Bandwidth and Memory Scaling
One of the surprises in comparing DDR2 performance on AM2 and Core 2 Duo was the much better memory bandwidth found on the AM2 platform courtesy of the on-chip memory controller. Unfortunately, this did not translate into significant performance improvements compared to a similar AMD processor running DDR. At that point we concluded that Core 2 Duo was not particularly bandwidth sensitive, since it made very good use of the memory bandwidth available.
In our earlier review we were really comparing the DDR2 memory controller on AM2 to the 975X chipset memory controller, since Intel continues to place the memory controller in the chipset. We have speculated since then whether an improved memory controller in a socket 775 chipset would bring with it improved performance.
P965 brought very minor changes, mainly in the straps and overclocking ability of the memory. The NVIDIA 680i/670/650 actually shows decreased buffered bandwidth, but unbuffered bandwidth is about the same as P965. This reinforced the notion that memory bandwidth didn't matter much with Core 2.
To begin our investigation into DDR3 performance, we compared Standard or Buffered bandwidth on the P965 running DDR2, the new P35 running DDR2, and the new P35 running DDR3. As you can see the results are very interesting.
While the purpose of this review was to compare DDR3 and DDR2 performance, something completely different emerged from the memory bandwidth tests. Namely, the memory controller on the P35 is definitely an improvement over the P965 memory controller. This is evident whether the P35 is running DDR2 or DDR3 memory.
In those cases where we can run timings the same or close to the same, as in 800 memory speed performance, DDR2 and DDR3 results are virtually identical. By 1067 the current slow DDR2-1067 timings of 7-7-7-20 are performing just as well as DDR2 running at 6-6-6-15. The superior timings of DDR2-1067 at 4-4-3 still provides the best bandwidth at that speed. Of course, DDR3 is currently alone at the 1333 memory speed, but even with the current slow 9-9-9-25 timings it performs nearly as well as DDR2-1067 at 4-4-3 timings.
We normally also test memory with buffering schemes like MMX, SSE, SSE2, SSE3, etc. turned off. While these features do provide apparent improved bandwidth, we have found the unbuffered bandwidth to correlate better with real-world application performance. Unbuffered performance does not always follow the patterns of buffered memory performance.
Unbuffered results show the same basic pattern as buffered results in this case. Here DDR3 is clearly the best performer at the same slow timings at DDR2-800, with DDR2 on the P35 behind about 3% and DDR2 on P965 about 12% lower. DDR2 is still faster at the better timings available with current DDR2 memory.
In Standard/Buffered memory bandwidth, the P35 (Bearlake) chipset is providing a 16% to 18% improvement in memory bandwidth compared to the P965. This is a significant improvement. The Unbuffered improvement is smaller, in the range of 4% to 8%. These bandwidth improvements may or may not translate into improved system performance. We will examine that in the SuperPi and Gaming benchmarks.
One of the surprises in comparing DDR2 performance on AM2 and Core 2 Duo was the much better memory bandwidth found on the AM2 platform courtesy of the on-chip memory controller. Unfortunately, this did not translate into significant performance improvements compared to a similar AMD processor running DDR. At that point we concluded that Core 2 Duo was not particularly bandwidth sensitive, since it made very good use of the memory bandwidth available.
In our earlier review we were really comparing the DDR2 memory controller on AM2 to the 975X chipset memory controller, since Intel continues to place the memory controller in the chipset. We have speculated since then whether an improved memory controller in a socket 775 chipset would bring with it improved performance.
P965 brought very minor changes, mainly in the straps and overclocking ability of the memory. The NVIDIA 680i/670/650 actually shows decreased buffered bandwidth, but unbuffered bandwidth is about the same as P965. This reinforced the notion that memory bandwidth didn't matter much with Core 2.
To begin our investigation into DDR3 performance, we compared Standard or Buffered bandwidth on the P965 running DDR2, the new P35 running DDR2, and the new P35 running DDR3. As you can see the results are very interesting.
| Standard (Buffered) Sandra XI.SP2 Memory Bandwidth - 2.66GHz | |||
| Memory Speed | P965 ASUS P5B Dlx |
P35 DDR2 ASUS P5K Dlx |
P35 DDR3 ASUS P5K3 Dlx |
| DDR2-800 3-3-3-9 | 5531 | 6456 | - |
| DDR2-800 5/6-6-6-15 DDR3-800 6-6-6-15 |
5207 | 6143 | 6156 |
| DDR2-1067 4-4-3-11 | 5782 | 6811 | - |
| DDR2-1067 5/6-6-6-15 | 5712 | 6621 | - |
| DDR3-1067 7-7-7-20 | - | - | 6613 |
| DDR3-1333 9-9-9-25 | - | - | 6757 |
While the purpose of this review was to compare DDR3 and DDR2 performance, something completely different emerged from the memory bandwidth tests. Namely, the memory controller on the P35 is definitely an improvement over the P965 memory controller. This is evident whether the P35 is running DDR2 or DDR3 memory.
In those cases where we can run timings the same or close to the same, as in 800 memory speed performance, DDR2 and DDR3 results are virtually identical. By 1067 the current slow DDR2-1067 timings of 7-7-7-20 are performing just as well as DDR2 running at 6-6-6-15. The superior timings of DDR2-1067 at 4-4-3 still provides the best bandwidth at that speed. Of course, DDR3 is currently alone at the 1333 memory speed, but even with the current slow 9-9-9-25 timings it performs nearly as well as DDR2-1067 at 4-4-3 timings.
We normally also test memory with buffering schemes like MMX, SSE, SSE2, SSE3, etc. turned off. While these features do provide apparent improved bandwidth, we have found the unbuffered bandwidth to correlate better with real-world application performance. Unbuffered performance does not always follow the patterns of buffered memory performance.
| Unbuffered Sandra XI.SP2 Memory Bandwidth - 2.66GHz | |||
| Memory Speed | P965 ASUS P5B Dlx |
P35 DDR2 ASUS P5K Dlx |
P35 DDR3 ASUS P5K3 Dlx |
| DDR2-800 3-3-3-9 | 4226 | 4536 | - |
| DDR2-800 5/6-6-6-15 DDR3-800 6-6-6-15 |
3668 | 3975 | 4098 |
| DDR2-1067 4-4-3-11 | 4608 | 4926 | - |
| DDR2-1067 5/6-6-6-15 | 4389 | 4557 | - |
| DDR3-1067 7-7-7-20 | - | - | 4547 |
| DDR3-1333 9-9-9-25 | - | - | 4702 |
Unbuffered results show the same basic pattern as buffered results in this case. Here DDR3 is clearly the best performer at the same slow timings at DDR2-800, with DDR2 on the P35 behind about 3% and DDR2 on P965 about 12% lower. DDR2 is still faster at the better timings available with current DDR2 memory.
In Standard/Buffered memory bandwidth, the P35 (Bearlake) chipset is providing a 16% to 18% improvement in memory bandwidth compared to the P965. This is a significant improvement. The Unbuffered improvement is smaller, in the range of 4% to 8%. These bandwidth improvements may or may not translate into improved system performance. We will examine that in the SuperPi and Gaming benchmarks.
Facebook
Twitter
RSS
45 Comments
View All Comments
13Gigatons - Wednesday, May 16, 2007 - link
Suddenly it doesn't seem like a bad decision on AMD's part to hold off on their move to AM3 and DDR3 until 2008/2009. I really don't get why we need to change the memory technology so fast, with DDR2 finally dropping in price so fast.I'd rather have 4GB of DDR2 then 1GB of DDR3.
Sunrise089 - Wednesday, May 16, 2007 - link
Actually, a 2%-5% performance jump is very impressive from anything other than a CPU or GPU. Running a Raptor versus a 7200RPM drive, or a high-end motherboard versus a budget model, or a add-on sound card versus onboard audio all are choices many people make without any huge double digit performance gains in most applications.Thats said, the 2%-5% gain isn't from the memory standard (did you even read the article?) but from the new chipset. So these numbers have absolutely no bearing on AMD's choices.
Googer - Tuesday, May 15, 2007 - link
If you wanted to test bandwidth effects, why use a processor that is not very bandwidth dependant? Instead a bandwidth hungry LGA-775 Prescott should have been one the CPU's used in these DDR3 benchmarks. I'd like to see this article updated with a dual core netburst processor added.TA152H - Wednesday, May 16, 2007 - link
Are Prescotts even relevant anymore though? I mean, how many people are going to be perspecacious enough to buy a P35 based motherboard, and care about memory performance, and then go out and buy something as foul as a Prescott? It might make for an interesting data point, but it's a very little practical value.vailr - Tuesday, May 15, 2007 - link
Please include, in your forthcoming P35 board review: enabling SATA AHCI mode. Still remains puzzling, especially when a board uses a non-Raid Intel chipset, such as the ICH8. Gigabyte says on their web site, that on the GAS-965P-DS3, that AHCI should only be enabled when running Vista. Several other questions remain, such as: AHCI seems to work fine under WinXP when using the latest 1.17.17 JMicron drivers (single HD connected to a JMicron SATA port). Connecting the same HD to an Intel port, and AHCI won't work. Intel's offical AHCI drivers only seem to install when a Raid array is present.In summary: please include comments on any differences and/or improvements in AHCI support between the 965 v. P35 chipsets.
Comdrpopnfresh - Tuesday, May 15, 2007 - link
There was an article a few months ago saying that speed increases on ddr2 didn't really matter, the architecture for the memory was old enough that there was a decline in performance advantage as speeds increase. If ddr3 is basically the same as ddr2, wouldn't the same be expected? Does anyone have any idea when the timings will come down? The voltage is nothing to gawk at- it's the reason for the increased speed. On the gate level, the less space between high and low, he faster a gate can transition. I'm most interested to see ddr3 performance and bandwidth numbers w/ amd processors.R3MF - Tuesday, May 15, 2007 - link
i currently havean X2 system with PC3200 dual channel
and a C2D system with PC6400 dual channel
i definitely see a quad core 3.2GHz chip running PC12800 in my future.
hooray for technology!
DeepThought86 - Tuesday, May 15, 2007 - link
I wonder how long before Intel gets impatient and starts not-so-gently shoving DDR3 down people's throats long before the price or performance justify it?"Oh look, our chipsets for Nehalem don't support DDR2, woops you'll have to dump your DDR2 and get this spiffy new stuff. Look, it went from 2 to 3, it must be better!"
theprodigalrebel - Wednesday, May 16, 2007 - link
I remember reading an article somewhere where the interviewer asked an AMD person about Intel pushing for DDR3. The guy admitted that DDR3 is the way of the future - though not ready/relevant today - and only a company like Intel can drive that change.In his words, AMD's move to DDR2 came at the right time - in terms of price and advances in bandwidth/latency where DDR2 finally defeated the best DDR kits. He admitted that the move wouldn't have been possible unless Intel had moved the market in that direction over the past year or so.
Intel has driven changes from AGP to PCI-Express, IDE to SATA and DDR to DDR2. It seems forced at first - and it probably is - but you don't HAVE to be an early adopter. You had the 925 chipset introducing DDR2 and 915 boards supporting DDR. That is exactly what is happening here.
TA152H - Tuesday, May 15, 2007 - link
You need for Intel to push stuff down people's throats, or you'd never get these changes. No other company is in a position to, it's like IBM used to be almost 20 years ago.The price of the memory will go down as production goes up, which of course is driven by demand, which of course has to be driven by Intel. If not Intel, then who?
By offering a chipset that offers both, they are slowly starting the transition and the prices should get closer. At some point, supporting DDR2 is just a waste of chipset space and is costing people money that have no intention of ever using it, so you get rid of it. At that point it might cost a little more still, but that's the price you pay for transitioning to a better technology, and making that technology cost effective. I think they're extremely important to the industry for exactly that reason, not a malicious force.