More Sockets, but Lower Performance?

When AMD briefed us on Quad FX, the performance focus was on heavy multitasking (AMD calls this "Megatasking") or very multi-threaded tests. We figured it was an innocent attempt to make sure we didn't run a bunch of single threaded benchmarks on Quad FX and proclaim it a failure. Given that the vast majority of our CPU test suite is multi-threaded to begin with, we didn't think there would be any problems showcasing where four cores is better than two, much like we did in our Kentsfield review.

However when running our SYSMark 2004SE tests we encountered a situation that didn't make total sense to us at first, and somewhat explained AMD's desire for us to strongly focus on megatasking/multithreaded tests. If we pulled one of the CPUs out of the Quad FX system, we actually got higher performance in SYSMark than with both CPUs in place. In other words, four cores was slower than two.

CPU SYSMark 2004SE Internet Content Creation Office Productivity
2 Sockets (4 cores) 261 373 182
1 Socket (2 cores) 288 393 211

You'll see that in some of the individual tests there is an advantage to having both CPUs installed, but in the vast majority of them performance goes down with four cores. It turns out that there are two explanations for the anomaly.

CPU Internet Content Creation 3D Creation 2D Creation Web Publication
2 Sockets (4 cores) 373 245 514 411
1 Socket (2 cores) 393 364 453 369

First, in Internet Content Creation SYSMark 2004SE, there appears to be an issue with having two physical CPUs in the system that results in the 3dsmax rendering test only spawning a single thread, lowering performance below that of a normal dual-core processor. This problem may be caused by a licensing violation within the benchmark where it is expecting to see one physical CPU with multiple cores and isn't prepared to deal with multiple CPUs. Regardless of the exact cause of the problem, it doesn't appear to be anything more than a benchmark issue. It's the performance in the Office Productivity suite that is far more worrisome because there is no issue with the benchmark that's causing the problem.

CPU Office Productivity Communication Document Creation Data Analysis
2 Sockets (4 cores) 182 171 259 137
1 Socket (2 cores) 211 187 285 176

The Office Productivity suite of SYSMark 2004SE wasn't the only situation where we saw lower performance on Quad FX than with a single dual core setup. 3D games seemed to suffer the most; take a look at what happens in our Oblivion and Half Life 2: Episode One tests:

CPU Oblivion - Bruma Oblivion - Dungeon Half Life 2: Episode One
2 Sockets (4 cores) 67.3 78.3 155.8
1 Socket (2 cores) 75.2 90.9 165.7

Once again, populate both sockets in the Quad FX system and performance goes down. The explanation for these anomalies lies in the result of one more benchmark, CPU-Z's memory latency test:

CPU CPU-Z Latency (8192KB, 128-byte)
2 Sockets (4 cores) 55.3 ns
1 Socket (2 cores) 43.3 ns

With both sockets populated, memory latency goes up by around 27% and thus in applications that are more latency sensitive and don't necessarily need all four cores, you get worse performance than with a single dual-core CPU. The added latency comes from the additional probing over the HT bus that's done for coherency whenever a memory request is made to see where the latest copy of the data resides.

It's a problem that will go away if you have a single quad-core CPU with one memory controller, but one that makes Quad FX a tougher pill to swallow compared to Intel's quad-core offerings.

How does a 3GHz Athlon 64 X2 Perform? Four cores, 1 Socket or Four cores, 2 Sockets?
Comments Locked

88 Comments

View All Comments

  • peternelson - Thursday, November 30, 2006 - link


    I image www.Scan.co.uk in the UK will stock both boards and cpus at realistic street pricing.

    And I am seriously thinking of getting one, to put lots of PCIe cards into, and for use WITH A NUMA AWARE OPERATING SYSTEM.

    Who knows, Asus may even release a bios upgrade for it!

    And also note that there is a similar Asus "Deluxe" board not just the WS workstation variant, so may be a little cheaper.

    My other options are a dual socket Opteron board (with expensive memory) or a Core2Quad, using Nvidia 680i chipset which gives less I/O capacity.
  • Anonymous Freak - Friday, December 1, 2006 - link

    Or, you could go for a dual socket Xeon system. It would cost near the same, and you'd have the option of two quad-core processors RIGHT NOW. (Rather than late next year.)

    Of course, then you run into more expensive (and more power-hungry) memory. But if you are actually looking at such a system, a Xeon might be a better fit.
  • mino - Friday, December 1, 2006 - link

    The problem IS the memory and I/O, these are the strenghts of this solution.

    Prices of cheap 4C WS solutions:

    HIGHEST PERFORMANCE 4C: -> systems have comparable performance (Vista/XPx64 in account)
    -------------------------------------------------
    ***Quad FX:
    1pc FX-74 $1000
    1pc MB $350
    4pcs 1G RAM $400
    1pc Fortron PSU $100
    0pcs 8-port SATA controller $0
    ----------------
    Total: $1750

    ***Core 2 Quad:
    1pc QX6700 $1000
    1pc MB $250
    4pcs 1G RAM $400
    1pc Fortron PSU $60
    1pc 8-port SATA controller $150
    ---------------
    Total: $1860

    BEST VALUE 4C: -> systems have comparable performance[except QX6700] (Vista/XPx64 in account)
    -------------------------------------------------
    ***Quad FX:
    1pc FX-70 $600
    1pc MB $350
    4pcs 1G RAM $400
    1pc Fortron PSU $100
    0pcs 8-port SATA controller $0
    ----------------
    Total: $1350

    ***Core 2 Quad:
    1pc QX6700 $1000
    1pc MB $250
    4pcs 1G RAM $400
    1pc Fortron PSU $60
    1pc 8-port SATA controller $150
    ---------------
    Total: $1860

    ***Opteron 2000:
    2pcs 2216 $1200
    1pc MB $400
    4pcs 1G RAM $600
    1pc Fortron PSU $100
    1pc 8-port SATA controller $150
    ---------------
    Total: $2450

    ***Xeon 5100:
    2pcs 5150 $1200
    1pc MB $400
    4pcs 1G RAM $700
    1pc Fortron PSU $100
    1pc 8-port SATA controller $150
    ---------------
    Total: $2550

    IDEAL WS/PC Solution: (best value proposition, upgrade possible when necessary)
    ----------------------------------------------
    1/2pcs FX-70 $300
    1pc MB $350
    2pcs 1G RAM $200
    1pc Fortron PSU $60
    0pcs 8-port SATA controller $0
    ---------------
    Total: $910

    The funny thing beeing, the biggest value of QuadFX is in the exceptional motherboard. Except those those 10 not routed USB ports, that board is a dream come true.

    And yes, those 12 SATA3G ports would be a blessing for my storage needs...
  • peternelson - Friday, December 1, 2006 - link


    Nice costings, I'm thinking the same here.

    That board has lots of I/O bandwidth.

    Since the Gbit ether ports support teaming (2 muxed together), I would have thought it worthwhile for Asus to implement all four rather than two.

    Four used separately would make nice routing between lans, and four used as two teamed pairs would give > 1Gbps performance in and out eg as a firewall, or as a server with redundant connection, or to talk to NAS boxes etc. Two extra nets wouldn't push the power consumption up by much.

    Hopefully Asus (or other) will produce a rev2 board with the extra ethernets.
  • Lonyo - Thursday, November 30, 2006 - link

    quote:

    You also get all twelve SATA ports, but there's only support for ten USB ports and two GigE ports. Obviously the number of people that will complain about not having all twenty USB ports and four GigE ports are limited, but with AMD expecting the L1N64-SLI WS to retail for around $370, we wanted all of the bells and whistles.


    There's support for 14 USB ports AFAIK, 4 on the back, and 5 headers on the mobo, which makes 14 I think, which still isn't 20, but it's more than enough, and it's more than 10 :P
  • mino - Friday, December 1, 2006 - link

    No, those 2 headers are for IEEE1394...
  • SLIM - Thursday, November 30, 2006 - link

    Is this just a prelim review only talking about the processors and the real review is still to come? I mean the only reason for this platform to exist is the 4 GPUs not the cpus. Let's see 4 8800gtx's with the FX74s playing at insane resolutions in games that intel simply can't fathom due to the lack of support for 4 gpus (at least as far as I know of).

    The 4 gpus are the point of this platform, not power consumption, not 3dsmax... GAMES WITH 4 GPUS! I don't mean to be too abrupt, but the boards got 4 graphics slots for a reason.
  • Spoelie - Friday, December 1, 2006 - link

    A single 8800gtx is already being bottlenecked by a core 2, let alone them in SLI.

    Quad SLI will not have any performance advantage over SLI, as the processing power to feed them is just not there.
  • PrinceGaz - Thursday, November 30, 2006 - link

    I dread to think what the power consumption of a system with the FX-74s and four 8800GTX cards would be. It would probably be close to 1KW, possibly even around 1100W or so. That would be insane!
  • JarredWalton - Thursday, November 30, 2006 - link

    Given that this "4x4" initiative is apparently going to use NVIDIA graphics cards, and NVIDIA makes both AMD and Intel chipsets, if they decide to do anything with quad G80 chips you can pretty much guarantee that it will be for both platforms. Thus, 4x4 is really 2x2 right now, whereas the 775 platform is 1x4 and Xeon is 2x4. In the future, AMD will have 2x4, and perhaps all the platforms will support some silly quad GPU configuration.

    Basically, quad SLI was all about the pissing contest. "Oh yeah, well I have for GPUs and you only have two!" Then there was the completely bogus marketing material that they sent out with quad SLI talking about how great the extra GPUs would be for accelerating physics calculations. I suppose it's possible that in the future such applications will become useful, but almost a year after the initial talk of NVIDIA physics acceleration and we have yet to see any actual demonstration of this capability. Anyway, this Quad FX is just the same thing as Quad SLI: potentially good marketing, but lackluster final performance and terrible heat and power requirements.

Log in

Don't have an account? Sign up now