Info Big Navi - Radeon 5950 XT specs leak.

[Hans de Vries]

Seems this might be a valid leak from inside partner SK Hynix..

Twitter

Big Navi - Radeon 5950XT: Twice the compute units as Navi 10.

Shading units: 5120
TMUs: 320
Compute units: 80
ROPs: 96
L2 cache: 12MB
Memory: 24GB (4 x HBM2e, 3 die)
Memory bus: 4096 bits
Band Width: 2048 GB/s

All these Big Navi numbers are perfectly consistent. The 96 ROPs are tiny pieces of logic at the edge of the memory tiles of the 12MB L2 cache which explanes the factor 3 in these numbers. An old example here https://bjorn3d.com/2010/01/nvidia-gf100-fermi-gpu/

SK Hynix will make 6GB HBM2e stacks with 3 dies per stack for consumer applications on request. SK Hynix recently announced HBM2e at 512GB/s at ISSCC-2020. Samsung went a step further with 640GB/s HBM2e (5 Gb/s/pin)

 

[DisEnchantment]

They also don't mention Zen+ directly (only as 12nm Zen, nothing about its enhanced IPC/cache) - don't read into it too much.

More important is that they announced Zen4 as 5nm, but did not do the same for RDNA3, makes me think either RDNA3 is delayed, or it's far enough out that they are considering waiting for the next process, either N5P or N3.

They announced Zen 4 only in the context of Data Center, so that enterprise customers get more visibility. All Consumer products don't have visibility beyond the next generation.

RDNA2 is influenced a lot by console timelines. Similarly RDNA3 is also influenced a lot by Samsung's Exynos plans. They cannot delay RDNA3, they need to match Samsungs Exynos 11 series.
There is a second team working on RDNA3 just like the Zen leapfrogging teams.
Expect more aggressive perf/watt improvements from RDNA3. That being said though I wonder if the RT HW will be present in the IP for Samsung, although I reckoned Samsung can access it if they desire. Makes me wonder though where does it fit in AMD's roadmap for PC.

"The opportunity was there and they took it"

 

[soresu]

Expect more aggressive perf/watt improvements from RDNA3. That being said though I wonder if the RT HW will be present in the IP for Samsung, although I reckoned Samsung can access it if they desire. Makes me wonder though where does it fit in AMD's roadmap for PC.

RT will be invaluable for standalone VR once Samsung really get back to competing in that arena.

 

[zinfamous]

Use 1080p/60hz and set Framerate Target to 60fps with Auto Voltage on, and you have achieved your goal.

peasant resolutions? NO SIR!

 

[Guru]

I think we are going to see a whole new generation, which will consist of the "big" navi, as well as trickle down mid range cards.

They are using the same process node, and since they are launching late 2020, they would have had plenty of time to optimize RDNA.

 

[EXCellR8]

Even more important than node design...

AMD promises open-air cooler designs for next-generation Radeon RX

AMD says there will be no reference blower NAVI 2X graphics cards designs - VideoCardz.com

AMD promises open-air cooler designs for next-generation Radeon RX Shortly after AMD’s Financial Analyst Day webcast ended, Scott Herkelman, the Vice President and General Manager for Radeon, has confirmed on Reddit that upcoming reference Radeon RX graphics cards will not use blower design...

videocardz.com

This may have just sold me on 2X

 

[extide]

They are using the same process node, and since they are launching late 2020, they would have had plenty of time to optimize RDNA.

Same process node as what? They said it won't be N7, that it would be better, but no details beyond that. So you can't really say what process they are using for sure yet.

 

[Guru]

Same process node as what? They said it won't be N7, that it would be better, but no details beyond that. So you can't really say what process they are using for sure yet.

RX 5000 series actually uses N7p process node(which is the improved n7 node), which will be the same exact process node RDNA2 uses. TSMC actually has 7+ mode, which is actually 6nm and offers up to 15% performance improvement. The n7p offers very small performance improvement over the original 7nm process, but it does offer a lot more density over the first 7nm.

Its safe to say that someone like Apple probably "Snatched up" the 6nm node and with TSMC supply running very short, AMD is going to be using 7nm second gen process.

 

[exquisitechar]

RX 5000 series actually uses N7p process node(which is the improved n7 node), which will be the same exact process node RDNA2 uses. TSMC actually has 7+ mode, which is actually 6nm and offers up to 15% performance improvement. The n7p offers very small performance improvement over the original 7nm process, but it does offer a lot more density over the first 7nm.

Its safe to say that someone like Apple probably "Snatched up" the 6nm node and with TSMC supply running very short, AMD is going to be using 7nm second gen process.

No. N7P brings no density increase and 7+ and 6nm aren’t the same thing.

 

[GodisanAtheist]

I'm salivating for new card releases. 2020 is going to be a good year...

 

[psolord]

I'm salivating for new card releases. 2020 is going to be a good year...

Me too.

I might even take the 4k step this year.

 

[Tabalan]

They announced Zen 4 only in the context of Data Center, so that enterprise customers get more visibility. All Consumer products don't have visibility beyond the next generation.

RDNA2 is influenced a lot by console timelines. Similarly RDNA3 is also influenced a lot by Samsung's Exynos plans. They cannot delay RDNA3, they need to match Samsungs Exynos 11 series.
There is a second team working on RDNA3 just like the Zen leapfrogging teams.
Expect more aggressive perf/watt improvements from RDNA3. That being said though I wonder if the RT HW will be present in the IP for Samsung, although I reckoned Samsung can access it if they desire. Makes me wonder though where does it fit in AMD's roadmap for PC.

"The opportunity was there and they took it"

Why would RDNA3 we heavily influenced by Samsung Exynos? If Samsung wants to change RDNA to their needs, they can do it on their own or ask AMD for help. There is no need to change RDNA3 to fit Samsung needs.

 

[soresu]

Why would RDNA3 we heavily influenced by Samsung Exynos? If Samsung wants to change RDNA to their needs, they can do it on their own or ask AMD for help. There is no need to change RDNA3 to fit Samsung needs.

I'm inclined to think Samsung a separate operator anyway - any IP they use will be on their own processes besides what AMD use for their own products.

It's likely that it will be closer to the arrangement with Hygon/Zen than the semi-custom deals with Sony and Microsoft, giving Samsung more leeway to design their own SoC's while having access to RDNA's successive IP generations from AMD.

 

[DisEnchantment]

Dang, Charlie was brutal against RTX in his latest article

Then we come to the last but most interesting bits, raytracing and variable rate sharing which are finally found in AMD’s RDNA2 architecture due this year. Nvidia has been making a huge stink about this, claiming superiority, being first to market and the like. Imagination beat them to market by years and the software support for Nvidia’s raytracing efforts has largely been relegated to those willing to accept large amounts of MDF, it hasn’t managed to stand on it’s own much less flourish without large cash infusions. Intel released variable rate shading last year with Ice lake, something also found in RDNA2. This one does have legs, is useful, and will stand on it’s own.

If you have any doubts, Sony also uses RDNA2/Navi 2X in the PS5 but with a slightly different software stack. Nvidia uses their own proprietary API that has all the lock-in that developers and customers hate, plus it isn’t compatible with either console. In short Nvidia released first in the vain hopes of a first mover advantage but it was folly. Game over, AMD wins for raytracing. When the software for the next gen consoles starts hitting the market, one solution will be there, the other will be a cash sink.

AMD talks direction and tech at Financial Analysts Day

AMD had their Financial Analyst Day yesterday and SemiAccurate found a lot of goodies.

www.semiaccurate.com

 

[extide]

RX 5000 series actually uses N7p process node(which is the improved n7 node), which will be the same exact process node RDNA2 uses. TSMC actually has 7+ mode, which is actually 6nm and offers up to 15% performance improvement. The n7p offers very small performance improvement over the original 7nm process, but it does offer a lot more density over the first 7nm.

Its safe to say that someone like Apple probably "Snatched up" the 6nm node and with TSMC supply running very short, AMD is going to be using 7nm second gen process.

You should try to use the actual node names -- and as mentioned N7+ is not N6. I have heard rumors that Navi is N7P -- but never Confirmed by AMD.

 

[Mopetar]

Dang, Charlie was brutal against RTX in his latest article

AMD talks direction and tech at Financial Analysts Day

AMD had their Financial Analyst Day yesterday and SemiAccurate found a lot of goodies.

www.semiaccurate.com

Charlie has always had a bit of an axe to grind with NVidia, but he’s usually pretty blunt in general.

It’s a fair point he makes, but I think he’s built this false dichotomy where either AMD or NVidia succeed. I think the more likely outcome is that RT is another flash in the pan console gimmick and that it doesn’t see very wide or pervasive use.

We’ve already seen that you basically need something at 2080 Ti levels to have a good frame rate without sacrificing resolution. I doubt AMD has some kind of magically superior implementation that lets them squeeze that kind of performance into a console GPU.

It’s going to take another five years at least for acceptable performance to reach mainstream GPUs which means that most consumers won’t use RT at all and that any developer effort is spent for a small minority of gamers. Console ports are only going to use it for minimal effects in specific instances so there won’t be a lot of developer effort spent extending it beyond that level either.

 

[mopardude87]

I'm salivating for new card releases. 2020 is going to be a good year...

Yes i keep preaching this to any who can hear. I haven't nerded out so much about a year since 2007. Not sure what excites me more though, the concept of nice AMD drivers, a killer 5950xt or when big Ampere comes round and puts the 5950xt possibly in its place? Gonna be a intense year for sure and yeah on the cpus even better.

 

[GodisanAtheist]

Yes i keep preaching this to any who can hear. I haven't nerded out so much about a year since 2007. Not sure what excites me more though, the concept of nice AMD drivers, a killer 5950xt or when big Ampere comes round and puts the 5950xt possibly in its place? Gonna be a intense year for sure and yeah on the cpus even better.

- Absolutely.

AMD seems more confident in its GPU stack than I've seen it in a long time, we know NV is going to drop some goodness at some point here, node changes, arch changes, RTRT from all manufacturers so now we have another variable to discuss and argue over, compute card lines really branching off from consumer focused lines, no more or greatly reduced ****ing bit mining to artificially distort prices (not that anyone will need an excuse...).

Lot of vague industry trends starting to solidify and come together all at once in 2020. And that's just GPUs.

Good time to be a GPU/DIY geek.

 

[Guru]

You should try to use the actual node names -- and as mentioned N7+ is not N6. I have heard rumors that Navi is N7P -- but never Confirmed by AMD.

Seems like TSMC has changed its naming scheme recently, so it seems there is now n7, n7p, n7+ and n6, seems like n6 uses the same packaging and tools as n7 and is compatible with n7 designs, but essentially offers very similar improvements over n7 as n7+ does.

 

[amrnuke]

Seems like TSMC has changed its naming scheme recently, so it seems there is now n7, n7p, n7+ and n6, seems like n6 uses the same packaging and tools as n7 and is compatible with n7 designs, but essentially offers very similar improvements over n7 as n7+ does.

N7 branches (on TSMC's roadmap) to N7+ -> N5 -> N3 (a live branch) as well as to N7P -> N6 which is dead-end branch but compatible with N7.

It'll be interesting to see where AMD take Zen3* and RDNA2 since N7+ sees 30-50% better power/performance gains than N7P.

 

[exquisitechar]

Charlie has always had a bit of an axe to grind with NVidia, but he’s usually pretty blunt in general.

It’s a fair point he makes, but I think he’s built this false dichotomy where either AMD or NVidia succeed. I think the more likely outcome is that RT is another flash in the pan console gimmick and that it doesn’t see very wide or pervasive use.

We’ve already seen that you basically need something at 2080 Ti levels to have a good frame rate without sacrificing resolution. I doubt AMD has some kind of magically superior implementation that lets them squeeze that kind of performance into a console GPU.

It’s going to take another five years at least for acceptable performance to reach mainstream GPUs which means that most consumers won’t use RT at all and that any developer effort is spent for a small minority of gamers. Console ports are only going to use it for minimal effects in specific instances so there won’t be a lot of developer effort spent extending it beyond that level either.

Consoles having 2080ti RT performance is pretty likely, really. I mean, with the Series X 12 TFLOPS RDNA2 GPU, the overall performance will probably be in the range of the 2080 Super, and I believe that RDNA2 will have significantly better RT performance than Turing (Ampere will too).

 

[mopardude87]

- Absolutely.

AMD seems more confident in its GPU stack than I've seen it in a long time,

Good time to be a GPU/DIY geek.

Yeah,i haven't been excited about a Amd gpu since the 7970 and i loved mine. I loved its gaming and overclocking ability, 1200 core was easy on the reference but the drivers had me returning it. I got faith in this big Navi even if somehow its just a short victory.

Shame i only see Cyberpunk as a reason i may upgrade 7700k/16gb/1080ti .Being at 1080p well i prob won't need a single upgrade anyways till Cyber. Worst case i build something next year and i go ahead and allocate funds into a new project dart/duster perhaps and that will keep my mind and myself busy.

 

[fleshconsumed]

Do I dare to get my hopes up that AMD is going to bring same level of excitement to GPU segment as they did for CPU in 2017? I'd love for that to happen, but I've been hurt badly by Fury/Vega/Navi... I'm not sure I can handle another disappointment...

 

[exquisitechar]

Do I dare to get my hopes up that AMD is going to bring same level of excitement to GPU segment as they did for CPU in 2017? I'd love for that to happen, but I've been hurt badly by Fury/Vega/Navi... I'm not sure I can handle another disappointment...

Eh, they’ll bring some serious performance. The time of flops like Vega is over. A Zen-like increase in performance/$ isn’t happening, though.

 

[amrnuke]

Eh, they’ll bring some serious performance. The time of flops like Vega is over. A Zen-like increase in performance/$ isn’t happening, though.

From what I understand, multi-GPU support may allow such an increase.

If everyone got together and pushed dev/app support for the Vulkan/DX12 parallelism, it would probably create a Zen-like increase in performance/$ across the board for AMD, Nvidia, and Intel. But I doubt any of them have a vested interest in that, due to the non-linear scaling of GPU cost as performance increases. Once the cat's out of the bag with multiGPU support, there's nothing stopping folks from sticking two $280 5700s in a system instead of buying, say, a $999 5950XT. Same applies for Nvidia's lineup. The GPU manufacturers, however, could choose to steer developers toward multi-GPU on-die support, but disable for support of multiple cards, hence allowing them to slap several chiplets on one piece of silicon, and still being able to price as they wish without consumer competition from people buying two cheaper cards for similar performance.

Though I'm not sure I have a great grasp on this, it seems like all the tools are there to make the leap, it's just that I am not sure AMD, Nvidia, and Intel see any major benefit to it.

 

[Pohemi]

From what I understand, multi-GPU support may allow such an increase.

Multi-GPU support (from both AMD and Nvidia) has been waning for years. The driver stacks don't even officially support it any longer afaik, and devs are not coding for it.