn't go for a Quadro
|
# ? Jun 5, 2021 19:58 |
I was lucky enough to get an EVGA RTX 3060 XC this week from their waitlist. Nice card and a decent upgrade over my old 1070. One word of caution is that the DisplayPort connectors are closer together. The cables I had been using for years on the 1070 had connectors that were 22mm wide. They wouldn't fit side-by-side on the 3060. Luckily I had a pair of cables with connectors that were 21mm wide and they work without issue. Turmoil fucked around with this message at 21:06 on Jun 6, 2021 |
|
# ? Jun 5, 2021 20:57 |
|
jisforjosh posted:There's no reason. The A6000 is like $5000 US and it's main draw in 48GB of VRAM and it's FP32/64 precision over the 3090 (someone that does ML correct me on this) not really a coder but i thought the big thing with machine learning is they don't want to use higher FP precision?
|
# ? Jun 5, 2021 21:14 |
|
the only real difference between the consumer cards and the workstation cards for ML besides the amount of memory is that the consumer cards have their mixed-precision performance nerfed at the driver level. they may also do that to the only-fp16 performance, i'm not sure. also, they say using consumer cards in data centers is against ToS so if you're a big business who wants to run a data center they won't give you customer support unless you buy the expensive ones. there's no such thing as a higher precision FP32 or FP64. a 32 bit floating point number is a standard format that is the same precision no matter what device it is on.
|
# ? Jun 5, 2021 23:26 |
|
Any San Diego goons wanna possibly form a group to camp Best Buy for the 3070ti launch? Thinking we could get something set up so people can take breaks or hold chairs/spots for people who can't be there all the time.
|
# ? Jun 5, 2021 23:36 |
|
taqueso posted:I'm trained on Ti being thai like titanium from bike parts back in the day before carbon I always think of titanal, from skiing.
|
# ? Jun 6, 2021 02:02 |
|
titanic, as in cannot possibly fail
|
# ? Jun 6, 2021 02:06 |
|
I'm still puzzled why they didn't brand the 3090 as a Titan and un-nerf stuff like FP16 performance - especially with the 3080 Ti sitting awkardly next to it. There were rumors for a while that there'd be a proper Titan but it looks like they're just pushing everyone to the A5000/6000s.
|
# ? Jun 6, 2021 02:11 |
|
drop a titan on it. lmfao
|
# ? Jun 6, 2021 02:16 |
|
the 3090 is like 10x too cheap to be not nerfed
|
# ? Jun 6, 2021 02:17 |
|
If people are actually trying to push saying X080ti as "tie" then I just see it as people pushing pronouncing GIF as "jif" - they are bad and dumb and wrong.
|
# ? Jun 6, 2021 02:40 |
CaptainSarcastic posted:If people are actually trying to push saying X080ti as "tie" then I just see it as people pushing pronouncing GIF as "jif" - they are bad and dumb and wrong. Agreed. It’s pronounced “Gee eye eff”.
|
|
# ? Jun 6, 2021 03:46 |
|
Crypto talk less enervating than ti talk
|
# ? Jun 6, 2021 03:53 |
|
Never trust an American on how to pronounce things.
|
# ? Jun 6, 2021 04:50 |
|
|
# ? Jun 10, 2024 11:29 |
|
Vintersorg posted:Never trust an American on how to pronounce things. That's probably fair, to be honest
|
# ? Jun 6, 2021 05:19 |