Compute

While I’m including compute performance for the sake of completeness here, the compute situation on Navi has not substantially changed since the launch of the Radeon RX 5700 series over 5 months ago. AMD’s Adrenaline 2020 software has improved the state of their OpenCL drivers slightly – there are fewer hard crashes and performance is up in some cases – but their drivers are still dysfunctional and not fit for production use. In particular, Folding@Home and parts of CompuBench are still unable to run (and SETI@Home users will want to stay clear too).

AMD is aware of the issue, unfortunately they don’t have any updates to offer on the situation. I am of the distinct impression that AMD has made OpenCL on Windows a low priority for now, and has opted to focus their software efforts on bringing up additional Navi GPUs, as well as improving Navi gaming performance and continuing to develop their ROCm platform for Linux. So anyone looking to do GPU compute on AMD’s GPUs would best be served by using Vega or Polaris cards if they’re using to Windows, or switching to Linux for these matters.

Compute: LuxMark 3.1 - Hotel

Compute: CompuBench 2.0 - Level Set Segmentation 256

Compute: CompuBench 2.0 - Optical Flow

The Division 2, Grant Theft Auto V, & Forza Horizon 4 Synthetics
Comments Locked

97 Comments

View All Comments

  • YB1064 - Thursday, December 12, 2019 - link

    Poor man's NVidia indeed.
  • flashbacck - Thursday, December 12, 2019 - link

    What do you mean by that?
  • Valantar - Thursday, December 12, 2019 - link

    Apparently the same performance at the same price with a better encode/decode block and better idle power usage but slightly worse gaming power usage and buggy OC controls equals "poor man's Nvidia". Who knew?
  • assyn - Thursday, December 12, 2019 - link

    Actually the Nvenc is better encode/decode because has wider format support and can decode 8K videos, while the AMD only capable of 4K.
  • Valantar - Thursday, December 12, 2019 - link

    The 1650 and 1650S have a cut-down NVENC block lacking pretty much every single improvement made with the Turing generation. Not all NVENC is equal.
  • Ryan Smith - Thursday, December 12, 2019 - link

    1650S has a full NVENC block. It's based on TU116, not TU117.
  • jgraham11 - Friday, December 13, 2019 - link

    Seems like this will be similar to the GCN cards in the past (RX5XX Series), at the beginning its ok, then over the years AMD's fine wine kicks in. Making incremental improvements that improve overall performance with every driver release. Do realize they are still using the GCN instruction set but RDNA is a new architecture, an architecture that will be in both the Xbox and PlayStation systems.
  • WaltC - Saturday, December 21, 2019 - link

    I feel as though AMD has hobbled this card. They put the GDDR6 VRAM onboard, effectively doubling more or less the bandwidth of GDDR5, but then they took it all back with the 128-bit bus...;) I think they missed an opportunity here--but I'm not privy to the manufacturing particulars for the product--so maybe not. Anyway, it seems like they might have made the 5500XT a GDDR6 8GBs, 256-bit bus GPU for $279; then made the 5500 with 4GB, and a 128-bit bus for ~$129 or so. The SPs by themselves are sufficiently cut down to the degree that even a 256-bit bus GPU like I've dreamed up here is going to be sludge next to a 5700--or, that should certainly be the case, I would think. The card almost seems "overly hobbled" to my way of thinking. I think that @ 256-bit 5500XT, 8GBs of GDDR6 might've been a smash seller for AMD @ a ~$279 MSRP, leaving the 5700 still clobbering it simply based on the much higher number of SPs. I just see a lot of potential here that seems overlooked--unless they have plans for a 5600XT, maybe, early 1 Quarter 2020....;)
  • WaltC - Sunday, December 22, 2019 - link

    Will someone please set me straight on whether this is an 8-lane PCIe card? I'm reading this stuff you-know-where but it doesn't make any sense at all to me. It's claimed that accordingly the 5500XT runs much faster on PCIe4 than PCIe3. Eh?

    https://www.reddit.com/r/Amd/comments/edo4u8/rx_55...

    RX-480/580/590 are all PCIex16 GPUs with 256-bit buses running GDDR5. Is the article linked here factual? thanks--to whomever can answer....;) Doesn't sound probable to me--but if it's true then AMD has hog-tied this card six ways to Sunday--trussed up tighter than a Christmas piglet...;)
  • WaltC - Sunday, December 22, 2019 - link

    OK, I saw the one line in the review I missed first time about "I suspect 8 PCIe lanes"--just very strange--I don't understand this product at all! Perhaps the full incarnation of the GPU is going into the PS5/XBox 2, it's just weird. To have to hobble it that much, the GPU must be very robust.

Log in

Don't have an account? Sign up now