Affordable Dual Core from AMD: Athlon 64 X2 3800+
by Anand Lal Shimpi on August 1, 2005 9:36 AM EST- Posted in
- CPUs
AMD's Efficiency Advantage?
Before we get to the actual barrage of performance tests, there is one issue that we have been wanting to tackle for quite some time now.AMD has often argued that their dual core architecture is inherently more efficient than Intel's, primarily because of their System Request Queue (SRQ). All core-to-core transfers occur via this queue instead of over a main, shared FSB, which is the case in the Pentium D.
Johan put AMD's architecture to the test by measuring the latency of cache-to-cache transfers in AMD's dual core chips vs. Intel's. The results were quite impressively in favor of AMD's architecture. Cache-to-cache transfers on Intel's dual core CPUs took over twice as long as on AMD's dual core CPUs, but at that time, we could not find any real world benefit to the architecture.
Armed with a bit more time, we went through all of our benchmarks and specifically focused on those that received the most performance gain from dual core architectures. Using these multithreaded and/or multitasking benchmarks, we looked at the performance improvements that the dual core processors offered over their single core counterparts. For AMD, making this comparison was easy; we took the Athlon 64 X2 3800+ and compared it to its single core equivalent, the Athlon 64 3200+. For Intel, the comparison is a bit more complicated. The inclusion of Hyper Threading makes the single-core to dual-core jump a little less impressive in some cases, thanks to the fact that virtually all single-core Pentium 4 processors these days can execute two threads simultaneously. Thus, for Intel, we had to look at HT enabled, dual core and dual core with HT enabled, all compared to single core performance to get a complete picture of Intel's multithreaded performance scaling.
Remember that all performance increases are with reference to a single core processor, and in the case of Intel, we are talking about a single core Pentium 4 with HT disabled. More specifically, we used a Pentium D 830 (3.0GHz) for the dual core tests and compared it to its single core counterpart - the Pentium 4 530 (3.0GHz).
First, we have our Winstone 2004 benchmark suite; we omitted Business Winstone 2004, since it shows virtually no performance boost from dual core CPUs and instead, focused on Multimedia Content Creation Winstone 2004 and the Multitasking Winstone tests.
It is also worth noting that although Hyper Threading improves performance with a single core, enabling HT on the dual core CPU actually yields lower overall performance than if we had left it off (+24.1% vs. +39.1%). Johan explained exactly why situations like this exist on the Pentium D in his "Quest for More Processing Power".
Next up is the SYSMark 2004 suite. In all but two of the tests, AMD scales slightly better than Intel when going to dual core. The scaling advantages aren't huge, but they are tangible in some of the tests.
Once again, while Hyper Threading itself tends to impress, HT + dual core gives us a mixed bag of results, sometimes outperforming dual core alone while falling behind other times.
Finally, we have our application-specific benchmarks; here, we have AMD scaling better than Intel in 3 out of the 5 tests, but then in the remaining 2, Intel scales better.
While the Athlon 64 X2 does have much better cache-to-cache transfer latencies than the Pentium D, it appears as if for the most part, those advantages don't surface in real-world desktop usage. That being the case, the Athlon 64 X2 3800+ must outperform the Pentium D 830 based on the performance advantages of its individual cores in order to win this battle, not based on any dual core architectural efficiencies. So, does it?
Facebook
Twitter
RSS
109 Comments
View All Comments
dougSF30 - Monday, August 1, 2005 - link
2.0 and 2.2 GHz parts with 512K L2 x 2 = 89W, whether "BV" (Manchester) OR "CD" (Toledo).2.2 GHz 1MB L2 x 2 and 2.4GHz 512K or 1MB L2 x 2 = 110W, whether "BV" (Manchester) OR "CD" (Toledo).
Once again, http://www.amdcompare.com/us-en/desktop/Default.as...">http://www.amdcompare.com/us-en/desktop/Default.as...
is useful. Select the X2 line.
yacoub - Monday, August 1, 2005 - link
p5:The Roxio VideoWave test in PCWorld’s WorldBench 5 suite completes 6 seconds quicker on the Pentium D 830 than it does on the Athlon 64 X2 3800+.
yacoub - Monday, August 1, 2005 - link
"In one of their strongest CPU paper-launches ever"HAHAHAHAH :)
dougSF30 - Monday, August 1, 2005 - link
"Strongest paper launches ever" ??? Face it, you guys blew it in your original X2 article, claiming there would be no retail availability until late this year. Turned out that in June anyone who wanted one could buy one from Newegg or Monarch.BTW, the 4200+ and 4600+ were Manchester (OPN core code: "BV", Rev E4) cores (147mm^2) when launched. Now there are also Toledo (OPN core code: "CD", Rev E6) versions, but they did not show up until recently.
Check the Desktop Processor Quick Reference Guide (and OPNs of parts sold since June from various vendors-- the 4200 and 4600 were "BV" Manchester parts):
http://www.amdcompare.com/us-en/desktop/">http://www.amdcompare.com/us-en/desktop/
(Select the X2 line)
masher - Monday, August 1, 2005 - link
Why no details on the testbed for each platform, specifically memory speeds used? It makes me wonder if the comparison used pricey low-latency ram for the X2 and bargain-barrel chips for the Pentium D..Also the article states "While AMD scales slightly worse than Intelin the MMCC Winstone and Multitasking 1 tests, AMD scales a lot better in the last two tests...". In one of those tests, AMD's "slightly worse" is 5.7%...whereas AMDS "a lot better" result on one of the other two was a measly 2.9%.
A 5.7% drop is "slightly worse", but a 2.9% increase is "a lot better"? What's funny is this obvious distortion was likely done subconsciously by the author, in his desire to boost his favorite.
When are you fanboys going to learn that processors are TOOLS. Use the best one for the job at the time...don't fall in love with them. If you were plumbers, you'd probably be masterbating each night with your pipe wrenches.
Houdani - Monday, August 1, 2005 - link
Eh? Where did you find the 5.7 and 2.9 numbers? Did the tables change since this comment was posted?For the numbers in question (table 1 on page 4) I'd rate the processors...
MMCC Winstone -- no advantage
Multitask 1 ---- Intel advantage (modest win) (+9.1%)
Multitask 2 ---- no advantage
Multitask 3 ---- AMD advantage (big win) (+29.3%)
Remove the MMCC and Multi_2 benches, and the comments are quite appropriate. Intel scales slightly better in Multi_1, while AMD scales a lot better in Multi_3.
justly - Monday, August 1, 2005 - link
I found (at least) some of the conclusions about “AMD's Efficiency Advantage?” bogus.The biggest error (and most obvious) is in the Winstone multitasking test 1, where the author commented that AMD is “significantly worse in the Multitasking 1 test”.
In that test duel core had no benefit because a single core AMD processor can handle that test without the need for a duel core. Looking at the actual numbers for that test on the “Multitasking Performance” page not only shows that the faster clocked single core AMD scale almost perfectly (based on clock speed) against its slower clocked duel core counterpart but it has the top performance in that test. Without looking at both the actual score and the percent of increase what is being proven has nothing to do with the “efficiency” of a duel core, but more to the point it shows how “inefficient” a single core is in that test. The 0% increase by moving from single core AMD to duel core AMD only goes to show how “efficient” the single core AMD is in that test scenario.
masher - Monday, August 1, 2005 - link
Aside from being near-intellible, this reply is far off the mark. No matter how efficient a single processor is, one expects two processors to be somewhat faster tha one. If this wasn't true, what would be the point of "duel-core" [sic] processors in the first place?There is no magic "100% efficiency" level for single a processor that cannot be exceeded. Even if there were, that efficiency would translate to the second core, making it faster as well. Two should be faster than one.
The efficiency being measured is not of the core itself, but of the scalability of multiple cores vs. a single one. In this particular case, there is a "100% efficient" standard-- linear scaling. If two cores run a benchmark twice as fast as one, then the dual-core implementation is 100% efficient...regardless of how efficient or ineffecient the actual cores themselves are.
Read the above carefully and consider it. It's not difficult to understand.
justly - Tuesday, August 2, 2005 - link
You can expect two processors to be faster than one, but that isn’t always true.What you failed to understand is the single core AMD processor is not limited by processing power in the Winstone multitasking test 1, so by adding the second core there is basically nothing for the second core to do, hence no performance increase.
Because of this it is inappropriate for the auther to gauge scalability for this test in a negative context as he did when he said “significantly worse”. In fact the worse a single core processor performs the more likely it will see a higher scaling when a second core is added because there is more work for the second core to process.
masher - Tuesday, August 2, 2005 - link
> "You can expect two processors to be faster than one, but that isn’t always true..."Of course it isn't. That's the purpose of this test...to determine the performance increase, if any.
> "...The single core AMD processor is not limited by processing power in the Winstone multitasking test 1, so by adding
> the second core there is basically nothing for the second core to do, hence no performance increase."
Oops, wrong again. You should have realized this simply from the fact that a higher clocked single-core A64 runs the benchmark faster. Claiming this test isn't even slightly CPU-bound and thus the poor result "isn't AMD's fault" is nonsense.
> "...the worse a single core processor performs the more likely it will see a higher scaling when a second core is added..."
Only for tests that are disk or memory-bound. And given the current differences in memory subsystems between AMD and Intel and the highly-integrated nature of AMD's memory controller, the "efficiency" of AMD's dual-core implementation is pretty much indistinguishable from the efficiency of their memory system.
So that just leaves highly disk-bound tests...which these tests were not. Even if they were, the results would still be valid. After all, if a single-core A64 runs as fast as possible due to disk constraints, then why should anyone spend more money on a dual-core chip?
Of course, one core of an X2 3800 is not nearly so fast as to run any of these benchmarks "as fast as theoretically possible". If multiple cores don't score higher, then its indicative of an underlying scaling issue, either with the chip or the test itself. Any other conclusion is nothing more than apologism and wishful thinking.