NVIDIA GeForce RTX 50 Series Faces Compute Performance Issues Due to Dropped 32-bit Support

[LastDudeALive]

Honestly, if the only lasting impact of the 50 series is forcing software devs to finally get off their asses and update all 32-bit code still in common programs, it will have been a good generation.

 

[_roman_]

Please no ngreedia posts. The topic is about Passmark / Nvidia Software Stack and nvidia graphic cars / maybe windows operating systems.

We should learn from that topic. Passmark is not really a decent benchmark software. Someone there is unable to read the release notes. They do not have proper coders or enough coders to keep their benchmarks up to date.

If you do not know which libaries you use in a project - you have an issue.

32bit is dead in my point of view for personal computers.

 

[bug]

Another clickbait, inflammatory article (from the same editor).

NVIDIA has silently discontinued support for 32-bit OpenCL and CUDA

There is no longer a supported 32bit Windows version, Nvidia's drivers are currently 64bit only. The change is also noted in the CUDA release notes (but you have to look). What were you expecting, paid TV ads?

This architectural change wasn't clearly documented by NVIDIA, whose developer website continues to display 32-bit code samples and documentation despite the removal of actual support.

Because, lo an behold, legacy software is still out there and needs maintenance.

Everything that worked before, still works. Just not with newer drivers. It's really nothing new, this is how Nvidia has always handled deprecation: they park stuff in a legacy branch, the branch still receives security and other fixes for years to come. Just not as often.

Edit: If anything, the news is that software that pretends to be 64bit, is still 32bit under the hood.

 

[igormp]

One of the reasons for dropping support for something in a compiler is that the hardware no longer supports it, without knowing what's actually happening under the hood it's hard to tell why they did this.

AFAIK that has no direct relation with the underlying hardware. Nvidia has deprecated building 32-bit stuff with their compiler ages ago, and now they've finally dropped any way to build such software with their latests SDK, and to run it with their newest ISA.
It may just be trimming down their stack, nothing more complicated than that.

GPUs work in a regime where almost everything is compiled first before running anyway, it's very bizarre to drop support for something unless there was a technical reason to do so. It's one thing to drop support for development and a different thing to remove the ability to run certain software entirely.

We are not specifically talking about the GPU itself here. If you try to write CUDA code in any device, even a windows QCOM laptop, and try to build it with the latest NVCC, it won't cross compile 32-bit software at all. Keep in mind that this "compiled" code usually turns into PTX, which is Nvidia's IR (similar to LLVM's IR), which is not directly executed by the GPU.
If you have software that was built with a previous version, so that 32-bit was still working, the runtime won't execute it if you're targeting to run it with Blackwell. Now I may be wrong, but AFAIK this is a limitation within the runtime SDK even before it gets to do any call upon the GPU, since the runtime is the part responsible for turning the above PTX into actual code that the GPU can run, so it's mostly a total lack of support from the software stack itself.

 

[Vayra86]

Come on AMD. The door is opened wide for you this time

Take it easy, AMD's been very busy limping towards the doorway

 

[Vya Domus]

but AFAIK this is a limitation within the runtime SDK even before it gets to do any call upon the GPU, since the runtime is the part responsible for turning the above PTX into actual code that the GPU can run, so it's mostly a total lack of support from the software stack itself.

This is the case with everything, every new architecture has a new ISA and PTX needs to be turned into a compatible binary using an updated compiler, the compiler has of course been updated as usual except it wont emit binaries for 32bit PTX. This tells you nothing about the reason as to why it wont do that, logically the hardware may not be suited to run 32 bit binaries, it's the same when you try and run a CUDA application on an older GPU which is shipped with a compiler that can't turn the given PTX into machine code.

Reminder that everything has to get turned into a compatible binary, even a regular shader, so why is there a compiler that can turn regular shaders in 32bit machine code (presumably, maybe it promotes everything to 64bit but I doubt it) but can't do the same for 32bit PTX ? Perhaps some incompatible instructions may be issued through 32bit PTX which can't be issued through shaders, hence lack of support, rather than "they couldn't be bothered" which seems implausible.

I mean nothing really encourages development of 32bit applications in general these days, why not drop support for literally everything 32bit running on the GPU, why just 32bit PTX ?

 

[thepath]

Probably the current implementation of passmark depicts a more real situation of the performance the user will obtain from the product.

Things must move on? Sure. This way? At all.

Imagine intel or AMD dropping suport of some 32bit instructions or sets without prior warning... any warning.

Intel and AMD GPU can't run hardware PhysX....They have same problem as RTX 50 if you enable PhysX in those old games.

 

[Dr. Dro]

This is more of a "Passmark" problem than an "Nvidia" problem, if you ask me. Not that I expect consumers to know, but as @igormp brought up, it's been deprecated since CUDA 9, released about 8 years ago. People just took it for granted it'd always be there, dropping an ISA is rough, there's no optimal schedule for this, because there's no scenario where everyone is ready for it.

Passmark should update their benchmark suite to be fully optimized to 64-bit environments and drop 32-bit support for their application entirely, since Windows 32-bit support was dropped more than half a decade ago.

 

[MxPhenom 216]

Capitalize on what? Software they haven’t supported in a decade?

This isnt that complicated....

Just capitalize on all the negative press (not just this software stuff). That's it

 

[HTC]

I suspect there may be something about Blackwell's ISA that hinders 32bit software from running natively. GPUs don't work like CPUs, you can't really write native software for them because GPU makers change the ISA often, a lot of the time from architecture to architecture, everything that runs on the GPU has to be compiled first, that's why I suspect there might be a hardware limitation somewhere so perhaps some 32bit instructions can't be issued or something like that.

On top of this Nvidia has the shitty habit of disclosing absolutely nothing about their ISA, like it's some sort of national secret.

I think the problem here is nVidia's "lack of disclosure".

From design to manufacture, a new generation of GPUs takes A LOT OF TIME to be made, so i would ASSUME nVidia knew about whatever it is that prevents them from enabling 32bit support A LOT EARLIER than when they disclosed the information to the software companies: WHY WAIT UNTIL NOW to do it?

 

[lilhasselhoffer]

I think looking at the history this is funny.

Windows 95 - first full 32 bit OS.
Windows 98 - 32 bit only
Windows 2000 - 32 bit only
Windows XP - 2001 - 32 bit and 64 bit
Windows Vista - 32 bit and 64 bit
Windows 7 - 32 bit and 64 bit
Windows 10 - 2024 - 32 bit and 64 bit
Windows 11 - 64 bit only

We've been living in a 32 bit and 64 bit windows world for 23 full years. I can absolutely see people who want to push technology ever forward saying that you've obsoleted the 32 bit instructions 8 years ago...but at that time there were plenty of people still on a 32 bit system that couldn't even use your 64 bit stuff...and it's why people basically assumed that without good reason these things would not disappear. Heck, simply translating the 32 bit instructions to run in 64 bit at basically half the efficiency (given half the bits would be filler) makes sense given that there are still people who can and do rock Windows XP...let alone those rocking 7 and 10...both of which had very popular 32 bit versions.

So it's amusing to me to see people want to fight about a benchmark having issues because they cut off code...when most of the time we hear about Nvidia and AMD optimizing for certain loads and benchmarks, because it means Nvidia either didn't understand what damage this would do or didn't care. If the former, it's weird to conceive of them optimizing for said benchmark. If the later, both AMD and Intel will bet artificially higher relative scores because they just didn't stop doing older things. That's...well, I take it as amusement given they don't otherwise seem to be having a good time after basically treating gamers as second class consumers because AI is splashing so much cash that they literally won't care about us until the AI bubble collapses. Payback's pretty frustration, especially if it's something you never concern yourself with. Hopefully Nvidia's decision to fully drop gets at least partially reversed into a translation layer, so this isn't a uniform damaging of performance in older games for no reason.

 

[b1k3rdude]

 

[Lycanwolfen]

Well If AMD releases the 9070XT at the right price of 699.00 and every supplier has stock and tons of it. Nvidia might learn a lesson again the gamers are not rich.

 

[Dr. Dro]

I think looking at the history this is funny.

Windows 95 - first full 32 bit OS.
Windows 98 - 32 bit only
Windows 2000 - 32 bit only
Windows XP - 2001 - 32 bit and 64 bit
Windows Vista - 32 bit and 64 bit
Windows 7 - 32 bit and 64 bit
Windows 10 - 2024 - 32 bit and 64 bit
Windows 11 - 64 bit only

We've been living in a 32 bit and 64 bit windows world for 23 full years. I can absolutely see people who want to push technology ever forward saying that you've obsoleted the 32 bit instructions 8 years ago...but at that time there were plenty of people still on a 32 bit system that couldn't even use your 64 bit stuff...and it's why people basically assumed that without good reason these things would not disappear. Heck, simply translating the 32 bit instructions to run in 64 bit at basically half the efficiency (given half the bits would be filler) makes sense given that there are still people who can and do rock Windows XP...let alone those rocking 7 and 10...both of which had very popular 32 bit versions.

So it's amusing to me to see people want to fight about a benchmark having issues because they cut off code...when most of the time we hear about Nvidia and AMD optimizing for certain loads and benchmarks, because it means Nvidia either didn't understand what damage this would do or didn't care. If the former, it's weird to conceive of them optimizing for said benchmark. If the later, both AMD and Intel will bet artificially higher relative scores because they just didn't stop doing older things. That's...well, I take it as amusement given they don't otherwise seem to be having a good time after basically treating gamers as second class consumers because AI is splashing so much cash that they literally won't care about us until the AI bubble collapses. Payback's pretty frustration, especially if it's something you never concern yourself with. Hopefully Nvidia's decision to fully drop gets at least partially reversed into a translation layer, so this isn't a uniform damaging of performance in older games for no reason.

To be more accurate, Windows 10 dropped active 32-bit support with version 2004 back in 2020, and banned OEMs from deploying it with new Windows licenses, so around two thirds of the way through its update cycles and around half a decade ago. A 32-bit version of newer versions, including 22H2 was still provided, but it is the last version of Windows in 32-bit and only available in the retail channel. It is therefore over 3 years old and its support is scheduled to end in around six months, when Windows 10 support is retired as a whole. Both Nvidia and AMD dropped 32-bit graphics driver updates in 2018, AMD actually did so earlier for all GPUs newer than the R9 290X, dropping Windows 8 in 2014 and Windows 8.1 in 2017. Furthermore, no hardware released in the past half decade includes 32-bit driver support of any manner.

No developers can really pull the card of "we didn't know it was coming", IMHO

Well If AMD releases the 9070XT at the right price of 699.00 and every supplier has stock and tons of it. Nvidia might learn a lesson again the gamers are not rich.

Missed the news?

 

[bonehead123]

Anutha day, anutha big, fat F.U. from nGreediya to consumer GPU buyers everywhere.......

And, I hate having to repeat my self, but..

HAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHA.....GOTCHA Suckas !

 

[boidsonly]

We'll see if NVIDIA can spell "lawsuit".

 

[igormp]

This tells you nothing about the reason as to why it wont do that, logically the hardware may not be suited to run 32 bit binaries, it's the same when you try and run a CUDA application on an older GPU which is shipped with a compiler that can't turn the given PTX into machine code.

Yeah, no way to know it without 100% certainty unless one works at nvidia.
In the same vein, we can't even be sure that it's the actual hardware that doesn't support that, or just the software stack.

Reminder that everything has to get turned into a compatible binary, even a regular shader,

Not really the case with CUDA since, as far as I know, the PTX is actually jitted into the actual SASS used by the underlying hardware. Iirc, this also applies to most (if not all) other APIs on top of GPUs.
The above is a bit pedantic, so feel free to ignore it if what you meant was something in a more broader sense.

so why is there a compiler that can turn regular shaders in 32bit machine code (presumably, maybe it promotes everything to 64bit but I doubt it) but can't do the same for 32bit PTX ?

Can it? I know that on linux you can't do so with Vulkan, not sure about Windows.
Nonetheless, in case it can, it corroborates with what I had said: the support was dropped from the software stack. The Vulkan/DirectX/whatever API you're running still has support to JIT 32-bit code into the proper SASS. The CUDA driver, however, has no such support for blackwell specifically. The GPU itself is not even aware of that.
Reminder that for each API, you have a specific driver that handles submissions to the GPU.

WIth the above said, there are other runtimes/compilers that allow you to emit 32-bit code and will support in some way. With open source drivers nothing stops you from emitting SPIR-V (not PTX) in any way you want and jitting those in the appropriate manner.

why not drop support for literally everything 32bit running on the GPU, why just 32bit PTX ?

Idk, but I believe they are slowly moving towards that.

I think the problem here is nVidia's "lack of disclosure".

From design to manufacture, a new generation of GPUs takes A LOT OF TIME to be made, so i would ASSUME nVidia knew about whatever it is that prevents them from enabling 32bit support A LOT EARLIER than when they disclosed the information to the software companies: WHY WAIT UNTIL NOW to do it?

But they have been saying so for years, that's what a deprecation notice is. Once something is deprecated, killing it may happen anytime, and nvidia announced so as soon as the new CUDA version supporting blackwell was released.

 

[lilhasselhoffer]

To be more accurate, Windows 10 dropped active 32-bit support with version 2004 back in 2020, and banned OEMs from deploying it with new Windows licenses, so around two thirds of the way through its update cycles and around half a decade ago. A 32-bit version of newer versions, including 22H2 was still provided, but it is the last version of Windows in 32-bit and only available in the retail channel. It is therefore over 3 years old and its support is scheduled to end in around six months, when Windows 10 support is retired as a whole. Both Nvidia and AMD dropped 32-bit graphics driver updates in 2018, AMD actually did so earlier for all GPUs newer than the R9 290X, dropping Windows 8 in 2014 and Windows 8.1 in 2017. Furthermore, no hardware released in the past half decade includes 32-bit driver support of any manner.

No developers can really pull the card of "we didn't know it was coming", IMHO

....

I hope my point is clear here. When 16 bit switched to 32 bit is was an on and off switch largely. I remember running a socket 2011 system with windows xp x64 because it was the only way to use all of the RAM that I could install...and it being a nightmare to get drivers until they just clicked and everything worked. I had to jump through hoops to play old 16 bit stuff...to the point where I largely didn't.

If you were to do the same, and give the maximum benefit of doubt, 20+ years when you could buy a 64 bit or 32 bit OS means that for some people's entire lives where they've operated a computer they had the choice...until very recently. And we're not talking toddler, we are talking legal to vote and drive cars (or drink and vote depending upon your region). That's kinda some silly long time to maintain something, that you always say is no longer supported, but keep supporting. I cannot make judgements, but it seems a bit silly to gimp performance eight years later...almost like somebody decided to finally do spring cleaning of code "that nobody used, obviously." They then find out somebody did still use it...which makes them either incompetent enough to not understand, incompetent enough to not have cleaned to now, or otherwise simply silly enough to add more fuel to the 20 car pile-up wreck that Blackwell has been.


I...like to think this was a low level developer finally cleaning out old code, looking to boost efficiency at no cost, and their dream of fixing things backfired because they don't understand the reason that nobody bothered to fix things was because the fix would hurt more than letting that stupid black box of code continue to fester as a virtually negligible overhead cost...because it's a beautiful tribute to "don't fix what ain't broken." An axiom that more engineers should live by.

 

[mechtech]

 

[Visible Noise]

I think the problem here is nVidia's "lack of disclosure".

From design to manufacture, a new generation of GPUs takes A LOT OF TIME to be made, so i would ASSUME nVidia knew about whatever it is that prevents them from enabling 32bit support A LOT EARLIER than when they disclosed the information to the software companies: WHY WAIT UNTIL NOW to do it?

Nvidia has been disclosing their plans for over seven years. How much notice do devs need? A decade?

 

[Dr. Dro]

Anutha day, anutha big, fat F.U. from nGreediya to consumer GPU buyers everywhere.......

And, I hate having to repeat my self, but..

HAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHAHA.....GOTCHA Suckas !

Right, now write that 200 times over and replace "ngreedia" with "amd" since Radeon never had CUDA or PhysX to begin with, yet you still happily purchase their cards anyway. Please don't act like a 5 year old, you can and generally do better.

 

[Vya Domus]

Not really the case with CUDA since, as far as I know, the PTX is actually jitted into the actual SASS

Whether it's compiled just in time or not the point is that it's strange that there are ways to compile some 32bit code but not 32bit PTX specifically, there are instructions which PTX will emit once compiled which you wont find else where, this is why I think there may be actual an actual hardware limitation, like for example new FP4 operations might be encoded in 64bit instructions only.

 

[Hecate91]

Right, now write that 200 times over and replace "ngreedia" with "amd" since Radeon never had CUDA or PhysX to begin with, yet you still happily purchase their cards anyway. Please don't act like a 5 year old, you can and generally do better.

This whataboutism isn't mature either, I'm so tired of seeing "but but other brand doesn't have the feature" because fans refuse to criticize their beloved brand no matter how shitty they're being to consumers and developers.

 

[igormp]

Whether it's compiled just in time or not the point is that it's strange that there are ways to compile some 32bit code but not 32bit PTX specifically, there are instructions which PTX will emit once compiled which you wont find else where, this is why I think there may be actual an actual hardware limitation, like for example new FP4 operations might be encoded in 64bit instructions only.

Hmmmm, could be. Nvidia never exposed things such as half precision to opencl, it was always a cuda-specific thing, so the actual underlying ISA on the gpu might not support such specific features in 32-bit mode.

This whataboutism isn't mature either, I'm so tired of seeing "but but other brand doesn't have the feature" because fans refuse to criticize their beloved brand no matter how shitty they're being to consumers and developers.

I don't see how this topic or the 32-bit thing has to do with treating devs in a shitty way. Nvidia has the entire moat exactly because it has the most dev-friendly API.

 

[Dr. Dro]

This whataboutism isn't mature either, I'm so tired of seeing "but but other brand doesn't have the feature" because fans refuse to criticize their beloved brand no matter how shitty they're being to consumers and developers.

Brother, literally stop with the whole victimism stuff. Other folks in the forum already called out on it. There is absolutely nothing wrong with finally removing a feature that was deprecated for 8 years. To make a big deal out of this while you buy the competitor's stuff regardless of it having it or not is simply hypocritical.

You want to play old PhysX games? Keep an older PC around to play game from 15-20+ years ago or stick to the 40 series. Life goes on. I'm not complaining they stopped manufacturing PS2's...