• 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.

*** The AMD UDNA Rumour Mill *** (next generation architecture)

Wouldn't it maker more sense to use GDDR7 instead (already in use in Nvidia's 5000 series).

Assuming 28 Gbps GDDR7, a 384 bit bus would result in = 1,344 GB/s.
 
Last edited:
eeii said:
Blimey 512bit bus. That’s not going to be cheap.
Click to expand...


Strange, AMD's whole thing with in recent times has been to use higher cache to offset the need for expensive vram configurations


Where as this leak suggests AMD is just going with the same vram configuration as a 5090
 
Last edited:
96 CU's would be 50% more than the 64 in the 9070 XT.

Vs the 9070 XT According to TPU at 4K.

5070 Ti: 104%
5080: 121%
4090: 134%
5090: 185%

This still isn't "High end" considering its going to compete with the 6000 series, tho it will be faster than the 4090, maybe significantly so, maybe closer to the 5090 than to the 4090, this is probably a true 6080 competitor.

Which IMO is fine, depending on price of course, how much faster do we need GPU's to be? And the 5090 has a die area of 750mm^2, that's massive, how much bigger will or can Nvidia make these GPU dies? And how much more than 600 Watts do they want to make them?

RX 1080 XT: 100% (New name, do it)
RTX 5090: 120%
RTX 6090: Who cares at this point?

Seriously this actually is fine....
 
Last edited:
I think with a new arch they should rename to these and have VRAM amounts similar to those below.

Easy to say, easy to remember and with a good choice of VRAM quantities for each gamers wants and needs.

____

RX 90 XTX (48GB) Enthusiast (If AMD are serious about doing this tier again).

RX 90 XT (32GB) High End
RX 90 (24GB)
RX 80 XT (24GB) Mid - High End
RX 80 (16GB)
RX 70 XT (16GB) Mid Range
RX 70 (16GB)
RX 60 XT (16GB) Low - Mid Range
RX 60 (12GB)
RX 50 XT (8GB) Low End
RX 50 (8GB)
____

Just my tuppence worth :cry:
 
MyBrainz said:
I think with a new arch they should rename to these and have VRAM amounts similar to those below.

Easy to say, easy to remember and with a good choice of VRAM quantities for each gamers wants and needs.

____

RX 90 XTX (48GB) Enthusiast (If AMD are serious about doing this tier again).

RX 90 XT (32GB) High End
RX 90 (24GB)
RX 80 XT (24GB) Mid - High End
RX 80 (16GB)
RX 70 XT (16GB) Mid Range
RX 70 (16GB)
RX 60 XT (16GB) Low - Mid Range
RX 60 (12GB)
RX 50 XT (8GB) Low End
RX 50 (8GB)
____

Just my tuppence worth :cry:
Click to expand...

I like that naming scheme...
 
humbug said:
96 CU's would be 50% more than the 64 in the 9070 XT.

Vs the 9070 XT According to TPU at 4K.

5070 Ti: 104%
5080: 121%
4090: 134%
5090: 185%

This still isn't "High end" considering its going to compete with the 6000 series, tho it will be faster than the 4090, maybe significantly so, maybe closer to the 5090 than to the 4090, this is probably a true 6080 competitor.

Which IMO is fine, depending on price of course, how much faster do we need GPU's to be? And the 5090 has a die area of 750mm^2, that's massive, how much bigger will or can Nvidia make these GPU dies? And how much more than 600 Watts do they want to make them?

RX 1080 XT: 100% (New name, do it)
RTX 5090: 120%
RTX 6090: Who cares at this point?

Seriously this actually is fine....
Click to expand...

Update to this, with the more than doubling of memory bandwidth the 50% increase CU count will scale 100%.

With that said rumours now also say there is a 10% IPC increase on RDNA 5 vs RDNA 4, so should be a 60% increase clock for clock.

9070 XT: 100%
5070 Ti: 104%
5080: 121%
4090: 134%
RDNA 5: 160%
5090: 185%

Or

RDNA 5: 100%
5090: 115%
 
Last edited:
Cerny and AMD discussed some upcoming technologies for future AMD GPU's and the PS6 console

In the video they discuss three features of next gen AMD GPU's that will also show up in the next gen PS6 console. That being:

* Radiance Cores: New cores for improved RT and PT performance
* Neural Arrays: New AI cores for AMD GPU's, to power future versions of FSR and Multi Frame Generation
* New Real Time Compression: To reduce the amount of memory games require
 
Last edited:
Grim5 said:
* New Real Time Compression: To reduce the amount of memory games require
Click to expand...
'requires speedy access to large amounts of memory'

They only discussed increasing ML cost as it evolves when more ML features are employed-which add memory cost- didn't see any specific mention of reducing games vram.
 
tommybhoy said:
'requires speedy access to large amounts of memory'

They only discussed increasing ML cost as it evolves when more ML features are employed-which add memory cost- didn't see any specific mention of reducing games vram.
Click to expand...

I remember back in the day how PS5's fast SSD and compression tech would allow games to be as big as never before, since that was the limiting part for all sorts of tricks to hide loading levels. :cry:
Turns out it was all BS.

To me, that whole talk is how they basically set the hardware up to be more efficient with RT/PT, catching up with PC advancements. After all is the direction the industry is going.
 
tommybhoy said:
'requires speedy access to large amounts of memory'

They only discussed increasing ML cost as it evolves when more ML features are employed-which add memory cost- didn't see any specific mention of reducing games vram.
Click to expand...
Minimising memory fab capacity that goes to consumer products makes sense, more of the fabs output goes to the high margin stuff.
 
TNA said:
Nexus was right all along, he foretold of this future.

What a time to be alive! :p:cry:
Click to expand...
Digital Foundry coined that native is defunct last year when John stated as such and that people should just forget about this whole bumming of "native or nothing" mentality that a weirdly large portion of PC gamers have for some reason. Look at the scene now, DLSS 4 produces as good as or better than "native" in many games, it's not just about upscaling, it's also about AA which DLSS does better than any other method currently available. DLAA works well too but doesn't have the performance of DLSS-SR.

Plus the fact you can use the latest SR model preset in any game that supports DLSS (or inject it into FSR only games) is another bonus.

It is funny though, now that consoles and AMD are on the wagon, it will be praised by various members of the online community and suddenly RT/PT are amazing to those eyes.
 
Last edited:
mrk said:
Digital Foundry coined that native is defunct last year when John stated as such and that people should just forget about this whole bumming of "native or nothing" mentality that a weirdly large portion of PC gamers have for some reason. Look at the scene now, DLSS 4 produces as good as or better than "native" in many games, it's not just about upscaling, it's also about AA which DLSS does better than any other method currently available. DLAA works well too but doesn't have the performance of DLSS-SR.

Plus the fact you can use the latest SR model preset in any game that supports DLSS (or inject it into FSR only games) is another bonus.

It is funny though, now that consoles and AMD are on the wagon, it will be praised by various members of the online community and suddenly RT/PT are amazing to those eyes.
Click to expand...

Yep. It was obvious this was coming. But quite a few people here were and still are native only users. Each to their own I guess.

I have been happy with DLSS since 2.0 and now we have FSR4 which is just as good if not better than that was so I don’t see why anyone would not se it if available.
 
You must log in or register to reply here.
Back
Top Bottom