AMD Radeon RDNA2 Navi (RX 6500, 6600, 6700, 6800, 6900 XT)

So the data is going to arrive in GPU memory and the PC can't do anything with it until the CPU processes it. Right.
 
There also is data cpu likes to use. It would be amazingly stupid if DirectStorage only worked for gpu use cases. Ideally DirectStorage could also be used to accelerate applications installs and such,...
 
We will probably get standard vulkan RT api once there is rt hw available from multiple vendors. Until then there is no way to get something ratified into standard.
 
DirectStorages whole point is to get the data directly to GPU bypassing CPU

No, it's whole point (at least insofar as the currently available information from Microsoft suggest) it to make the process of transferring data from SSD to the GPU more efficient by reducing the amount of CPU overhead required. That's completely different to using P2P DMA to transfer data directly from SSD to GPU bypassing the CPU completely. That may indeed be the way it works, but unless you can show me a statement from Microsoft explicitly stating that, then you're just making an assumption. Here's what Microsoft have to say about the way Direct Storage works:

Microsoft said:
Previous gen games had an asset streaming budget on the order of 50MB/s which even at smaller 64k block sizes (ie. one texture tile) amounts to only hundreds of IO requests per second. With multi-gigabyte a second capable NVMe drives, to take advantage of the full bandwidth, this quickly explodes to tens of thousands of IO requests a second. Taking the Series X’s 2.4GB/s capable drive and the same 64k block sizes as an example, that amounts to >35,000 IO requests per second to saturate it.

Existing APIs require the application to manage and handle each of these requests one at a time first by submitting the request, waiting for it to complete, and then handling its completion. The overhead of each request is not very large and wasn’t a choke point for older games running on slower hard drives, but multiplied tens of thousands of times per second, IO overhead can quickly become too expensive preventing games from being able to take advantage of the increased NVMe drive bandwidths.

On top of that, many of these assets are compressed. In order to be used by the CPU or GPU, they must first be decompressed. A game can pull as much data off the disk as it wants, but you still need an efficient way to decompress and get it to the GPU for rendering. By using DirectStorage, your games are able to leverage the best current and upcoming decompression technologies.

In a world where a game knows it needs to load and decompress thousands of blocks for the next frame, the one-at-a-time model results in loss of efficiency at various points in the data block’s journey. The DirectStorage API is architected in a way that takes all this into account and maximizes performance throughout the entire pipeline from NVMe drive all the way to the GPU.

It does this in several ways: by reducing per-request NVMe overhead, enabling batched many-at-a-time parallel IO requests which can be efficiently fed to the GPU, and giving games finer grain control over when they get notified of IO request completion instead of having to react to every tiny IO completion.

In this way, developers are given an extremely efficient way to submit/handle many orders of magnitude more IO requests than ever before ultimately minimizing the time you wait to get in game, and bringing you larger, more detailed virtual worlds that load in as fast as your game character can move through it."

No-where do they mention decompression being moved off the CPU or the flow of the data changing from SSD->CPU->GPU to SSD->GPU->CPU. Again, I'm not saying that's definitely not the case. Merely that at the moment we have no evidence of it so it would be better if we stopped stating assumptions as if they're facts.
 
Very nice showing from AMD, although some weirdness as well. The 3 SKUs aren't all that different in capabilities, thus the pricing is a little weird as well. $579 vs $649 for 6800 and XT is just... odd... Feels like a late price increase for the 6800 as perhaps it fared better against the 3070 as they initially thought? The price separation is so tiny that it just feels out of place and then the jump to 6900 is $350 with not much to show for it.

The 16GB across the range is a good thing to have and quite frankly imo, these put the current Ampere's in a tough spot. 3090 is overpriced and the other two are questionable on memory. I don't want to buy a new graphics card and question the memory amount on day 1. I'm not sure if they are having supply issues with the GDDR6X, but they need to roll out those 16/20GB SKUs out quickly.
 
They do look very interesting. Good to see competitive cards from AMD, the first in a while that have appealed to me at a time when I'm pondering an upgrade. A situation not hurt by the offerings from The Other Lot being ... distinctly odd IMO.
 
The 6900XT kills the 3090 in value if you're not concerned about ray tracing and are primarily gaming. If you're buying it for one of those tools like Blender or Octane, the extra $500 for the 3090 may have value if the ray tracing performance is significantly better, which I'm guessing it probably is.
 
No, it's whole point (at least insofar as the currently available information from Microsoft suggest) it to make the process of transferring data from SSD to the GPU more efficient by reducing the amount of CPU overhead required. That's completely different to using P2P DMA to transfer data directly from SSD to GPU bypassing the CPU completely. That may indeed be the way it works, but unless you can show me a statement from Microsoft explicitly stating that, then you're just making an assumption. Here's what Microsoft have to say about the way Direct Storage works:



No-where do they mention decompression being moved off the CPU or the flow of the data changing from SSD->CPU->GPU to SSD->GPU->CPU. Again, I'm not saying that's definitely not the case. Merely that at the moment we have no evidence of it so it would be better if we stopped stating assumptions as if they're facts.

At the very least because we all agree that high performance GPGPU decompression is a very questionable thing to announce.

;)
 
The 6900XT kills the 3090 in value if you're not concerned about ray tracing and are primarily gaming. If you're buying it for one of those tools like Blender or Octane, the extra $500 for the 3090 may have value if the ray tracing performance is significantly better, which I'm guessing it probably is.

My guess is if you're only gaming then you care about RT.

I hope RT perfs will appear soon
 
No, it's whole point (at least insofar as the currently available information from Microsoft suggest) it to make the process of transferring data from SSD to the GPU more efficient by reducing the amount of CPU overhead required. That's completely different to using P2P DMA to transfer data directly from SSD to GPU bypassing the CPU completely. That may indeed be the way it works, but unless you can show me a statement from Microsoft explicitly stating that, then you're just making an assumption. Here's what Microsoft have to say about the way Direct Storage works:

No-where do they mention decompression being moved off the CPU or the flow of the data changing from SSD->CPU->GPU to SSD->GPU->CPU. Again, I'm not saying that's definitely not the case. Merely that at the moment we have no evidence of it so it would be better if we stopped stating assumptions as if they're facts.

Direct Storage is an API. RTX IO works with Direct Storage and not in lieu of it. So whatever RTX IO does, it does so through functionality provided to it by Direct Storage.

"P2P DMA to transfer data directly from SSD to GPU bypassing the CPU completely"...this doesn't happen without DS as Windows doesn't readily cede control of its memory space to a third party device.
 
Last edited:
So AFAIU directstorage promises less overhead and RTX-IO promives decompression on the GPU (resulting in less processing on the CPU) but I've seen neither really talk about P2P DMA explicitly? (though some RTX-IO diagram certainly make it look like it is skipping the CPU completely).

I'm kind of sceptical for two reasons:

1) You need to figure out how to avoid all CPU processing including the filesystem layer. This is going to be messy with things like full disk encryption and we haven't heard anything about such restrictions yet.
2) For significant P2P DMA you need access to a large PCIe BAR, unless you can only DMA into a 256MiB subset of VRAM. This needs BIOS support and is rare enough on gaming machines that AMD just marketed it as "Smart Access Memory", so I doubt NVidia can rely on this without mentioning this limitation at all.
 
Direct Storage is an API. RTX IO works with Direct Storage and not in lieu of it. So whatever RTX IO does, it does so through functionality provided to it by Direct Storage.

"P2P DMA to transfer data directly from SSD to GPU bypassing the CPU completely"...this doesn't happen without DS as Windows doesn't readily cede control of its memory space to a third party device.

Perhaps, so like DirectX, there may be options that hardware vendors can choose to support or not. Simply supporting Direct Storage doesn't necessarily mean you implement everything it has to offer.

Alternatively, since RTX-IO uses its own API it may be more akin to RTX raytracing. Working alongside DirectX but not being a core part of it.
 
So... will there be any stock available or is this gonna enrage gamers everywhere :LOL:

My attempts to get a 3070 for msrp price failed. Maybe destiny wants me to go team red
 
Back
Top