First Look: AM2 DDR2 vs. 939 DDR Performance
by Wesley Fink on April 17, 2006 12:05 AM EST- Posted in
- CPUs
Comparing Memory Performance
When comparing many results, it is sometimes easier to see relative positioning of test results visually in a graph instead of interpreting numbers in a table. The full range of memory speeds were compared, first in Memory Benchmarks and Super Pi Calculation Performance
Synthetic Memory benchmarks show DDR400 latency and bandwidth roughly comparable to DDR2-533 3-3-3 performance in our tests. DDR2-667 3-3-3 is a bit higher, with best performance on this 4th spin of AM2 at DDR2-800 3-3-3.
The AM2 memory controller is finally meeting the promise of DDR2 with a dedicated on-processor memory controller. Latency is faster than DDR on AMD and quite a bit faster than latency on the Intel DDR2 memory controller in the motherboard chipset. Efficiency still lags considerably with a top Everest Read+Write Bandwidth of 7.2GB/s compared to a theoretical max of 12.8. This compares to the DDR400 Everest bandwidth that is close to the theoretical maximum of 6.4. As already said, this opens the door for further improvements in DDR2 memory performance as AM2 matures and shrinks to 65nm, more cores, and larger caches.
When comparing many results, it is sometimes easier to see relative positioning of test results visually in a graph instead of interpreting numbers in a table. The full range of memory speeds were compared, first in Memory Benchmarks and Super Pi Calculation Performance
Synthetic Memory benchmarks show DDR400 latency and bandwidth roughly comparable to DDR2-533 3-3-3 performance in our tests. DDR2-667 3-3-3 is a bit higher, with best performance on this 4th spin of AM2 at DDR2-800 3-3-3.
The AM2 memory controller is finally meeting the promise of DDR2 with a dedicated on-processor memory controller. Latency is faster than DDR on AMD and quite a bit faster than latency on the Intel DDR2 memory controller in the motherboard chipset. Efficiency still lags considerably with a top Everest Read+Write Bandwidth of 7.2GB/s compared to a theoretical max of 12.8. This compares to the DDR400 Everest bandwidth that is close to the theoretical maximum of 6.4. As already said, this opens the door for further improvements in DDR2 memory performance as AM2 matures and shrinks to 65nm, more cores, and larger caches.
Facebook
Twitter
RSS
37 Comments
View All Comments
peternelson - Saturday, April 15, 2006 - link
No I don't work for VIA or Transmeta but I do work in the IT Industry ;-)Efficeon chips were used in Orion Multisystems DT-12 and DS-96 cluster in a box computers and some notebooks.
Whilst they have low power, they do lack performance for some applications compared to the latest chips. But using VLIW based code-morphing they do indeed run x86 code.
Just call it a "good idea" rather than "unique" ;-)
Bladen - Friday, April 14, 2006 - link
Although DDR2 667 at 3-3-3 doesn't seem that common, all I can find is DDR2 667 4-4-4 or 5-5-5. Here in Australia anyway.Maybe when AM2 is released a rehash article featuring the higher latencies is in order.
Gary Key - Saturday, April 15, 2006 - link
In our recent experiences with Infineon based DDR2-667 modules rated at 4-4-4, the majority of these modules will run at DDR2-667 with 3-3-3 settings with a voltage setting around 2.2V. Your mileage will vary based upon supplier but going with one of the more performance oriented providers will usually result in the better timings.
We fully expect a wave of higher performance DDR2 modules to be launched in conjunction with the AM2 product. The majority of these new modules settling in around the DDR2-667 and DDR2-800 levels or above as we recently witnessed in our DDR2-1000 article -http://www.anandtech.com/memory/showdoc.aspx?i=273...">DDR2-1000 goes Higher.....
AnandThenMan - Friday, April 14, 2006 - link
It will be interesting to come back to this statement after AM2 and Conroe are out in the wild to see how accurate it was.
Interesting article, but nothing very surprising to me. The Athlon64 core is pretty much at its computational limit at a given clock, feeding it with more memory bandwidth does little. Which means that early adopters of the AM2 platform will get 939 performance with an updgrade path, which is not too shabby.
Conroe better live up to expectations though...
Wesley Fink - Saturday, April 15, 2006 - link
It would be a very pleasant surprise if AMD has us all in the dark and launches a Conroe competitive part or a Conroe-killer. Competition is good for buyers, especially when performance is very close. The close performance results in lower prices, as we are now seeing in the ATI/nVidia video cards from the most recent generation.However, we have to evaluate things with the best information we have available on time to Fab, launch dates, and the available revs that have been provided to AMD partners to design their companion products for the AM2 launch. There is always room for an unexpected surprise, but it looks less likely the closer we get to 6/6/06.
Viditor - Saturday, April 15, 2006 - link
I have no doubt that AM2 is a very weak upgrade...management at AMD said as much in their recent conference call.The only critique I have would be of the line
You should have added the word "desktop" before marketplace as the server marketplace should still be solidly AMD, and we have yet to see what will happen in the mobile space...JMHO
Wesley Fink - Saturday, April 15, 2006 - link
A very fair comment. I added the "desktop" qualifier since it makes sense.