ASUS Chinese GM—Tony Yu—has shown off a graphics card concept on Bilibili that has a rather unusual feature, a slot for an M.2 NVMe SSD. The card is based on NVIDIA's GeForce RT 4060 Ti GPU and although not all details are clear at this point in time, but ASUS is taking advantage of the unused PCIe lanes on the card, since the AD106 GPU only uses eight PCIe lanes, the PCIe connector on the card has space for a further eight lanes. In theory ASUS could have added a pair of SSDs, since there are a total of eight lanes available, but as this was just a proof of concept, they seemingly stuck with a single SSD.
It's unclear if ASUS relies on bifurcation or if the company has added some kind of bridge chip, but bifurcation makes more sense, as a bridge chip would add a lot more cost. The neat thing with the NVMe drive being on the GPU, is that it also connects to the heatsink of the graphics card, which means the cooling should be rather good. However, for this to work properly, the SSD would have to be mounted back to front compared to how it would be mounted on a motherboard. Based on the test results, the SSD runs at a cool 42 degrees C, even when the GPU is being stress tested. It's likely that this product will not make it to markets outside of China, if it's ever launched into retail.
68 Comments on ASUS has a GeForce RTX 4060 Ti Card with an M.2 SSD Slot
I'm still not sure if the GPU 'takes in' the full x16, and bifurcates it internally, somehow.
Or, if there's basic PCIe packet-switching in the GPU? If this requires Bifurcation-support, expect a lot of negative feedback from users that didn't know their board didn't properly support the feature.
Asus is just 'with the times'. I hear that Their high-end boards supposedly like to be a lil sadistic on the voltage, too.
Don't judge Asus for being into kink :p Me too.
The Radeon Pro SSG
It was (and is) fairly unique for exposing the NVMe SSDs as extended-VRAM.
IIRC, it used an in-house (Xilinx?) FPGA; presumably, with PCIe Switch 'programming' and AMD's own integrations.
What you speak of is kinda the other side of the coin in my desire for M.2-adaptable x4-lane GPUs: I just want to be able to fully-utilize all the lanes available to me. With DirectStorage promised to be a thing (supposedly) in the near-future, I could genuinely see this become more common. The overall shorter trace-lengths going to/from the Primary CPU-connected PCI-e x16, are also going to help reliability/speed for storage.
Why they didn't opt for 8+4+4?
Asus being full retards as usual. I call it a sloppy job.
Oh they actually have, but made one pain in the arse to remove...
This idea from ASUS could probably offer a partially fix for this limitation, by offering a x8 + x4 configuration. That leaves us of course with 4 more lanes unused, so if this idea gets adopted from more models and graphics cards manufacturers, we might even see graphics cards with two M.2 slots on them.
I presume you've seen the are already M.2 ML/AI accelerators, but they're not GPU based.
So your GPU could cache, while it caches and you cache yourself a chicken dinner.
See here
www.asus.com/support/FAQ/1037507
I'm a bit unsure if x4+x4+x4+x4 means that GPU can only do x4 or if x4+x4 still allows for x8 or that it really must be x8+x8 or x8+x4+x4.
If it's the former many support it, if it's not, it's a lot harder to find a cheap(er) board.
I think it's a bit of a pointless product however. Pricing must be seriously higher which kind of defeats the purpose of the entire product.
If they're going to sell it it's probably going to be OEM only.
Of course, now the CPU manufacturers will use it as an excuse to give us even fewer PCIe lanes. Used to be that dual x16 slots were the norm on mid-range boards, now you're lucky if you get a second x8 on a high-end one, and you have to pay through the nose for "prosumer" CPUs and boards to get that basic functionality.