AMD Radeon RX 5800 And RX 5900 Navi GPUs Documented In Trademark Filings

[exetras]

https://hothardware.com/news/sapphire-amd-radeon-rx-5800-5900-navi-navi-20

Quote

When AMD first announced its initial round of first 7nm Navi 10-based graphics cards, they were aimed squarely at the mainstream segment of the gaming market to enable 1440p experiences. Those cards include theRadeon RX 5700, Radeon 5700 XT and Radeon 5700 XT 50^th Anniversary Edition; all of which will launch on July 7^th.

 

We’re now getting some potential further confirmation of more powerful variants of Navi that could be released in the months to come. Sapphire has registered a whole slew of product names with the EEC for upcoming graphics cards including the Radeon RX 5500 at the low end and a Radeon RX 5950 XT at the high-end. The already announced Radeon RX 5700 series sits squarely in the middle.

Here is the leaked filling:

 

As Always:

Quote

With that being said, we’ll have to take the above product names with a grain of salt at this time, as it’s very likely that a number of the entries are simply placeholders at this moment according to VideoCardz.

 

This answers the question, Are we getting more variations of the Vega architecture? Yes we are.

 

My guess is the R5900 will be aimed at the 2080 TI and 4K gaming, while the 5500 would be aimed at 1080P gaming.

 

These leaks are fun!

[Doobeedoo]

Now we're talking, some new high end Navi chips. 

[Master Disaster]

So Navi 21 then? Also based on historical naming convention 5950 will be dual GPU on a single board which is both exciting and scary in equal measure.

[RejZoR]

It doesn't necessarily mean these parts are actually planned. Filling a trademark just means protecting the name, so NVIDIA or Intel can't call their products RX 5900XT for whatever reason. Though it is likely for these parts to appear given that AMD has extended plans for RDNA and models after RX 5700 series.

[SenKa]

This is a total pipe dream, but with Vega II coming out maybe dual GPU's will be back in style for the mainstream???

 

It won't happen, and I know it won't happen. But I still want it to happen.

[SenKa]

Just now, Master Disaster said:

So Navi 21 then? Also based on historical naming convention 5950 will be dual GPU on a single board which is both exciting and scary in equal measure.

My exact thought! I hope AMD followed suit with their naming scheme

[Master Disaster]

Just now, SenpaiKaplan said:

This is a total pipe dream, but with Vega II coming out maybe dual GPU's will be back in style for the mainstream???

 

It won't happen, and I know it won't happen. But I still want it to happen.

5950 kind of hints this will be the case

[SenKa]

1 minute ago, Master Disaster said:

5950 kind of hints this will be the case

I think I may put off upgrading my GTX670 for a bit. It runs all my game flawlessly ATM and I don't see that changing any time soon... 5950 is making me think happy thoughts.

[Humbug]

1 hour ago, SenpaiKaplan said:

This is a total pipe dream, but with Vega II coming out maybe dual GPU's will be back in style for the mainstream???

 

It won't happen, and I know it won't happen. But I still want it to happen.

Personally much more looking forward to big Navi rather than dual GPUs. The RDNA architecture is showing great potential with cards having the same number of compute units as Polaris now performing like Vega 64. Which was never remotely possible on gcn, even if they had hit these clockspeeds. So once they make big GPUs with 60+ compute units it's gonna be exciting.

 

On the other hand I find it hard to look forward to dual GPUs bevause we are relying on Nvidia and AMD creating SLI/crossfire drover profiles and game devs for whom this is the lowest priority.

[SenKa]

2 hours ago, Humbug said:

Personally much more looking forward to big Navi rather than dual GPUs. The RDNA architecture is showing great potential with cards having the same number of compute units as Polaris now performing like Vega 64. Which was never remotely possible on gcn, even if they had hit these clockspeeds. So once they make big GPUs with 60+ compute units it's gonna be exciting.

 

On the other hand I find it hard to look forward to dual GPUs bevause we are relying on Nvidia and AMD creating SLI/crossfire drover profiles and game devs for whom this is the lowest priority.

There are a lot of games that still scale amazingly well with Crossfire/SLI. They are increasingly rare, sadly, but I think that is simply because less people are buying 2 or more GPU's due to cost.

 

I would like to think AMD Is revisiting dual GPU cards for the consumer/prosumer market, but at the same time I highly doubt it. A single Radeon Pro Duo (Polaris) costs as more than my whole system.

[cj09beira]

58 minutes ago, SenpaiKaplan said:

There are a lot of games that still scale amazingly well with Crossfire/SLI. They are increasingly rare, sadly, but I think that is simply because less people are buying 2 or more GPU's due to cost.

 

I would like to think AMD Is revisiting dual GPU cards for the consumer/prosumer market, but at the same time I highly doubt it. A single Radeon Pro Duo (Polaris) costs as more than my whole system.

they seem to be going on that direction but with enterprise market first as they know that devs are too lazy to support it if it still shows on the driver as 2 cards 

[pas008]

43 minutes ago, SenpaiKaplan said:

There are a lot of games that still scale amazingly well with Crossfire/SLI. They are increasingly rare, sadly, but I think that is simply because less people are buying 2 or more GPU's due to cost.

 

I would like to think AMD Is revisiting dual GPU cards for the consumer/prosumer market, but at the same time I highly doubt it. A single Radeon Pro Duo (Polaris) costs as more than my whole system.

yes its work for multiple companies now

 

I have seen the intel quad chip gpu rumors

 

I have always wondered why they couldnt do this once they figured out the latency

hydra 100 and 200 chips were around for awhile before where they did work load dividing i guess if i remember correctly

but why not something like this on the gpu itself to divide the resolution work into quadrants or even halves

like 4k is 4x1080p displays a chip for each quadrant on the gpu

1440 is 4x720p etc etc

[cj09beira]

2 minutes ago, pas008 said:

yes its work for multiple companies now

 

I have seen the intel quad chip gpu rumors

 

I have always wondered why they couldnt do this once they figured out the latency

hydra 100 and 200 chips were around for awhile before where they did work load dividing i guess if i remember correctly

but why not something like this on the gpu itself to divide the resolution work into quadrants or even halves

like 4k is 4x1080p displays a chip for each quadrant on the gpu

1440 is 4x720p etc etc

the problem there is things like temporal anti aliasing which needs pixels near it so the boundaries would be messed up (if there was 0 cross talk), so at most they can do most of the work but things like post processing has to be on a single chip unless the communication is really fast

[pas008]

10 minutes ago, cj09beira said:

the problem there is things like temporal anti aliasing which needs pixels near it so the boundaries would be messed up (if there was 0 cross talk), so at most they can do most of the work but things like post processing has to be on a single chip unless the communication is really fast

taa even has its problems by far on sli but some cases/drivers/games/etc it works though

which shows it can be done and would be up to the amd/nvidia/intel to implement it properly because all would be on the card

[Belgarathian]

Only 5950 XT... Oh well, can't wait for the 5970 and 5990 XTX

[grayperview]

I'm guessing the 5800's will have 56 CU's and the 5900's will have 60 and 5950 64, all binned to run at the limit.

[Trik'Stari]

What's this? AMD decided to maybe start actually fucking competing in a way that matters? (I mean in the GPU industry)

 

THANK. FUCK

[CarlBar]

6 hours ago, cj09beira said:

has to be on a single chip unless the communication is really fast

 

Well with AMD pushing their IF interconnects as a solution to everything it's quite possibble actually. The EPYC 2 info released indicates a single Zen 2 CPU chiplet supports upto 410Gbytes/second of transfer speed so they could do it. And from the PoV of the datacenter it may make sense to stat designing things to be able to do that, in which case backporting to desktop wouldn't be a completely terrible decision. If they can make it work it's effectively chiplets for GPU's. And that would be a game changer.

[The King of the Undead]

15 hours ago, CarlBar said:

 

Well with AMD pushing their IF interconnects as a solution to everything it's quite possibble actually. The EPYC 2 info released indicates a single Zen 2 CPU chiplet supports upto 410Gbytes/second of transfer speed so they could do it. And from the PoV of the datacenter it may make sense to stat designing things to be able to do that, in which case backporting to desktop wouldn't be a completely terrible decision. If they can make it work it's effectively chiplets for GPU's. And that would be a game changer.

what the hell????????

that's a little faster than my sata 2 or my pcie 1 x16 ever will be.