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?


View All Comments

  • VooDooAddict - Saturday, December 2, 2006 - link

    It strikes me as odd that you haven't reviewed a Dual Xeon 5300 system.

    While the motherboards aren't geared toward the enthusiast. It would seem anyone who could really benefit from two Dual core Athlon FXs would get incredibly more bang for the $$ out of a Dual Xeon 5300 system giving them 8 cores on an established and sure to be supported platform.

    I only see two major drawbacks to the Dual Xeon as opposed to 4x4.

    You don't have access to SLI. That's unimportant though as people who want SLI would be better served with an Intel QX6700 on an NVIDIA 680i.

    You need FB-DIMMs. Considering the expense you are already at with any Dual CPU system, and the types of activities you'd need to be doing to get real benefit. ECC probably wouldn't be a bad thing to have.

    I'd love to see a review of a Xeon 5300 System with a 8800GTX.
  • yyrkoon - Friday, December 8, 2006 - link

    To the best of my knowledge ECC memory is usually slower compared to non ECC memory. Also, if I didn't read it here, I read it somewhere else, not only is FB-DIMM memory more expensive, it also does not perform as well compared to standard DDR2 memory.

    I think, for now, if you're wanting the best of both PC/server worlds, you would have to settle for a motherboard similar to Asus' Workstation, or Pro line. However, you would still be stuck with a single CPU, with multiple cores, but you would have SLI, PCI-X, and 2-3 PCI 2.3 slots :) As for memory quantity, anything more than 4 GB on a 32 bit OS is a waste (unless you use the right OS with the equivalent of the /PEA boot option, which I understand can cause some system instabilities)
  • yyrkoon - Friday, December 1, 2006 - link

    As far as I can remember, using multiple CPUs (as in socket, not core), has always shown diminishing returns. In fact, I think I remember reading something to that effect here on this web site (or maybe it was TH?). Regardless, having several Dual CPU boxen around, and have had the chance to play around with a few Quad CPU systems, I've seen this with my own two eyes.

    I did not start reading this article with the belief that AMDs solution would beat an Intel quad setup, but I did have hopes that AMDS system would be at least feasible. The way I see it, from these results, AMD should have never even bothered with this design, and would have better spent their time working on future technology.

    A few more mistakes like this, and pretty soon the only choice we're going to have is Intel, and it doesn't matter which company you buy from. We all should be hoping this never happens. My current system right now, is an AM2 system, and I have to say, I'm a bit miffed at AMD for leaving customers like myself out in the cold, when it comes to performance CPUs. So now, since I spent less than $200 for my current Motherboard / CPU, I think I'll Leave AMD in the dust (before they get the chance to do the same to me), and opt for an Intel system board, and CPU, for my next upgrade.
  • stepz - Sunday, December 3, 2006 - link


    he way I see it, from these results, AMD should have never even bothered with this design, and would have better spent their time working on future technology.

    To be fair, the engineering effort was close to zero. They sell almost exactly the same stuff as Opterons, only in shinier boxes and surrounded by a lot more nonsense buzzwords. This product is brought to you mainly by the marketing department.
  • fitten - Friday, December 1, 2006 - link


    As far as I can remember, using multiple CPUs (as in socket, not core), has always shown diminishing returns.

    Amdahl's Law.
  • T Rush - Thursday, November 30, 2006 - link

    I'm really not that thrilled to see a brand new Intel CPU barely squeak by a similar config from AMD that is using a CPU architecture that is over 4 years old

    I just gave Anand Lal Shimpi's article a closer inspection, we do see an advantage to the new Intel Core2 Q 65nm Kentsfield, but its not as large as I think it should be compared to the AMD64 90nm K8/Hammer dual cores times two...and however the pricing also doesn't reflect the large advantage you would expect to find with a newer more efficient design: as tested the two FX-74's are priced the same as the Extreme QX6700, and then next we see that two FX-72's under pricing the coming Quad Q6600 by a noticeable amount, and last we don't see a Quad Core2 to compete with the two FX-70's at the lowest $599 for the pair...I would expect the newer tech to not only be a much better performer, but also less expensive as well
    ....but to my surprise Shimpi almost caused me to fall out of my chair when I got to the last page and one of his biggest complaints was the higher price of the competition's motherboard!!!!, this to him seems to be a problem for AMD now, but the Intel systems he has tested before he was willing to with this comparison he has a problem with the total system cost(mobo+cpus) being higher do to the motherboard being over one third more expensive as tested; $370(which he later even 'rounds up' to $400) for the AMD vs. $260 for the Intel, looking at $110 difference
    ....oddly with even his own widely revered "Tremendous Value Through Overclocking" that he wrote, where the focus is in fact "Value"(not something that I would think is "as big" of an issue when looking at the highest priced/performing/featured/etc. systems that you could possibly buy today) he completely was not bothered by the fact that he was using a $250 motherboard for the "Value" Intel CPUs at as much as $120 difference more to the $130 AMD board he used where value was a factor in that earlier case ..what!!!
  • fitten - Friday, December 1, 2006 - link

    Well... the parts that "barely squeak by" are also running at only 80% of the clock frequency... 2.4GHz Q6600 compared to 3.0GHz AMD FX... and Intel's Core2 architecture seems to have plenty of headroom to go to higher clockspeeds on Intel's whim... 3.6GHz overclocks are pretty common with the Core2Duo parts (and the Quad parts are the same exact cores).

    There's no doubt that the Core2 architecture from Intel is the largest CPU performance bump in a number of years... previous years were almost stagnant as we saw 200MHz to 400MHz clock frequency increases on the same cores as the previous year. Core2 was a performance jump that was similar to the days of the Pentium3 timeframe.
  • T Rush - Thursday, November 30, 2006 - link

    was anyone really expecting AMD's "4X4" system using two CPUs built using the 4 or 5 year old K8/Hammer CPU architecture to out-perform Intel's latest release using its newest recent Core2 architecture?...I would hope not
    ...if anything Intel should have been able to pull off a much larger victory in price and performance....where the only clear major advantage that shows the large difference you would expect to see when comparing very new tech to much older was with the power consumption, which I wouldn't think is that great of concern with any system that is running four cores and most likely dual high end video cards(or more) in a "all out" highest performing/benchmarking gaming desktop system
    the over all win does go to the Intel on this dual dual core battle..but personally I would have liked to of seen a much broader margin before putting all my eggs in one basket as many will do
  • cryptonomicon - Friday, December 1, 2006 - link

    Thats an incredible amount of power... 450w!! Now that's what I need to keep my room warm this winter :D Reply
  • JarredWalton - Friday, December 1, 2006 - link

    You can overclock an E6300 even higher on a lot of P965 boards than on the $225 ASUS P5W DH. The motherboard chosen was used for consistency in benchmarking, not because it is the recommended "budget overclocking" motherboard for socket 775. As for QX6700 barely outperforming FX-74, there's a couple of things to keep in mind. First, you can't buy the AMD chips yet, whereas QX6700 is actually available (albeit a bit expensive). Second, the past indicates that 3.0 GHz is pretty clearly to limit of AMD's current chips on 90 nm.

    Meanwhile you have QX6700 systems shipping factory overclocked at 3.33 GHz and beyond. Sure, with that level of overclocking, the power draw might actually begin to be similar between the two platforms. However, a 25% overclock of a Core 2 Quad certainly isn't going to close the margins. And given that these are indigenous platforms we are talking about, overclocking is going to be far more common than not I would say.

    I'm not sure on Q6600 pricing either (since it's not available), though I've heard at least one rumor suggest the price has dropped to around $500. Guess we'll wait and see there as well. Basically, we are see Core 2 Quad beat AMD's latest and greatest in almost every test, without even pushing the clock speed limits. In seven months, we can evaluate again, and maybe AMD will have the lead.

    Basically, we should expect the margin of victory to be slightly lower with a single socket quad core going up against dual socket dual core then it was with Core 2 Duo going up against Athlon X2. The dual sockets give more memory bandwidth, so particularly when running a NUMA aware operating system, things should be pretty close when comparing FX-74 against QX6700. AMD was basically able to close the performance gap somewhat by throwing more power and hardware at the problem, sort of similar to what Intel did with Pentium D. It wasn't a good solution when Intel did it, and it's not a good solution now. It is "OK" at best.

Log in

Don't have an account? Sign up now