DeepSeek V4 Flash on a Single AMD MI300X

(github.com)

116 points | by zhoutong 3 hours ago

4 comments

  • xorfish 13 minutes ago
    This is still quite a bit away from the performance that deepseek gets on their H800. In their DSpark paper they report a throughput of 15k tokens/s/gpu. The MI300 should be able to compete with the H800 so there are probably still quite a few optimizations that can be made.
  • majke 2 hours ago
    I don't think you can buy a single "MI300X" unit, right? Only the box with x8 of these at a cost of ~250K EUR.
    • zhoutong 2 hours ago
      It’s available on demand from a few cloud providers. Seems like the cheapest is AMD Developer Cloud (https://www.amd.com/en/developer/resources/cloud-access/amd-...) powered by Digital Ocean at $1.99/hour.

      Edit: Now I think about it, this might be the cheapest way to run the DeepSeek V4 Flash 0731 on a dedicated inference server at original weights. I haven’t run mixed load benchmarks but I guess it’s possible to generate $3-$4 worth of tokens per hour and still maintain a usable per-user throughput.

      • WASDx 1 hour ago
        At 830tok/s * 1 hour that's almost 3M tokens which is just $0.54 worth of tokens at Deepseeks current output price.
        • lnenad 1 hour ago
          830t/s is burst aggregate. ~500 is sustained and it's for 8 concurrent users. Meaning for $1.99/hour if you serve 8 users it's 8*$0.54, not just $0.54.

          You shouldn't rent one out if you're just serving it for yourself, but from a financial standpoint if you sell to users you can take a 100% margin.

          • minraws 35 minutes ago
            Bro 500 Aggregate. so that's 500 * 60 * 60 = 1.8M output which is .5$ at best... Not including pre-fill and stuff.

            This is not the real margins, even if you are selling to 8 users it's 90 tps per median stream. So assuming that .6-.7$

            This is not even remotely worth it.

            You need to 3x this tps(~1500 tps) to be worth it, and that's what most providers are doing, at 20-30 users at 50-60 tps with better optimized batch processing and kernels you can make some profit.

            • lnenad 16 minutes ago
              You are right, it's not 500x8 it's 90x8.
        • Almondsetat 55 minutes ago
          You get privacy for 4 times the cost
        • krisknez 1 hour ago
          How is that economically viable? They are selling at a loss?
          • gpugreg 56 minutes ago
            Agentic workloads are somewhere around 1%/0.5%/98.5% input/output/cached tokens. Cached tokens are pretty much free for inference providers (if they implement sparse and compressed attention properly) and throughput for input tokens is much higher.

            Lets assume that you've got 2 million input tokens, 1 million output tokens and 98.5 million cached tokens to process. That would cost 2 * $0.14 + 1 * $0.28 + 98.5 * $0.0028 = $0.8358 with DeepSeek API pricing.

            For comparison, it would take 2M / 8000 + 1M / 800 = 1500 seconds to process this amount of tokens with the linked framework, which is about $0.83 when we assume $2/hr for one MI300X.

            However, other inference providers have 10 times higher prices for cached tokens, which results in a comfortable margin.

            And we should not discount that DeepSeek also gets paid in data, which is probably more valuable to them.

            And I believe that this framework still has some room for optimization for generation with high batch sizes.

          • dietr1ch 1 hour ago
            They claim their advantage is knowing how to serve their models efficiently, which is quite possible since they design for it.
          • pama 52 minutes ago
            Use nvidia hardware instead and use a larger cluster serving many more users concurrently. Easily 10x–20x higher token rate per GPU with public solutions like dynamo and sglang.
        • thrownaway561 1 hour ago
          This is exactly what I came to say. The price of Flash is so cheap that trying to run it locally or with your own hardware is pointless. I was using it about a month ago to program some stuff and ran it for 4 days non-stop and it cost me about $2.
          • NitpickLawyer 16 minutes ago
            > trying to run it locally or with your own hardware is pointless.

            Serving local models has advantages other than price. If you work in restricted industries, or have a strong need to protect your IP, or if you just value privacy more than cost, you now have options.

          • jorvi 15 minutes ago
            With the cost of electricity, hardware depreciation and tok/s it rarely makes sense to run locally.
      • langs 1 hour ago
        You need to optimize the KVCache part(save to disk to save compute) to achieve this goal.
    • Lwerewolf 1 hour ago
      The MI350p exists and should run a decent quant (say, the ~96GB antirez mix) well, but you can get two rtx pro 6000s for one of these, or 8x (actually more) r9700 + probably the gear to run them, etc.

      Otherwise, you can probably buy one of these second hand from somewhere (SXM A100s are available that way) and run it in an adapter board.

    • _joel 55 minutes ago
      I thought it was a consumer grade GPU until I saw the 192GB of HBM and 256GB or RAM.
      • varispeed 45 minutes ago
        To be fair the development of GPUs have stalled over the years. If they kept up with the progress instead of focusing on enterprise market, likely 256GB consumer GPU would be a norm today.
    • baalimago 2 hours ago
      Give it an AI-bubble pop and these will be flooding the market.
      • tamimio 3 minutes ago
        Thing is, GPUs will always be on demand, look at their history, initially for gaming, then for hash cracking, then 3D rendering, then for crypto mining, and now AI training and fine tuning. When AI bubble bursts, there will be another bubble taking over.

        The only solution is more companies making high end units, only competition will make it better for consumers.

      • _factor 2 hours ago
        They will be instantly bought out by companies, not individuals. The consumer bubble won’t pop for quite a while yet. Production also won’t ramp up while lack of real competition keeps the demand high.
      • aurareturn 1 hour ago
        When is it popping? Is the AI bubble in the room with us now?
        • amrit3128 38 minutes ago
          Tomorrow? Next year? In 5 years? Nobody can say. But we do know that AI is overvalued, so it WILL pop.
          • aurareturn 26 minutes ago
            Well, if no body can say when it will pop, then can we really say it's a bubble and it's overvalued?

            I can tell you it will pop in 10 years and when it pops, it will still be 20x bigger than in 2026. Does that even make any sense?

            People said AI bubble will pop soon in 2024 and that it was overvalued. Turns out, many AI stocks 10x, 20x since 2024. Actual usage has gone exponential as well. Anthropic revenue went from $100m ARR at start of 2024 to $80b ARR today.

            • ekidd 8 minutes ago
              > Well, if no body can say when it will pop, then can we really say it's a bubble and it's overvalued?

              Well, given the literal trillions being spent, the only ways this pays off are:

              1. AI replaces a non-trivial fraction of human employees.

              2. Someone builds a Culture Mind, and humans become (hopefully) pampered pets of AIs we don't understand. Seems unlikely, but it would arguably count as a payoff even if it made money meaningless.

              Or maybe the AIs don't want pets, and you get SkyNet. Which definitely doesn't care about paying off anyone's investments.

              When you look at various news articles about investors, yeah, there are definitely a lot of rich people who think that they're going to automate all human labor or just bring about the Singularity. Possibly with them in charge of the rest of us. If you don't make these kinds of wild assumptions, then yeah, this is looking like one of the biggest bubbles ever.

            • Joel_Mckay 9 minutes ago
              Many are saying July 2027, as in the past these market corrections have correlated with Shrek movie releases.

              Debt-backed investors have to pay up eventually. =3

        • baalimago 1 hour ago
          Next month perpetually
        • slaw 34 minutes ago
          The AI bubble will pop when China gets access to EUV, so the earliest it could happen is 2030
  • jkwang 2 hours ago
    [flagged]
  • hn0tdqaek4 12 minutes ago
    [dead]