AMD “SSG” API Manual: Using a 2TB Solid State Drive with the AMD SSG GPU [pdf] | Hacker News Reader

AMD “SSG” API Manual: Using a 2TB Solid State Drive with the AMD SSG GPU [pdf]

pro.radeon.com

22 pointsdragontamer17 comments

I heard about this mysterious "Solid State GPU" around the internet, so I looked around for how to actually use it. Unfortunately, the $7000ish price tag is a bit too high for me to personally invest into, but the entire concept of making files on the GPU directly is intriguing. I'm curious if anyone has done anything with this hardware yet.

The manual is relatively straight forward, it shows 3 vendor-API extensions to OpenCL, OpenGL, and DirectX which allows the programmer to create files on the GPU. AMD markets this GPU as being able to play 8k video or perhaps do some kind of huge memory computation (without ever being bottlenecked by the PCIe bus).

EDIT: There's a second, shorter "device driver" manual which shows how the card works. Apparently, there are 4 NVMe SSDs on board the GPU.

http://pro.radeon.com/_downloads/SSG_Driver_Installer_Guide....

Interesting concept, downside is it is only available for Windows 10.

Well that's one way to destroy any chances of this gpu being useful for machine learning.

Why? One of my DL rigs runs on Win 10 reliably. Radeons anyway won't be a good candidate for DL, though for mining are best.

I'm not sure the 2TB of onboard NVMe would help machine learning tasks. Most machine learning tasks are already well designed for the ~16GB limit on standard GPUs.

The product page suggests that this sort of GPU can handle streaming 8k video. So maybe the card was designed for OpenCL accelerated video editing or some other 3d-task (maybe 3d rendering) with very large texture data.

You're absolutely correct on the video editing front.

The ability to work on full res 4k/8k video (in realtime) without generating down-scaled proxies will be a huge time-saver for editors in addition to saving money during post.

A lot of deep learning workloads aren't necessarily bounded by things like disk transfers, though, but stuff like bandwidth to the card itself. You probably have enough RAM to just pre-warm your disk cache anyway if you have exceedingly large training samples. This is more useful for things like massive CAD models, if I had to guess.

On Nvidia if you want this, I think you can basically achieve it using GPUDirect anyway if you really want to bypass CPU transfers and go straight from SSD to GPU, via PCIe. I guess it depends exactly on how this card is connected to the disk.

Having it on-card is pretty convenient, though. GPUDirect has practical limitations on what boards/systems it will work with.

Intel's Larrabee actually ran FreeBSD with a proper filesystem and everything. You could telnet into your GPU, install your own API implementation, pretty cool stuff: http://tomforsyth1000.github.io/blog.wiki.html#%5B%5BWhy%20d...

If they had a regular CPU on board too this would make more sense. But this could be useful for some volumetric work, like oil & gas visualisation. Maybe parallel database stuff too, but lots of porting. Just can't see what bottleneck it is fixing apart from super high res uncompressed video.

Dropped to ~$4600 now. https://www.tweaktown.com/news/59728/amds-new-radeon-pro-ssg...

If it worked on linux I would consider working with the device. Windows is a non-starter.

Its really odd that AMD is pushing so strongly with Open Source Linux Drivers / Open Source Vulkan Drivers / ROCm and then they come out with a $7000-card that doesn't work on Linux.

With that said, their "standard Vega" chips: the WX 9100, Vega Frontier, and the MI25 seem to work with Linux. Its just that this "SSG" is a $7000 anomaly.

Most likely, some big corporate customer wanted this card for maybe Adobe products (Photoshop? Premier?) or maybe 3d effects like Maya. Considering that AMD's "test product" was an older SSG "Developer Edition" a few years ago, clearly some big customer wanted this card, and clearly wanted it on Windows.

I guess it's a bleeding edge stuff if you want to avoid GPU being bottlenecked by PCIe transfer during computations, and their SW expertise is likely better on Windows. Even with the crazy SW hiring AMD does right now, it might take some time until Linux catches up.

From my perspective, this is all opposite land.

ROCm / HIP / HCC / AMD's next OpenCL implementation are all Linux-only right now, with no Windows support for the HCC compiler (even though its based on Microsoft C++ AMP).

Software over there seems like it must be a total mess. There is a pretty good sized market for people training DL models that Nvidia currently owns completely that they have not yet entered simply because they haven't made Tensorflow and Pytorch work with ROCm. Even the ROCm web pages are amateurish and appear to have virtually no work behind them.

Seriously, it's three very good developers and a few months of work that's holding them back from maybe a hundred million in sales, and at least 20 billion in their stock valuation from being perceived as being in the DL game.

maybe it just doesn't work well enough, yet. would be a pr decision to leave it out if i consider the price.

Support for linux was added recently.

You can the info here

http://support.amd.com/en-us/kb-articles/Pages/AMDGPU-Pro-Dr...

http://support.amd.com/en-us/kb-articles/Pages/Radeon-PRO-SS...

Reply on news.ycombinator.com