16-core ”Warsaw” CPUs added to Opteron lineup

AMD has added a couple of server processors to the Opteron 6300 series. These new offerings are based on Warsaw, which appears to be a power-optimized version of the silicon anchoring existing members of the 6300 family. Like the current models, the latest additions are built on 32-nm fabrication technology and based on Piledriver CPU cores. They have the same cache configurations, and they’re reliant on the same dual-die approach to provide six- and eight-module configurations with 12 and 16 cores, respectively.

The new models include the Opteron 6370P, a 16-core chip with a 2GHz base clock speed and a 2.5GHz Turbo peak, and the Opteron 6338P, which has 12 cores clocked at 2.3/2.8GHz. Here’s how those stack up against the other 12- and 16-core chips in the Opteron 6300 series.

Model Cores Base clock Max Turbo clock Power band

Price
6386 SE 16 2.8GHz 3.5GHz 140W $1392
6380 16 2.5GHz 3.4GHz 115W $1088
6378 16 2.4GHz 3.3GHz 115W $867
6376 16 2.3GHz 3.2GHz 115W $703
6370P 16 2.0GHz 2.5GHz 99W $598
6348 12 2.8GHz 3.4GHz 115W $575
6344 12 2.6GHz 3.2GHz 115W $415
6338P 12 2.3GHz 2.8GHz 99W $377

The Warsaw-based models have lower base and Turbo frequencies, but they’re also cheaper than the higher-clocked options. That makes sense, since the chips are "optimized to deliver improved performance per-watt for virtualized private cloud deployments with less power and at lower cost points," according to AMD.

AMD’s press release provides "power band" figures instead of traditional TDP ratings. However, the accompanying quick reference guide suggests that’s just new terminology for the same metric. The power band ratings for the older members of the Opteron 6300 series match the TDP ratings for those chips. The thermal envelope for these latest Opterons is only 99W, 16W lower than for the bulk of the 12- and 16-core lineup.

The Opteron 6370P and 6338P are available starting today from multiple system integrators. They’ve also been qualified for servers from Sugon and Supermicro.

Comments closed
    • green
    • 7 years ago


    6370P vs 6348

    2.8Ghz – 2.0Ghz = 0.8Ghz
    0.8Ghz / 2.0Ghz = 40% faster

    3.4Ghz – 2.5Ghz = 0.9Ghz
    0.9Ghz / 2.5Ghz = 36% faster

    16 cores – 12 cores = 4 cores
    4 cores / 12 cores = 33% more cores

    i guess i depends on the workloads you’re working on, but wouldn’t the 6348 be better in a few respects?
    personally if i could go 50% more cores or 50% faster at “close to” the same power, i’d go for speed

      • ronch
      • 7 years ago

      Yep. Would I choose a 16-core Opteron running at 2.0GHz or one that’s a 4.0GHz 8-core? The latter of course. And by this reasoning an FX-8350 should be generally better than an Opteron 6370P in terms of performance. Of course with Opterons you do get some extra features, particularly multi-socket support, quad-channel memory support, RAS, etc.

    • sschaem
    • 7 years ago

    Not very exciting… And I’m amazed anyone still uses 32nm fabrication for “power optimized” chips.

    GlobalFoundry must have AMD total screwed over to have them order more of those legacy chips in 2014…

    I still wonder if AMD would have been better off financially if they burned the Dresden factory to the ground…. seriously…

      • tipoo
      • 7 years ago

      You think that’s amazing, Nintendo went for 45nm chips!

    • ronch
    • 7 years ago

    These are still dual-die processors. I thought there was a leak about AMD planning a native 16-core Opteron lineup?

      • Xenolith
      • 7 years ago

      Yes, next gen. Not this gen.

    • chuckula
    • 7 years ago

    In a semi-related note, our good friend Charlie put this spin on the exact same story: “AMD launches a pair of [b<][i<]Steamroller[/b<][/i<] based Opterons" (emphasis added and link not included intentionally) Just remember that he wants you to pay $50 a pop for his... "insights"... yeah. For a mouth-breathing AMD sycophant, you'd think Charlie would at least get AMD's own CPU core codenames right in the headlines of his own articles....

      • Unknown-Error
      • 7 years ago

      $50?

        • chuckula
        • 7 years ago

        There’s no record that anyone, other than Fighterpilot, has actually paid.

      • Fighterpilot
      • 7 years ago

      At least Charlie gets paid for being one…./irony

        • maxxcool
        • 7 years ago

        ^ hooooooooooo!

      • NeoForever
      • 7 years ago

      Then again, he’s done such “typos” before to flush out the plagiarists.

      • just brew it!
      • 7 years ago

      He’s corrected it to Piledriver now.

      And the article is hardly glowing — “AMD is announcing two new [b<]old[/b<] Opterons" -- so I'm not sure how you're interpreting it as evidence of being an AMD sycophant.

        • NeoForever
        • 7 years ago

        It’s funny he brought up Charlie in the first place. The rest is just icing on the cake.

    • Unknown-Error
    • 7 years ago

    99W TDP? I hope this is not like the 125W TDP of FX-8350 or 45W/65W TDP of A8-7600 which turned out to be just labels.

      • Duct Tape Dude
      • 7 years ago

      Has TDP ever [i<]not[/i<] been a label? Did I miss something?

        • Waco
        • 7 years ago

        I think his point is that those CPUs/APUs tend to actually use more power under certain loads than their TDP indicates.

          • sschaem
          • 7 years ago

          Do you have a link to a definitive article on that?

          Many review do add a “power usage” section but they are not done professionally, so its hard to tell.
          (And on top of that, motherboard/bios can skew numbers)

            • Waco
            • 7 years ago

            I wish I did – I remember reading a few a while back when the 8350 first came out and for the life of me I can’t find them.

            It wasn’t a huge discrepancy (and it could easily have been BIOS related) but I remember seeing something along the lines of 130-140 watts used for CPUs rated at 125 watts. Hell, I don’t even remember how they measured the CPU usage on its own (or how they isolated it) but I do recall they had some setup to do so…

    • ronch
    • 7 years ago

    I’m getting sick of chips that the manufacturer (I’m looking at you, AMD) say are ‘power optimized’ but run at lower clock speeds. The only good thing about having these is how lower TDPs are available to you now at lower prices, instead of buying, say, a 6376, clocking it 300MHz lower and getting a lower (by 16w) TDP.

    This is the opposite of offering a ‘5.0GHz’ CPU at stratospheric TDP levels.

    It’s just sad how AMD is struggling so much lately. My 990FX board bit the dust a few days ago and today i spent a few hours scouring my country’s ‘computer capital’, only to see how only two stores are offering a 990FX-based board, and both are Asrock’s 990FX Extreme3 at that. Not the best board for my FX-8350. Apparently no one wants to carry FX processors or 990FX-based boards anymore because very few people want them.

      • Srsly_Bro
      • 7 years ago

      How else could AMD or any other company off power optimized chips without lowering clock speeds AND have sufficient quantities to sell?

        • ronch
        • 7 years ago

        They’re not really power optimized. They just lowered the clocks so obviously power consumption will go down. Nothing special, probably not even specially binned chips. More proof? Downclock your CPU. Now is it power optimized? No. Just lower performance along with the lower power consumption that naturally comes with it.

      • DPete27
      • 7 years ago

      Are you not in the US ronch? [url=http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=100007625%20600166242&IsNodeId=1&name=AMD%20990FX<]newegg has 12 990FX boards.[/url<] Not a great variety, but better than 1

        • ronch
        • 7 years ago

        No, I don’t live in the U.S.

      • Deanjo
      • 7 years ago

      [quote<]Apparently no one wants to carry FX processors or 990FX-based boards anymore because very few people want them.[/quote<] More to the point, motherboard manufacturers do not want to invest into developing new boards on a stagnant platform that is very outdated and a dead end.

      • LaChupacabra
      • 7 years ago

      A lot of my infrastructure is using small virtual linux machines that only need 1 “core” and half a gig of memory. It’s very disk I/O bound. I’m sure there are a lot of other people in the same boat.

      Guess not everyone sees the world the same, though.

    • chuckula
    • 7 years ago

    In keeping with video game tie-ins, I would like to see UT-2004 War Cow branded Opterons.

      • OneArmedScissor
      • 7 years ago

      Great. Now AMD will do a Redeemer…

    • GabrielBelmont
    • 7 years ago

    What happened to G-Sync analyses articles?
    We were getting a gsync a day. I miss those articles.

      • internetsandman
      • 7 years ago

      The industry hasn’t yet finished rendering the next frame of news, so your news outlets have yet to receive and update their displays with the relevant information

        • Arclight
        • 7 years ago

        Basically we are getting the same news posted for a longer until the next ones have finished rendering.

        • ALiLPinkMonster
        • 7 years ago

        Sounds like they’re still using AMD editors. Do they even know?

          • internetsandman
          • 7 years ago

          I think you missed the joke whereby there are no multiples of a position at AMD; one driver writer, one editor, one marketer 😛

            • ALiLPinkMonster
            • 7 years ago

            Oh, yeah that works too. I took it as more of a non-targeted joke about synchronized refresh rates in general.

Pin It on Pinterest

Share This