OCZ PC2-4200EB: Reaching for DDR2-800
by Wesley Fink on December 21, 2004 3:37 PM EST- Posted in
- Memory
tRAS and DDR2
OCZ rates the new PC2-4200EB at 3-2-2-8, with 8 a faster tRAS setting than we found in the DDR2 memory roundup. In the roundup we found a tRas setting of 10 was optimum for DDR2 running at 4-4-4 timings. To determine if the OCZ DDR2 ran best at a faster tRAS timing, we first ran Memtest86, a commonly used memory tool that measures memory bandwidth from a custom bootable OS.Memtest86 showed three distinct tiers of memory bandwidth performance. The range from tRas 2 to tRas 9 showed absolutely no difference in performance. 10 to 12 showed a slight performance drop and 13 to 15, a larger performance drop. Based on Memtest86 performance, a tRAS of 8 seemed reasonable.
| OCZ PC2-4200EB - 3-2-2-tRAS | |
| tRAS | Memtest86 Bandwidth |
| 2 | 3229 |
| 3 | 3229 |
| 4 | 3229 |
| 5 | 3229 |
| 6 | 3229 |
| 7 | 3229 |
| 8 | 3229 |
| 9 | 3229 |
| 10 | 3135 |
| 11 | 3135 |
| 12 | 3135 |
| 13 | 3050 |
| 14 | 3050 |
| 15 | 3050 |
We have seen reports on the web of claimed fantastic performance using tRAS 2 to 5 settings with this OCZ memory on the Asus P5AD2-E. This is the same board used in our new DDR2 memory test bed. So as a further check, performance tests were run varying tRAS from 2 to 15 using Everest (formerly Aida32). With Everest read/write/composite, performance was measured with 3-2-2- timings and a variable tRAS. Performance patterns in Everest were exactly the same as in Memtest86, with extremely small variations from 2 to 9 tRAS. An example is the tRAS 2 composite of 5663 compared to the tRAS 8 composite of 5686.
Our conclusion is tRAS from 2 to 9 makes no performance difference with the OCZ PC2-4200EB, so we ran all performance tests at 8. If you see enthusiasts claiming improved DDR2 performance with tRAS 2 to 5, ask for proof - as we can find no real performance difference in the 2 to 9 range with all other settings being the same.
Facebook
Twitter
RSS
22 Comments
View All Comments
MS - Tuesday, December 21, 2004 - link
JasonIt is certainly true that if you have a system backbone that is not capable of taking advantage of the peripherals it will put a little damper on the enthusiasm. On the other hand, don't blame the components for that.
One of the biggest issues is that the P4, no matter what you do cannot take advantage of latencies or bandwidth. The Prescott is a little better than the Northwood in that respect but it is still the bottleneck. Chances are that Intel finally will wake up and do something about this problem but maybe not. However, from the standpoint of a memory manufacturer, all we can do is try to provide THE very best solution and I firmly believe that we have done just that. Whether anybody wants it or not is a different story --- even though the latter could become our problem... :)
Wesley Fink - Tuesday, December 21, 2004 - link
#9 - We don't often mention exact price in a review because it is always changing and reviews are read and reread in the future. This is particularly true with memory. Right now, a 2x1GB kit is about $818. The 1GB kit with SS 512MB dimms is about $435.As we said in the review, this is expensive memory.
formulav8 - Tuesday, December 21, 2004 - link
This was most likely a HAND-Picked Dimm that OCZ sent to anandtech. Wait until there is real experiance with this memory to see if it is real or not.Jason
skunkbuster - Tuesday, December 21, 2004 - link
how much does this cost? i dont think i saw it mentioned in the reviewbcoupland - Tuesday, December 21, 2004 - link
What I find ironic with all this huge bandwidth on the 1066 fsb p4's, is that a S754 3700+ with 3.2 GB/s can still beat it in most tests, some more than others. Nice Ram, though.bigtoe36 - Tuesday, December 21, 2004 - link
GuysI haven't seen any other DDR2 clock as good as these dimms. Running 3-2-2- at DDR700+ is pretty impressive. Maybe we need to blame the boards or the cpu's for the lack of speed. Im sure if AMD moves to DDR2 running dimms at DDR700 3-2-2 would be pretty damned fast.
CBone - Tuesday, December 21, 2004 - link
How did the other sticks do in the 1066 FSB bench? I'm going to guess that if there was a difference, it was so slight as to be negligible.It seems that everyone is waiting for the great white hope in DDR2, but manufacturers are delivering the great white hype. So far it looks like ALL DDR2 performs and overclocks about the same so you should buy as cheap as you can and not bother getting the overpriced Corsair or OCZ.
formulav8 - Tuesday, December 21, 2004 - link
Was it mainly just for the oc or the lower timings? I guess if for ocing it is nice. But for timings it is worthless?? That is what I was trying to understand with my above post.Jason
formulav8 - Tuesday, December 21, 2004 - link
Unless I was looking at the wrong numbers, this memory is worthless. There was NOT EVEN 1% increase in performance? It increased the points in some benches but by a wopping 0.5%-0.9% on average???I am not sure why this review is so excited about this memory??
You get a much higher performance increase with DDR1 at low timings compared to high timings. Up to 4-5% increase in almost every bench.
Did I compare the wrong numbers or something??
Jason
Icehawk - Tuesday, December 21, 2004 - link
I'm very impressed that they got dual channel 1gb sticks working so well in DDR2 - wish they could do the same for DDR :( A64 + 2gb DC would be nice...