Xeon 5600 series brings six-core chips to 2P systems

Intel’s gradual roll-out of 32-nm microprocessors has already paid dividends on the desktop, and today, servers and workstations are getting the same treatment.  The chip code-named Westmere-EP in its development is the backbone of the new Xeon 5600 series, a top-to-bottom refresh of Intel’s 2P Xeon server and workstation lineups.

The potency of Westmere-EP silicon will be familiar to TR readers who read our review of the Core i7-980X Extreme last week.  (The code name for the desktop versions of Westmere-EP is Gulftown, but they’re essentially the same chip.)  Westmere-EP was designed to be a drop-in replacement for Xeon 5500-series (Nehalem-EP) processors, with an upgrade to six cores and 12MB of L3 cache per socket.  The Xeon 5600 series has the same basic clock frequencies and power envelopes as the 5500 series, but obviously, performance should be quite a bit higher with 50% more cores and cache. 

Clock-for-clock performance should be up due to some architectural enhancements in the Westmere family, as well, including a group of seven new instructions, dubbed AES-NI, intended to accelerate common encryption algorithms.  We covered most of these changes in our Gulftown review, but the 2P server market is undoubtedly Westmere-EP’s true home.  For instance, some targeted changes should improve virtualization performance, and the chip’s memory controller enables support for the low-voltage DDR3 memory now becoming available from major memory makers such as Samsung.  The memory controller supports dual DIMMs per channel at 1333MHz clock speeds, too, where Nehalem-EP supported only one per channel.  The addition of more buffers should allow for higher memory bandwidth at the same memory frequency, as well, according to Westmere’s architects.

A new power gate to cut off power to certain parts of the processor’s “uncore” section promises lower power draw at idle.  Both sockets must be idle in a 2P system in order for the gate to take effect, however, since both chips will be responsible for memory locations that the other processor may wish to access.

The table below shows the revamped Xeon lineup, with key specs for the different models.  Any products in the 5600 series should be Westmere-based, though some have had cores or features disabled.

Most of the entrants in the table above are fairly straightforward, though several are notable for their strangeness.  Several low-voltage Xeons with two of their cores disabled can fit into 40W power envelopes, which is fairly remarkable.  A handful of others are workstation-tailored parts with 130W power envelopes, including the Xeon X3680, which looks to be essentially the same thing as the Core i7-980X, though with Xeon validation and branding behind it.

No doubt most of the attention from IT administrators will focus on the 95W and 80W parts, which are the mainstream offerings.  Note that many of the 80W Xeons in the 5600 series are quad-core, eight-thread models with two defunct cores.  The six-core parts carry a bit of a price premium.

They should be worth the extra money, though, if our experience with the single-socket Gulftown is any indication.  The combination of more cores and cache added up to excellent performance scaling, even in our scientific computing benchmarks, which had traditionally been limited by memory bandwidth.  Meanwhile, per-socket peak power consumption remained essentially steady versus the 45-nm quad-core Nehalems, yielding a substantial improvement in performance per watt.  If anything, Westmere-EP should be an even more compelling proposition in 2P servers and workstations, where the software infrastructure generally stands ready to take full advantage of 12 cores and 24 threads.  The Xeon 5600 series will face somewhat more formidable competition in the form of the six-core “Istanbul” Opterons, although signs still point to a clean sweep of the performance and power-efficiency contests by these new Xeons.

We’d best not get ahead of ourselves on that point, however.  Several members of the Xeon 5600 series have already made their way into Damage Labs, and we have a full review planned. We’re still assembling some of the test equipment, including our fancy new Yokogawa power meter, enterprise-class SLC SSDs, some noteworthy new benchmarks, and the latest hardware from the competition at AMD.  Stayed tuned for a full review in the coming weeks.
Comments closed
    • maroon1
    • 10 years ago

    §[<http://it.anandtech.com/IT/showdoc.aspx?i=3769&p=9<]§ If you look at benchmarks in the review above you would notice that Dual Xeon X5670 is performing as good as Quad Opetron 8430 (6-core Istanbul) I'm pretty sure that four Opetron 8430 cost more than two Xeon X5670

    • RagingDragon
    • 10 years ago

    I like the X5677 specs, though not the price. A less expensive single socket version, say a Xeon W36xx or an i7-9xx would be nice.

    • oldDummy
    • 10 years ago

    Gulftown is for sale on Newegg.

    §[<http://www.newegg.com/Product/Product.aspx?Item=N82E16819115223<]§ EDIT: Oops out of stock.

      • NeelyCam
      • 10 years ago

      Yeah, and $1,099. Just a matter of time before that’s $1,199…

      Market sets the price. That’s just how it is.

    • StashTheVampede
    • 10 years ago

    In the Damage labs, can you please use Battlefield Bad Company 2 as a benchmark? That game scales with cores (seen with and without HT), please include.

      • StuG
      • 10 years ago

      I second that!

      • Krogoth
      • 10 years ago

      It only goes up to four threads and is still limited mostly by the GPU when played at higher resolutions.

        • StashTheVampede
        • 10 years ago

        This shows 2vs4 cores having big gains:
        §[<http://www.nvnews.net/vbulletin/showthread.php?t=148199<]§ Still can't find a comparison between 4 and 8.

          • Krogoth
          • 10 years ago

          Four threads does yield a benefit over two threads, but look at the graphs and figures again. It is a noticeable to a nice gain, but not earth-shattering (10-70%). It is all at a relatively low-resolution with no AA/AF. It is interesting that K10 based chips benefit more from it than Westmere and Nehalem-based stuff.

          Anyhow, GPU still matters far more than CPU. I doubt workstation platforms will see any benefit since they are operate at full load with eight or more threads.

            • StashTheVampede
            • 10 years ago

            I’m not questioning the scaling of threads, but I just want to see a benchmark in relation to the threads. Naturally a GPU will be important for a higher resolutions, but as a pure CPU benchmark for scaling, BC2 could be used.

    • grantmeaname
    • 10 years ago

    l[

    • valrandir
    • 10 years ago

    Unfortunately for Intel, POWER7 runs circles around Xeon Nehalems. It comes with 32MB of L3 cache, 4 SMT threads per core, in 8 or 16 core MCM modules at 4GHz+ clockspeeds (with similar IPC). Thus a dual socket POWER7 system can have up to 128 threads in flight.

    §[<http://en.wikipedia.org/wiki/POWER7<]§ Even Nehalem EX wont compete (well, maybe at a price lower level)

      • SomeOtherGeek
      • 10 years ago

      Yes, it does! It is unbelievable. Including the price, in the ballpark of 100K per core?

      • StashTheVampede
      • 10 years ago

      Can Power7 run any x86 apps? How about an x86 VM?

      When PowerX can do that, MAYBE you’ll see some dent in Xeon sales, but I’m doubting it.

      • sschaem
      • 10 years ago

      I only see a mention of 8 cores per chip, rated at a total of 256GFlops.
      And IBM wants 174,000$ for a 4 socket / 32core (total) server.
      §[<http://www-03.ibm.com/systems/power/hardware/<]§ A 5870 is rated at 2 teraflops ? So it seem a Radeon card can run circle around a Power7 server, for 1100x time less $ per GFlops. From this math, A 128 HW thread Power7 system could not run Crysis :) Seriously, for that kind of cash you can rackmount allot of AMD or Intel blade servers, and still have money to add dozens of GPU to run stream programs.

        • loophole
        • 10 years ago

        Indeed – IBM’s Power systems very rarely compete on price. And in fact IBM doesn’t pretend that they are competitive with x86 systems. For that space they’ve got their own x86 systems. Most of those who buy Power systems do so for their high-availability and RAS features, or have software that’s been optimised for the architecture (and usually runs on AIX rather than Linux).

        With Nehalem-EX though, you’re getting a lot of the high-end RAS features that were previously not found on x86 systems.

        • stdRaichu
        • 10 years ago

        Disclaimer: we have a bunch of POWER servers at work for backend databases. x86 just doesn’t cut it when you need absolute balls to the wall performance – most of our databases live on oracle/RHEL running on VMware, but all the real heavy lifting gets done on the POWER boxes. They’re certainly not cheap but you’re paying for a) reliability b) service c) ludicrous levels of backwards compatibility (plus the ability to LPAR) and d) performance I’ve not seen equaled on another platform. They’re essentially a stepping stone between big x86 tin and mainframe iron.

        I doubt GPU’s could provide the amount of float accuracy needed for the financial applications whilst still retaining the speed (and that’s assuming the algorithm is embarrassingly parallel enough to run on a GPU in the first place); on top of that, POWER includes several specialised execution units, including a very funky one that works in decimal, something I don’t think any other processor family can do.

        As such, despite them often being compared to x86 on basis of FLOPS alone, they’re utterly different beasts.

      • JumpingJack
      • 10 years ago

      Power 7 is a monster isn’t it. Unfortunately, it is a niche’ product and isn’t likely to really do much in the volume/commerical space. For years, the RISC market has been dwindling to the point where I have seen some editorialize that it is just a matter of time before it evaporates all together.

      I disagree, I think there will be a solid market for it in supercomputing/mainframe/HPC space.

    • valrandir
    • 10 years ago

    Mmm, I bet we’ll see a 40W 6-core version soon.

    • shank15217
    • 10 years ago

    Mangy Cours is supposed to run on 2P systems. Infact that’s their biggest selling point. Manufacturing cost for AMD has always been higher because Intel is perpetually ahead in manufacturing tech. AMD will not make the kind of money Intel makes on their 12/8 core cpus. This isn’t going to change much with bulldozer as AMD is still not using SMT so their chips will be bigger for a given core count, however their biggest selling point should be higher performance at a given price, ease of upgrades and reduced validation times between processor generations…. Dang I thought I replied to Damage

      • Damage
      • 10 years ago

      Magny-Cours may run on 2P systems, but I think you’re looking at a new class of 2P systems based on the G34 socket with more memory bandwidth, higher memory capacity, and quite likely higher prices than existing 2P stuff. I think Intel will have an answer for that market that isn’t Westmere-EP.

        • shank15217
        • 10 years ago

        Yes, it will be an interesting year.

          • NeelyCam
          • 10 years ago

          No – it’ll be a boring year. Intel has 2010 in the pocket.

          But 2011 WILL be an interesting year.

      • AlvinTheNerd
      • 10 years ago

      Bulldozer doesn’t have SMT, but they have a heterogeneous core layout that will help their silicon per core ratio. The bulldozer system has two integer cores for each module. That means they add 50% more silicon for the second integer core instead of 100% like it is currently. Considering that 2 real cores are going to perform better, this should give them the advantage.

        • shank15217
        • 10 years ago

        Agreed, but that configuration also increases die size significantly. AMD is betting on overall higher performance.

    • Thresher
    • 10 years ago

    I don’t really keep up on server chips. Could someone decode what the X, W, E, etc. mean?

      • Steel
      • 10 years ago

      My guess is, since Intel doesn’t really make it obvious on their site, is W = single socket workstations, L = low power, E = mainstream (economy?) and, of course, the X stands for Xtra.

        • Game_boy
        • 10 years ago

        The W5580 is dual-socket though. Strange that their new top DP chip is not.

        §[<https://techreport.com/articles.x/16656<]§

        • Umbongo
        • 10 years ago

        It is an indicator of the TDP range.

    • SomeOtherGeek
    • 10 years ago

    Reply FAIL!

    • OneArmedScissor
    • 10 years ago

    I hate to be the finger pointer, but there are a few more than minor issues with this article.

    “The Xeon 5600 series will face somewhat more formidable competition in the form of the six-core “Istanbul” Opterons, although signs still point to a clean sweep of the performance and power-efficiency contests by these new Xeons.”

    Wow. Double face palmer. Way to forget the new 6, 8, and 12 core AMD CPUs, which should see availability at about the same time as these new Intel CPUs, and also emphasize improved power efficiency. It’s one thing to “forget,” something, but that went too far by effectively handing Intel the efficiency performance crown for a race that hasn’t even started.

    I come here for objective news, and I’m a bit taken aback by such a gaping hole left in the big picture there. I don’t know everything, which is why I come here. I may have caught that “mishap,” but someone else invariably won’t, and it makes me wonder what else I’m not catching.

    “Several low-voltage Xeons with two of their cores disabled can fit into 40W power envelopes, which is fairly remarkable. ”

    At those prices and clock speeds, for crippled chips, that’s highway robbery. Both Intel and AMD have been selling 45w quad-cores for half a year. One of Intel’s own Xeon 45w quad-cores is a much larger, fully functioning chip, and has its own completely unique turbo boost profile…but it’s still about half the price of the cheapest Westmere listed there. I don’t seem why Intel should be commended for that. Someone should be slapping them around.

    “The combination of more cores and cache added up to excellent performance scaling”

    You didn’t compare it to a six core chip with less L3, which this very chart shows there are several of. That’s just making assumptions. You could say that more cores scaled, but why in the world tack on the part about the cache? I’m sorry, but I really do not agree with this, “piling on the cache = moar fastar” dogma that PC sites perpetuate, without ever actually proving. It is not hard to do, but no one bothers, which gives Intel too much power to market by numbers vs. reality. People buy things based on what sites like TR say, and I think it’s your responsibility to back up statements like that.

    This is what leads to people asking for suggestions on a $500 build with a Pentium dual-core being spammed with responses on how they need the $200 huge chip. All too often, it’s totally blown out of proportion.

      • SomeOtherGeek
      • 10 years ago

      Wow! I think you jumped the gun here… But I think TR made a pretty good “assumption”.

      You didn’t compare it to a six core chip with less L3, which this very chart shows there are several of. That’s just making assumptions. You could say that more cores scaled, but why in the world tack on the part about the cache? I’m sorry, but I really do not agree with this, “piling on the cache = moar fastar” dogma that PC sites perpetuate, without ever actually proving. It is not hard to do, but no one bothers, which gives Intel too much power to market by numbers vs. reality. People buy things based on what sites like TR say, and I think it’s your responsibility to back up statements like that.

      I can state for a fact that more cache = better performance. Take my E8400 and compare it with the AMD Athlon II X2 250 that I just bought for my wife’s new system has the SAME clock speed and only 2 MB cache to the E8400’s 6MB. And sorry to say, but they DO NOT even compare! My E8400 blows that chip left and right except for the price of course. Even my E5200 with 2MB doesn’t compare, but that chip is a lot slower, but even OCing it to match the speed didn’t give the same numbers. So, I have to “assume” that the cache does make a difference.

      Way to forget the new 6, 8, and 12 core AMD CPUs, which should see availability at about the same time as these new Intel CPUs, and also emphasize improved power efficiency.

      I don’t know and TR doesn’t either, they said so: We’d best not get ahead of ourselves on that point, however.. We have looked at reports from other sites about this what a year ago and they showed a clear lead. And we all know that the server prices have ALWAYS been overpriced, but that is the name of the game when their is no competition. So, I’m rooting for AMD to produce some awesome chips.

      I’m not being one sided with Intel or even TR, but just stating some facts.

      • Damage
      • 10 years ago

      Since you’ve asked what the story is, let me address a couple of your points:

      -The reasons I didn’t mention the upcoming Magny-Cours parts from AMD are several. Although AMD has said they’re shipping, Magny-Cours-based Opterons haven’t been formally announced yet. Furthermore, Westmere-EP is a smaller chip than AMD’s Shanghai quad core, let alone the Istanbul six core. Two 45nm six cores together on a single package don’t strike me as entirely direct competition to Westmere-EP. Of course, AMD may price them that way. What happens there remains to be seen. My sense is that fullest versions of Magny-Cours, at least, will compete more directly with Nehalem-EX than Westmere-EP.

      No, I didn’t elucidate all of that thinking in this little post, but I didn’t intend to mislead in focusing on existing parts and this class of system. I was simply focusing on the most relevant comparison at present–and quite possibly in the future, as well.

      -My comments about performance scaling with which you take issue were directly in reference to the Gulftown performance results linked in the text. They were not intended to describe actual 2P systems and were simply being used as a point of reference. Properly understood, they are accurate and refer to actual test data.

      You will note that, after referring to that data and saying where it would seem to point, I stopped and I said we shouldn’t get ahead of ourselves, because we have plans to test the actual 2P solutions at issue. We *do* back up our performance talk with actual test data and have a long track record of doing so.

      I should say that we’re hoping our tests will include Magny-Cours processors as well as the rest. We have been working toward that goal and look forward to making it happen, if we can.

      • NeelyCam
      • 10 years ago

      I know that you, as an AMD fan, are a bit pissed off being on the losing side, but I wouldn’t mind if you give some credit where credit is due, and agree that Intel released an awesome product, instead of – again – bitching how expensive things are, and how much better AMD is.

      I’m so tired of reading that stuff from you time over time.

        • Game_boy
        • 10 years ago

        How can you say AMD is losing when you don’t know the performance/price/power of AMD’s 12 cores? At the same price and power envelope they could beat Intel’s 6 cores.

        I’ll concede Nehalem EX is faster, but i) it’s not out either, ii) it’s more expensive, both CPUs and as a platform and iii) we don’t know its power consumption.

          • NeelyCam
          • 10 years ago

          But AMD is losing, regardless of if we’re looking at current offerings or near future.

          Currently, these 6-core Xeons are the performance leaders – nobody questions that. In the near future, M-C might level the playing field a bit until Nehalem EX arrives.

          Trying to find the one metric where AMD’s offerings beats Intel’s is a bit is borderline pathetic.

          “Clock speed doesn’t matter – it’s the performance. Look at that P4, you don’t even need to pay for heating.”

          “Idle power doesn’t matter. Global warming isn’t real anyways”

          “Better performance? Ah, but AMD high-end Phenoms are clocked higher than Lynnfields.”

          “4.5GHz OC? Ooh ooh, but AMD has more cores! Real men use real cores! Multithreading is for kids.”

          “AMD is superior in $95-$102 range at performance/dollar*SQRT(watt_load)/(watt_idle)^(3/2)+size-of-the-cpu-cooler-in-sqmm”

          “AMD’s platform has no equal. AMD’s IGP plays Crysis at 14FPS!!”

          “I can reuse my four-year-old DDR2 sticks. I save at least $40! Speed doesn’t matter.”

          “Yeah, but with an Athlon I can unlock two cores and 4MB of cache!! (And I feel lucky)”

          “QPI & on-chip memory controller? Pathetic – they had to copy AMD! Faster, you say? It burns too much power! Low power consumption is what really matters!”

            • NeelyCam
            • 10 years ago

            Unsuccessful troll is unsuccessful

          • JumpingJack
          • 10 years ago

          You can get a pretty good idea of where MC will end up in 2P by looking at the 4P results for an 8439, and scaling for clock speed. Just comparing top bins, it looks like 6C/12T is indeed better than 12 real cores. Hard to tell, in some cases is will be close, in other cases not so close.

          AMD, though, will be well positioned on price/performance front. When you go balls out on your product and you land on the spot of the best you can do, then price needs to adjust from there to be competitive. MC will show notable wins over X5600 though, and in places where it likely really countes (VM for example)

          None of the arguing about value/performance changes the very real possibility that a 6 core Xeon DP out performing, in most cases, 12 cores of Opteron… it’s even more absurd when you compare die sizes, AMD will be making and selling a CPU with almost 3x the amount of silicon real estate just to come close to or match Intel’s latest offering.

          I found it odd that AMD embraced the MCP approach, considering the criticism they levied over the years when Intel lauched those type of products and given the fact that the topology of a glueless itnerconnect with IMC makes such an approach very difficult … they are effectively welding two sockets together. My guess is that this is a stop-gap (much like Dunnington was a stop gap in 4P for Intel) to hold out for Bulldozer.

            • NeelyCam
            • 10 years ago

            Um… Wow. I’m speechless. I have nothing to add – agreed on every point.

            Good post. +5

            • Game_boy
            • 10 years ago

            No, you can’t.

            MC has an IPC boost over Istanbul, and a single Magny-Cours will beat a single Istanbul. It also uses the same power at the wall at idle and load. This is all official.

            AMD was never against MCM, they were against MCM without a fast on-package connection (FSB went through the northbridge). AMD now has an HT link between the dies. When AMD’s John Fruehe initially said that to the press, the fine detail was lost.

            • NeelyCam
            • 10 years ago

            You guys are exhausting.

            Let’s just wait for the damn benchmark scores, m’kay?

        • kc77
        • 10 years ago

        Yes, he (OneArmedScissor) went a little overboard. However, there is a point to be made. Desktop vs. Server performance are horses of a different color. Not all workloads perform the same especially in the enterprise market. Small changes here and there can make huge differences which makes results not so easily qualified or quantified. It’s the main reason why you don’t see too many server reviews.

        Personally there won’t be much of a difference at 2S. However, at 4S and
        up, and with running DOM (O)’s with 64-bit kernels the AMD platform isn’t as outgunned as it is on the desktop. It’s interconnects are pretty good and quite scalable. It’s one of the main reasons people do purchase them for server but not for desktop. Overall it’s best to wait to see what the results are before making sweeping statements, one way or the other.

    • firestorm02
    • 10 years ago

    Wow, the E5620 just became the new sweet spot for 2P workstations. Any one want to buy 2 E5520’s, lol.

    • jokinin
    • 10 years ago

    Wake me up when the Xeon X5677 has a <200euro consumer versions with 65W TDP, or AMD has something competitive with that thing for that price.
    Meanwhile my Core 2 Duo E8400 is happily running everything I need quite well.

    • Krogoth
    • 10 years ago

    Intel has the DP market by the balls.

    Opterons cannot compete on a time is $$$$ basis. They are only are desirable if budget is a serious consideration or have pre-existing infrastructure.

      • Game_boy
      • 10 years ago

      Wait for Magny-Cours before saying that.

      12 cores at up to 2.3GHz; IPC boost over Istanbul; priced against Westmere 6 cores (75W ACP, 2.2GHz 12-core will be $989), and using similar power at idle and load to the 2.8GHz Istanbul (at the wall, not ACP).

      For ~$1000, would you take 12 Greyhound cores at 2.2GHz (plus whatever the IPC boost is) or 6 Westmere cores at 2.66GHz? Let’s assume similar power consumption and platform costs, and server workloads.

        • Krogoth
        • 10 years ago

        Magny-Cours are going to be server chips though. They aren’t going to catch up to Gulftowns in workstation related tasks. They might win in areas where threads matter since 12 real cores is still better than 6 cores with SMT.

          • Game_boy
          • 10 years ago

          The original comment referred to DP. Hence I was talking about servers.

    • Usacomp2k3
    • 10 years ago

    Is that a square root symbol used as a check mark? I like!

    • P5-133XL
    • 10 years ago

    It is my understanding that Intel will be offering 8/16 Xeons for the 4 socket crowd shortly. That’s when the Intel fans will rejoice!

      • mototime
      • 10 years ago

      You will see 8 cores on Mar. 30th. Nahelem EX

    • Meadows
    • 10 years ago

    New power meter, you say.
    Let us hope that ends the whining.

      • Krogoth
      • 10 years ago

      What kind of whining?

        • TravelMug
        • 10 years ago

        The kind that got George Ou banned from here the last time?

          • Krogoth
          • 10 years ago

          He was trying to be a clever troll.

          I doubt it was “whining”.

          The continued insistence that the Opteron platform consumed less power than obviously superior Bloomfield-DP platform and ad hominems towards the reviewer are red-lights for trolling.

    • MadManOriginal
    • 10 years ago

    Second post!!11

    Scott, any hints on whether 4c/8t variants will be releaed as desktop chips, or if the Xeons will work fine in desktop motherboards? It’s of no relevance to me but I know there are some enthusiasts who hoped for 32nm quad cores although at these prices maybe not so much.

      • Umbongo
      • 10 years ago

      You can use them in desktop boards that have the correct firmware, but as the performance appears to be no different to their 45nm predecessor at the same clock speed and it hasn’t been shown they overclock better it doesn’t seem worth it at those prices.

    • shank15217
    • 10 years ago

    Humm.. lets see, the cheapest 6 core processor one can buy is $958. Intel fan boys rejoice!

      • ClickClick5
      • 10 years ago

      And AMD’s (desktop wise) will cost…

        • Mr Bill
        • 10 years ago

        Around $200; unless you want it server certified?

      • djgandy
      • 10 years ago

      So you buy $2000 of Intel processors instead of buying 3 entire AMD machines?

        • BlackStar
        • 10 years ago

        Nobody was fired for buying IBM.

        • dpaus
        • 10 years ago

        Actually, right now, a 4P system with 6-core Istanbul Opterons would still give you a better $$/perf ratio (I think). But that will change when Intel releases these for 4P systems. Except, of course, you can get 8P systems for AMD. The fun never ends….

          • shank15217
          • 10 years ago

          Mangy-Cours will come out at the end of the month. That would provide competition for these processors. Nahelum EX is 4P and up so AMD has a good chance to compete in the 2P space.

            • loophole
            • 10 years ago

            Actually you can find 2 socket Nehalem-EX systems – IBM’s x3690 X5 is one such example. Admittedly this is the x3690’s base config and it can be scaled up to 4 sockets; but they do start at 2.

      • NeelyCam
      • 10 years ago

      You pay for performance and quality.

Pin It on Pinterest

Share This