Firstly, really not sure on the product naming but I guess is it so different from naming them after islands or after pivotal inventors in history? Well, yes…I don't think I couldn't not smirk every time I see “Druid generation” GPU. But maybe that's the publicity they're going for, weird naming and discussion about the weird naming in the foreground…
Secondly, they had to show it playing Crysis to put to bed “can it play Crysis”, ha ha!
I'm interested in intel GPUs in so far as it's a third player in the market but all opinion on their capabilities will be reserved minus my opinion on Mr Koduri and his activities from the Fury series and onwards at AMD.
Due to my need of cuda the only things that interest me in the intel gpu's are:
- Does it bring down prices on nvidia gpu's
- Do they have a cuda alternative that my software supports meaning I could consider something other than nvidia lol
Having said that if it's a viable alternative it might be worth it for a media/gaming pc if it's fairly priced etc.
Thats all nice and such, but what are the hash rates on these suckers.
LSG501
- Do they have a cuda alternative that my software supports…
The Xe-HPG GPUs will be supported by Intel's oneAPI back-ends for several libraries. They also provide tools for conversion of cuda code.
Looks like Codeplay is creating a cuda implementation of the oneapi backend, so, presumably, you'll also be able to accelerate Intel's libraries with your cuda GPUs.
There may be some serious bargains to be had when Intel releases DG2. After all they will need to create a market for their products. Sensible pricing would work.
Cuda commenters miss the existence of OpenCL (I run into this a lot with physics researchers who insist on coding with Cuda instead of portable libraries “because they don't understand it”)
Let's see if this breaks Nvidia's stranglehold on people's thinking
JayN
The Xe-HPG GPUs will be supported by Intel's oneAPI back-ends for several libraries. They also provide tools for conversion of cuda code.
Looks like Codeplay is creating a cuda implementation of the oneapi backend, so, presumably, you'll also be able to accelerate Intel's libraries with your cuda GPUs.
I know intel has their api but the issue is more about if the programs I use actually add the code for it, vray for example dropped AMD support even though AMD use opencl (not saying opencl is good in 3D rendering because it's not).
stoatwblr
Cuda commenters miss the existence of OpenCL (I run into this a lot with physics researchers who insist on coding with Cuda instead of portable libraries “because they don't understand it”)
Let's see if this breaks Nvidia's stranglehold on people's thinking
Actually I do know about opencl and to put it bluntly it's pretty useless in my use case of 3D rendering… it's considerably weaker option to use due to lack of ‘features’ that are on cuda these days.
You're also forgetting one MAJOR thing that Nvidia has going for it over AMD and intel…. it also supports opencl as well as cuda, albeit not as well performing, it may also support intels version too based on JayN's comment about codeplay.
LSG501
Due to my need of cuda the only things that interest me in the intel gpu's are:
- Does it bring down prices on nvidia gpu's
- Do they have a cuda alternative that my software supports meaning I could consider something other than nvidia lol
Having said that if it's a viable alternative it might be worth it for a media/gaming pc if it's fairly priced etc.
As long as Intel are competing for capacity at TSMC, I can't see this helping with the cost of the products we actually care about :(
While Nvidia was working on CUDA Intel's management was debating who is more important to get the blue badge; no hard feelings that
Nvidia AI tech is making its way into 5G base stations and Nvidia is an ever expanding giant that currently is simply UNSTOPPABLE.
Hmm…
Alchemist
Battlemage
Celestial
Druid
What are the odds their 5th gen will be “Elf”?
iworrall
Hmm…
Alchemist
Battlemage
Celestial
Druid
What are the odds their 5th gen will be “Elf”?
What annoys me is there is no consistency. Alchemist, Battlemage and Druid are professions whereas Celestial are beings.
*nerd rage*
If they're playing rift breaker then they might be playing with FSR already
Ray traced,upscaled BSODs then? ;)
Xlucine
If they're playing rift breaker then they might be playing with FSR already
I think Intel already confirmed they were looking into it for their Xe integrated graphics so would make sense if they're also looking at it for their dGPUs. While they're also designing their A
ISS systems as well.
Tabbykatze
While they're also designing their AISS systems as well.
Read that as ASS systems.
CAT-THE-FIFTH
Read that as ASS systems.
iworrall
Hmm…
Alchemist
Battlemage
Celestial
Druid
What are the odds their 5th gen will be “Elf”?
My money is on “Thief”…
Taiwan's UDN has talked to an insider who says that Intel Arc GPUs will be produced by TSMC on its N6 process.
Link via
Dan Nystedt on Twitter.
mtyson
Taiwan's UDN has talked to an insider who says that Intel Arc GPUs will be produced by TSMC on its N6 process.
Link via Dan Nystedt on Twitter.
Interesting, I wonder if this will mean silicon production lines will be disparate from the N7/N7+ lines meaning we won't just be watering down the silicon supply and therefore keeping the same amount of product available but from 3 vendors instead of 2.
Tabbykatze
Interesting, I wonder if this will mean silicon production lines will be disparate from the N7/N7+ lines meaning we won't just be watering down the silicon supply and therefore keeping the same amount of product available but from 3 vendors instead of 2.
I don't think so.
TSMC are all about low risk, and I though 7nm to 6nm was basically very similar but 6nm uses EUV steps (or more EUV steps) to decrease process time.
I think for 6nm TSMC started a new line to do 6nm, they then encourage their customers to move to 6nm (same design rules, not the same masks: but better characteristics and possible less process steps). Then as 7nm demand trails off, I would expect them to retool some 7nm lines as 6nm.