A Closer Look at Latency and Scaling
As was explained in the Core 2 Duo launch review, Core 2 Duo has not physically added a memory controller on the processor. The memory controller is still part of the motherboard chipset that drives Core 2 Duo. Intel added features that perform intelligent look-aheads on the memory controller to behave like lower latency. As you saw on pages 2 and 3, ScienceMark 2.0 shows the "intelligent look-aheads" in Core 2 Duo to be extremely effective, with Core 2 Duo memory now exhibiting lower apparent latency than AM2. However, not all latency benchmarks show the same results. Everest from Lavalys shows latency improvements in the new CPU revisions, but it shows Latency more as we would expect in evaluating Conroe. For that reason, our detailed benchmarks for latency will use both Everest 1.51.195, which fully supports the Core 2 Duo processor, and ScienceMark 2.0.
Latency, or how fast memory is accessed, is not a static measurement. It varies with memory speed and generally improves (goes down) as memory speed increases. To better understand what is happening with memory accesses we first looked at Sciencemark 2.0 Latency on both AM2 and Conroe.
ScienceMark shows Conroe Latency with a 45ns to 61ns lead at DDR2-400. Latency continues to decrease as memory speed increases with Core 2 Duo, reaching a value of about 30ns at DDR2-1067. The Trend line for AM2 is steeper than Core 2 Duo, increasing at a rapid rate until latency is virtually the same at DDR2-800.
It is very interesting that ScienceMark shows lower latency on Core 2 Duo than AM2, since we all know the on-chip AM2 controller has to be faster. We thought perhaps it was because all of the tested memory accesses could be contained in the shared 4MB cache of Core 2 Duo, but Alex Goodrich,one of the authors of ScienceMark, states that Version 2 is designed to test up to 16MB of memory, forseeing the day of larger caches. In addition he states the Core 2 duo prefetcher is clever enough to pick up all the patterns ScienceMark uses to "fool" hardware prefetchers. ScienceMark plans a revision with an algoritm that is harder to fool, but Alex commented that Conroe fooling their benchmark was "in itself a great indicator of performance".
Everest uses a different algorithm for measuring Latency, and it shows the on-chip AM2 DDR2 controller in the lead at all memory speeds, with Latency almost the same at the Core 2 Duo memory speed range of DDR2-400 to DDR2-533. However, the Everest trend lines are similar to those in ScienceMark, in that AM2 latency improves at a steeper rate than Core 2 Duo as memory speed increases.
The point to the Latency discussion is that, as expected, AMD has much more opportunity for performance improvement with memory speed increases in AM2. Intel will eventually reach the point, if the lines were extended, where they would have to move to an on-chip memory controller to further improve latency. This is not to take anything away from Intel's intelligent design on Core 2 Duo. They have found a solution that fixes a performance issue without requiring an on-chip controller - for now.
As was explained in the Core 2 Duo launch review, Core 2 Duo has not physically added a memory controller on the processor. The memory controller is still part of the motherboard chipset that drives Core 2 Duo. Intel added features that perform intelligent look-aheads on the memory controller to behave like lower latency. As you saw on pages 2 and 3, ScienceMark 2.0 shows the "intelligent look-aheads" in Core 2 Duo to be extremely effective, with Core 2 Duo memory now exhibiting lower apparent latency than AM2. However, not all latency benchmarks show the same results. Everest from Lavalys shows latency improvements in the new CPU revisions, but it shows Latency more as we would expect in evaluating Conroe. For that reason, our detailed benchmarks for latency will use both Everest 1.51.195, which fully supports the Core 2 Duo processor, and ScienceMark 2.0.
Latency, or how fast memory is accessed, is not a static measurement. It varies with memory speed and generally improves (goes down) as memory speed increases. To better understand what is happening with memory accesses we first looked at Sciencemark 2.0 Latency on both AM2 and Conroe.
ScienceMark shows Conroe Latency with a 45ns to 61ns lead at DDR2-400. Latency continues to decrease as memory speed increases with Core 2 Duo, reaching a value of about 30ns at DDR2-1067. The Trend line for AM2 is steeper than Core 2 Duo, increasing at a rapid rate until latency is virtually the same at DDR2-800.
It is very interesting that ScienceMark shows lower latency on Core 2 Duo than AM2, since we all know the on-chip AM2 controller has to be faster. We thought perhaps it was because all of the tested memory accesses could be contained in the shared 4MB cache of Core 2 Duo, but Alex Goodrich,one of the authors of ScienceMark, states that Version 2 is designed to test up to 16MB of memory, forseeing the day of larger caches. In addition he states the Core 2 duo prefetcher is clever enough to pick up all the patterns ScienceMark uses to "fool" hardware prefetchers. ScienceMark plans a revision with an algoritm that is harder to fool, but Alex commented that Conroe fooling their benchmark was "in itself a great indicator of performance".
Everest uses a different algorithm for measuring Latency, and it shows the on-chip AM2 DDR2 controller in the lead at all memory speeds, with Latency almost the same at the Core 2 Duo memory speed range of DDR2-400 to DDR2-533. However, the Everest trend lines are similar to those in ScienceMark, in that AM2 latency improves at a steeper rate than Core 2 Duo as memory speed increases.
The point to the Latency discussion is that, as expected, AMD has much more opportunity for performance improvement with memory speed increases in AM2. Intel will eventually reach the point, if the lines were extended, where they would have to move to an on-chip memory controller to further improve latency. This is not to take anything away from Intel's intelligent design on Core 2 Duo. They have found a solution that fixes a performance issue without requiring an on-chip controller - for now.
Facebook
Twitter
RSS
118 Comments
View All Comments
drebo - Tuesday, July 25, 2006 - link
Excuse me, but the E6400 wasn't even IN the benchmarks Anandtech ran, but by extrapolation(considering the E6700 did not beat out the 5000+ by a large margin), it cannot perform better than the 5000+.And, in the majority of benchmarks, such as the ones http://www.anandtech.com/cpuchipsets/showdoc.aspx?...">here, which are benchmarks that have traditionally favored Intel processors in the past, the AMD chips do not fall as you paint them. They fall as I have painted them. The 4200+ outperforms the E6300 in ALL of them...and costs LESS. The 5000+ matches performance in 2/3 of them with the E6600. And costs LESS.
Sorry, but this isn't about which company is better or provides a better product. This is about which company provides a better product at a better price. We all know the X6800 is the best desktop processor available. NO ONE is disputing that. The dispute is that review sites are proclaiming that the Core 2 Duos are god's gift to Intel and that Athlon64s are completely worthless now, which is NOT the case.
In the $1k processor range, sure, Core 2 Duo is the king...but AMD doesn't have a processor in that range anymore. There is no competition to the X6800 anymore. At every other price point, however, the Athlon64 X2 processors are extremely competative.
coldpower27 - Tuesday, July 25, 2006 - link
Take a look at these benchmarks show the E6400 in them, and for the most part it beats out the 4600+http://xbitlabs.com/articles/cpu/display/core2duo-...">http://xbitlabs.com/articles/cpu/display/core2duo-...
coldpower27 - Tuesday, July 25, 2006 - link
Oh please, you looked at 1 page of benchmark to make your conclusion??Did you just look at 3D Rendering to draq your conclusions?
What about the encoding and gaming performance? Did you look at those?
IntelUser2000 - Wednesday, July 26, 2006 - link
It is obvious you haven't seen both of my posts fully. Please do. Core 2 Duo outperforms.
This is type of performance increase people would have been crazy about, but somehow its different. Maybe its conspiracy to kill Intel??
coldpower27 - Wednesday, July 26, 2006 - link
IntelUser2000.Chill. I am reply to drebo in this reply not you. Anandtech comment posting system is just strange.
IntelUser2000 - Tuesday, July 25, 2006 - link
Please stop being an idiot. If you see Anandtech's benchmarks you can clearly see that E6600 is FASTER overall than FX-62. That's why the saying of $316 FX-62 came about.
For those that are in denial about the FACT that E6600 with $316 price beats FX-62
Application Performance using SYSMark 2004 SE
Average performance: 18.6% in favor of E6600
Application Performance using PC WorldBench 5
1.4% advantage of Core 2 Duo E6600
Application Performance using Winstone 2004
2.9% advantage of FX-62, E6600 equal or faster than 5000+
3D Rendering Performance using 3dsmax 7 & CineBench 9.5
2.7% advantage of FX-62, E6600 faster than 5000+
Encoding Performance using DivX 6.1, WME9, Quicktime (H.264) & iTunes
6.0% advantage of E6600
Overall Gaming Performance
11.6% advantage of E6600
Total Performance increase of E6600 over Athlon FX-62
Core 2 Duo E6600 is 9.9% faster overall than Athlon FX-62!!
IntelUser2000 - Tuesday, July 25, 2006 - link
Normally, the kind of performance advantage E6600 gives over the FX-62 would have meant people would have flocked over to E6600 even if it was previous high-end mainstream price for Intel processors, which is $637.Stop spreading false info people(anti-intel, and i am saying lightly), cause not only your beloved Athlon 64 X2 5000+ slower than E6600, so is AMD's top end offering, the FX-62!!!
I remember people saying Core Duo was worth it(yes the Yonah). Based on the logic people are spreading for Core 2 Duo argument, Core Duo wasn't worth it at all. $637 for the top end Core Duo(Yonah), and being barely faster per clock than FX-62 when clocked outrageously high(30% over stock, while FX-62 is at stock), plus expensive motherboards that are hard to find.
If Core Duo was worth it at all, well... Core 2 Duo is like a dream.
Accord99 - Tuesday, July 25, 2006 - link
And the E6300 does better in other tests, such as the Sysmark, video encoding and games. And the E6600 doesn't compete with the 5000+, it competes with the FX-64.drebo - Tuesday, July 25, 2006 - link
I don't know what benchmarks you're looking at, but they sure as hell aren't posted on any review site I've ever seen.I never expected my opinion that AMD is not dead and Intel is not master of all to be popular, but I atleast expected some constructive responses. Yes, Intel has the best performing processor. We know this. No one cares. I'm sure AMD could spin a 3.2ghz FX-70 or something and sell it for $2000. But why bother, for 2% of the population to use? The vast majority of computer users are looking for performance-per-dollar, and there is no difference between Athlon64 X2 processors and Core 2 Duo processors in that ratio...except, of course, motherboard...which tends to cost near to twice as much on the Core 2 Duo side.
Yes, Intel released a competative platform. We're all very happy with them, and once they're available, it'll be great...but they're hardly a nail in the coffin for AMD. AMD still holds the superior chips for single-core, and performance-per-dollar is just as good as Core 2 Duo, and I can tell you from personal experience, that's what matters.
Go go objectivity!
Accord99 - Tuesday, July 25, 2006 - link
I'm just looking at all the Anandtech benchmarks. It's quite clear the E6300 is faster than the 4200+ overall, while the E6600 is faster than the FX-62 overall.So performance/dollar is higher for Conroe, the current motherboards are more expensive but that can be made up from purchasing less high-end memory. And Conroe uses less power.