-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
AMD Navi 23 ‘NVIDIA Killer’ GPU Rumored to Support Hardware Ray Tracing, Coming Next Year
- Thread starter Gregster
- Start date
- Status
Associate
- Joined
- 31 Mar 2016
- Posts
- 1,412
- Location
- Moonbase Alpha
Damn OC forums have been down for hours, presumably cause of the RTX 3080 launch mess
Nvidia cannot be held responsible for the functionality of the OC site.
Associate
- Joined
- 31 Mar 2016
- Posts
- 1,412
- Location
- Moonbase Alpha
I've heard about it but I for one have no idea of the implications of it.
Soldato
- Joined
- 21 Jul 2005
- Posts
- 21,201
- Location
- Officially least sunny location -Ronskistats
I went on another site (as its up) and after pretending to buy a 3080, it said 'likely in stock - 30th October'. Jesus thats quite a while lol.
I'll just put this here...
I've edited this back in - No need for a separate thread - Armageus
These's some rumours going around about a 128MB Infinity Cache.
I'd say its very likely to be correct, a patent was published today from AMD called 'ADAPTIVE CACHE RECONFIGURATION VIA CLUSTERING', link here here:
https://www.freepatentsonline.com/y2020/0293445.html
Basically, what I think it will allow, is greater L1 cache capacity for GPUs. It does this by dynamically creating 'a plurality of compute units' clusters. The dynamically created clusters allow a 'backup' pool of shared L1 cache to be accessed, if the L1 cache for a CU (Compute Unit) isn't sufficient.
L1 cache is the fastest (lowest latency) type of GPU cache. So the more L1 capacity, the less the slower L2 and L3 caches are used.
My thoughts - The 5700 XT uses 128KB L1 Cache per Graphics Engine (4 in total). Maybe L1 cache sizes would also need to be increased? The RTX 3080 has a total of 8.7MB of L1 cache (128 KB per SM), quite a bit higher.
----------------------------------------------------------------------------------------------------------
More general rumours here:
https://www.notebookcheck.net/AMD-R...-PlayStation-5-GPU-clock-speeds.494558.0.html
I've edited this back in - No need for a separate thread - Armageus
These's some rumours going around about a 128MB Infinity Cache.
I'd say its very likely to be correct, a patent was published today from AMD called 'ADAPTIVE CACHE RECONFIGURATION VIA CLUSTERING', link here here:
https://www.freepatentsonline.com/y2020/0293445.html
Basically, what I think it will allow, is greater L1 cache capacity for GPUs. It does this by dynamically creating 'a plurality of compute units' clusters. The dynamically created clusters allow a 'backup' pool of shared L1 cache to be accessed, if the L1 cache for a CU (Compute Unit) isn't sufficient.
L1 cache is the fastest (lowest latency) type of GPU cache. So the more L1 capacity, the less the slower L2 and L3 caches are used.
My thoughts - The 5700 XT uses 128KB L1 Cache per Graphics Engine (4 in total). Maybe L1 cache sizes would also need to be increased? The RTX 3080 has a total of 8.7MB of L1 cache (128 KB per SM), quite a bit higher.
----------------------------------------------------------------------------------------------------------
More general rumours here:
https://www.notebookcheck.net/AMD-R...-PlayStation-5-GPU-clock-speeds.494558.0.html
Last edited by a moderator:
Soldato
- Joined
- 14 Aug 2009
- Posts
- 3,554
Well, maybe we'll see the day AMD will actually do something about that.
Soldato
- Joined
- 21 Jul 2005
- Posts
- 21,201
- Location
- Officially least sunny location -Ronskistats
I think its warming up nicely @g67575 too many rumours now for something that is going to address the regular DDR6 issue. I am also interested in this other feature that can spread the load between Ryzen CPU and RDNA2 graphics as if this is a thing too it could be nifty considering all the console targeted games.
About their drivers? They've been good for years now, you get the odd buggy set but that's par for the course for either of the 2.
Soldato
- Joined
- 21 Jul 2005
- Posts
- 21,201
- Location
- Officially least sunny location -Ronskistats
Yeah thats an old wives tale for sure, if its putting you off and you only buy nvidia sure you will never know.
Well after saying earlier in the week about not being sure whether to wait for AMD to release before jumping on the GPU upgrade wagon, the shambles of the nvidia 'paper' launch and the blatant and disgusting price gouging by retailers has made my mind up to wait it out.
I just hope AMDs launch has a better start and doesn't decend into the chaos that nvidias has.
I just hope AMDs launch has a better start and doesn't decend into the chaos that nvidias has.
Soldato
- Joined
- 21 Jul 2005
- Posts
- 21,201
- Location
- Officially least sunny location -Ronskistats
They are going to reveal nvidias headstone. Rdna3 in 2021, MCM on TSMC nm will bury NVIDIA.
While it doesn't absolutely require it - to make the best use of MCM will require some significant OS, driver and hardware changes and we'd likely see more Windows updates along the lines of the rollout of hardware accelerated scheduling well in advance of MCM GPUs come along.
EDIT: Not sure if it was just prototyping but apparently one of the approaches that has been played with is almost the opposite of what some people talk about - a way of having multiple chips that can be allocated to multiple virtual GPUs which can have their resource allocation dynamically adjusted on the fly - which can be farmed out to indiscriminately from the software/application side but adjusted at a lower level depending on load on each one.
Last edited:
Yea they have great products coming
This gen Big Dog Navis is the first stone thrown or set.
They beaten Intel and now nvidia is feeling the pressure.
Last edited:
Machines 1 : Humanity 0
Im laughing that many people think Nvidia dont have poor supply of their GPUs, this was pretty much a paper launch to get out ahead of what they know is coming.
People are blaming the retailers for gouging, it was always going to happen with the limited stock supply.
Reports of over 80% yield for the cpu for the new Xbox Series S, sounds promising, i do expect however the GPU's to be limited supply for a few months especially with Cpus coming as well, but nothing like Nvidias issues.
I think we may see a turning point for the Radeon team with this launch, i cant wait, will trade in my 5700XT for big Navi
People are blaming the retailers for gouging, it was always going to happen with the limited stock supply.
Reports of over 80% yield for the cpu for the new Xbox Series S, sounds promising, i do expect however the GPU's to be limited supply for a few months especially with Cpus coming as well, but nothing like Nvidias issues.
I think we may see a turning point for the Radeon team with this launch, i cant wait, will trade in my 5700XT for big Navi
Caporegime
- Joined
- 13 May 2003
- Posts
- 34,643
- Location
- Warwickshire
Rogame et al on Twitter seem pretty convinced that Navi 21 is 16GB of GDDR6 on a 256 bit bus and Navi 22 is 12GB of GDDR6.
I think we can expect Navi 21 to be flirting with 3080 performance levels at around $100 cheaper with 50W less power required.
I think we can expect Navi 21 to be flirting with 3080 performance levels at around $100 cheaper with 50W less power required.
- Status