High Speed DDR2: Buffalo and Crucial Deliver 1000+
by Wesley Fink on July 7, 2006 12:05 AM EST- Posted in
- Memory
Crucial PC2-8000 C5: Stock Memory Performance
While the Intel platform does not really allow for easily testing different memory speeds at the same CPU speed, our test Asus motherboard does provide a wide range of memory ratios that match available DDR2 memory. This allows end-users to select the memory ratio that matches their DDR2 memory speed if they wish, or to choose a speed value over or below the rated memory speed. For those reasons, we first tested all of the stock ratios at the fastest stable timings we could achieve at the given ratio. With ratios, processor speed remains constant at 3.46GHz and memory speed is varied by choosing different ratios.
Since the Crucial Ballistix is also rated at DDR2-1000, the 1000 speed was tested for performance. To achieve that speed the 889 memory speed was selected at the standard 1066 FSB (266 setting quad pumped). The FSB was then overclocked to 300 (1200 FSB) resulting in a memory speed of DDR2-1000. The Ballistix performed much better than the rated timings of 5-5-5-15. The Ballistix memory was completely stable at DDR2-1000 at 4-3-4-12 timings. This matches the best memory timings we have found with DDR2-1000 as the timings are the same as OCZ EL PC2-8000 XTC.
While the Crucial generally matched the wide memory speed range of OCZ EL, the required memory timings at most memory speeds were a bit slower than those required by OCZ. This resulted in performance that was slightly slower than OCZ at most speeds. While Crucial, Buffalo, and OCZ all use the same Micron memory chips, each manufacturer "speed-bins" the chips using their own internal procedures. They also choose the PCB to use with the memory chips, and program the SPD for best performance. These additional steps result in some differences in performance with a given chip such as you are seeing in this review. Apparently OCZ squeezes a bit more performance out of these chips, although the performance advantage for OCZ is small.
The Crucial PC2-8000 topped out at DDR2-1084, which is slightly below DDR2-1100 achieved with the Buffalo and OCZ. However, the slower memory timings required at the top of the performance range make the speed around 800 to 1000 the best-performing range with any of these three DDR2-1000 rated memories. In that range the Crucial Ballistix is very competitive.
At the rated speed of DDR2-1000 the CPU is running at 13x300 or a 1200FSB. The CPU overclock and higher bus speed certainly contribute to the higher performance at DDR2-1000. Comparing non-overclocked CPU and bus speeds, the memory performance curve is linear, generally improving as speed increases. Where faster memory timings are possible (lower memory timing settings) there is a comparable improvement in memory performance.
While the Intel platform does not really allow for easily testing different memory speeds at the same CPU speed, our test Asus motherboard does provide a wide range of memory ratios that match available DDR2 memory. This allows end-users to select the memory ratio that matches their DDR2 memory speed if they wish, or to choose a speed value over or below the rated memory speed. For those reasons, we first tested all of the stock ratios at the fastest stable timings we could achieve at the given ratio. With ratios, processor speed remains constant at 3.46GHz and memory speed is varied by choosing different ratios.
| Crucial (Stock Memory Ratios) - 2x1GB Double-Bank | ||||||||
| CPU Ratio at 3.47 GHz |
Memory Speed |
Best Timings (Voltage) |
Far Cry (fps) |
Sandra Unbuffered |
Sandra Buffered |
SuperPI 2M Mod 1.5 (seconds) |
HL2 Lost Coast (fps) |
Quake 4 (fps) |
| (4:3) | 400 DDR2 | 3-2-2-5 1.8V |
60.8 | INT 2820 FLT 2827 |
INT 5519 FLT 5548 |
88.6 | 82.6 | 68.8 |
| (1:1) | 533 DDR2 | 3-2-3-8 2.0V |
61.4 | INT 3494 FLT 3521 |
INT 6459 FLT 6467 |
86.2 | 87.9 | 73.4 |
| (4:5) | 667 DDR2 | 3-3-3-10 2.1V |
62.2 | INT 3961 FLT 4004 |
INT 6571 FLT 6565 |
85.5 | 88.7 | 75.7 |
| (2:3) | 800 DDR2 | 4-3-4-11 2.1V |
62.6 | INT 4211 FLT 4252 |
INT 6722 FLT 6702 |
84.9 | 89.5 | 76.5 |
| (1:2) | 1067 DDR2 | 4-4-5-14 2.35V |
65.7 | INT 4560 FLT 4632 |
INT 6836 FLT 6843 |
83.9 | 91.6 | 79.6 |
| (3:5) OC 300x13 |
1000 DDR2* *Rated Speed |
4-3-4-12 2.2V |
73.3 | INT 4989 FLT 5064 |
INT 7671 FLT 7658 |
74.9 | 101.3 | 87.8 |
| Highest Mem Speed (1:2) OC 271x13 |
1084 DDR2 | 5-4-5-15 2.35V |
66.2 | INT 4521 FLT 4587 |
INT 6905 FLT 6898 |
83 | 91.8 | 80.3 |
Since the Crucial Ballistix is also rated at DDR2-1000, the 1000 speed was tested for performance. To achieve that speed the 889 memory speed was selected at the standard 1066 FSB (266 setting quad pumped). The FSB was then overclocked to 300 (1200 FSB) resulting in a memory speed of DDR2-1000. The Ballistix performed much better than the rated timings of 5-5-5-15. The Ballistix memory was completely stable at DDR2-1000 at 4-3-4-12 timings. This matches the best memory timings we have found with DDR2-1000 as the timings are the same as OCZ EL PC2-8000 XTC.
While the Crucial generally matched the wide memory speed range of OCZ EL, the required memory timings at most memory speeds were a bit slower than those required by OCZ. This resulted in performance that was slightly slower than OCZ at most speeds. While Crucial, Buffalo, and OCZ all use the same Micron memory chips, each manufacturer "speed-bins" the chips using their own internal procedures. They also choose the PCB to use with the memory chips, and program the SPD for best performance. These additional steps result in some differences in performance with a given chip such as you are seeing in this review. Apparently OCZ squeezes a bit more performance out of these chips, although the performance advantage for OCZ is small.
The Crucial PC2-8000 topped out at DDR2-1084, which is slightly below DDR2-1100 achieved with the Buffalo and OCZ. However, the slower memory timings required at the top of the performance range make the speed around 800 to 1000 the best-performing range with any of these three DDR2-1000 rated memories. In that range the Crucial Ballistix is very competitive.
At the rated speed of DDR2-1000 the CPU is running at 13x300 or a 1200FSB. The CPU overclock and higher bus speed certainly contribute to the higher performance at DDR2-1000. Comparing non-overclocked CPU and bus speeds, the memory performance curve is linear, generally improving as speed increases. Where faster memory timings are possible (lower memory timing settings) there is a comparable improvement in memory performance.
Facebook
Twitter
RSS
24 Comments
View All Comments
JarredWalton - Friday, July 7, 2006 - link
We're working on more memory reviews, and Corsair's offering will be reviewed soon.araczynski - Monday, July 10, 2006 - link
didn't know that was an option, but that's ok, with the adblock solution i get to kill many birds with one stone.araczynski - Friday, July 7, 2006 - link
...i'm getting fed up with the damn intellitext ads all over the place, anyone know of a surefire way to block that crap?TIA
araczynski - Friday, July 7, 2006 - link
NM, found adblock for firefox, brilliant!JarredWalton - Friday, July 7, 2006 - link
Another option is to simply http://www.anandtech.com/siteinfo.aspx?intelli=y">turn it off in your site preferences. :)lopri - Friday, July 7, 2006 - link
Excellent review in that it shed light on DDR2 testing methodology on Intel platform. I should admit that I'd been quite ignorant about DDR2 scaling to this date. This article exaplains alot about the way Intel platform works and how dividers are used - in plain English. It seems, in a sense, it's easier to test memory on Intel platform, especially once Conroe arrives, with so many dividers available at any given CPU clock.Thanks for an excellent review. BTW, when can we expect the P5W-DH review and/or DFI 975X Infinity review?
lop
*nitpick1 : On page 5, there is a typo. ;)
Wesley Fink - Friday, July 7, 2006 - link
Thanks, Lopri. The typo is now corrected.Now that AMD has moved to DDR2 with AM2, memory dividers and memory speeds work exactly the same, with standard supported speeds to DDR2-800.
The P5W-DH review will post next week.
semo - Saturday, July 8, 2006 - link
why don't you make a dedicated error reporting section for each article?last page fifth paragraph:
shoudln't that be "amd platform"
Wesley Fink - Saturday, July 8, 2006 - link
We were talking about how Intel handles memory speed which is why AMD was not also mentioned. DDR-533 is DDR-533 on BOTH the Intel and AMD platform. However, the CPU clock speed on the AMD platform is 400, while the bus speed on the Intel is 1066. AMD double pumps 200 on the clock to achieve 400, while the latest Intel processors quad pump 266 to achieve 1066. The BASE speed is still 200 AMD and 266 Intel.1:1 on the AMD is therefore DDR2-400, but Hyper Transport and the on-chip memory controller on AM2 means the memory speed or ratios don't really matter. There is no penalty for running AM2 at ratios, but there is a very slight penalty for running Intel at other than 1:1 (DDR2-533) - in the range of 1% to 5% memory bandwidth reduction.
The AM2 has massive DDR2 memory bandwidth, but it is not memory-starved and really can't use the extra bandwidth in the current CPU design. It might matter more in future AMD designs.
semo - Sunday, July 9, 2006 - link
ok thanks i think i get it now. and it's a shame amd can't take advantage of the advances ddr2 has made (apart from the reduction in penalty going from ddr to ddr2)