This expensive ram, all expensive ram has a horrendous price to performance ratio and I think that's what you're asking. You're talking about 3-5% performance increases from the very best ram to the cheapest Kingston Value Ram.. It's amazing people readily spend 100-200% more money for 3-5% more performance. In fact I can't think of a poorer cost/benefit ratio of parts inside computers than in expensive ram, though SCSI, EE's, FX's come close.
Anyway check out toms, he did such a review last year.
It seems RAM is still the Achilles heel of the A64.
Do you guys plan to review the OCZ PC3700 Gold Rev 3 and OCZ EL PC4000 Gold Rev 2?
It might not bring anything to the table for A64 but from talking to OCZ and some minor early review leaks by them I get the impression that they think the new Hynix chips will perform much better then previous chips on A64.
Worth reviewing anyway. Especially since it includes the first PC4000 RAM that can run 2.5,3,3,6 at only 2.6 volt at stock speed.
#11 - To keep results as consistent as possible we have always set BIOS on the P4C800-E with Performance Mode on Auto and Performance Acceleration on Auto. We do not test memory or overclocks with either of these options enabled.
#14 - The manufacturers tell us heatspreaders cool memory, but I am not a believer. I find memory is often cooler without heatspreaders, but I have no objective proof for this claim. Heatspreaders do a good job of hiding the name of the memory chip maker. A high output heatsink fan nearby or a fan over the memory modules cools memory better than any heatspreader I have seen.
#13 - We did Doom 3 week with Doom 3 benches on every CPU and video card we could find. We have no plans of using Doom 3 for memory testing. You will see Doom 3 benchmarks in future video card tests at AT. We are also considering using Doom 3 in future motherboard tests.
Wesley, thank you for your response.
It seems then that anyone not about to overclock the ram much is better off saving some bucks going with value RAM, or getting larger capacity modules instead of lower-latency ones...
Due to lack of money I've already decided against the low-latency modules, I'd be looking forward to a possible Value RAM shootout at this site :)
Another question popped in my mind seeing comment #12... Do RAM heatspreaders actually help the cooling, when overvolting or not, or are they just for decoration?
#5 and #6- you must have read my mind as I considered asking exactly the same having read the article this morning. We get to see how all these premium sticks perform at various speeds but I'd be very interested to know how the cheap ones compare.
Wesley- although its possible to guesstimate the performance of various brand-name PC3200 ValueRAM / CAS3 or CAS2.5 modules, actually testing them is the only way to get a reliable figure. More importantly it should give some guide as to whether they are able to overclock to any useful extent, and at what timings and reasonable voltage increases. I don't think any of us would want to stick 3V into a stick of cheapo (but brand-name) PC3200 without a heat-speader if we wanted it to last for an extended period, but upping the voltage to 2.7 or even 2.75 isn't going to harm any module.
I'm sure a lot, maybe even the majority of AnandTech readers use budget or intermediate RAM for price/performance reasons rather than the premium sticks which are usually reviewed, so it would make sense including at least a few of the major brands.
The Asus P4C800-E is the motherboard you have been using in recent memory reviews.
Can you please tell me what bios setting you are using for the "Turbo" option under the JumperFree Configuration Menu and what setting is being used for the "Performance Acceleration Mode" option under the Advanced Chipset settings Menu.
#5 & #6 -
The info you are looking for is already in the charts but it takes a little digging. The Samsung PC3700 (DDR466) performed at 2.5-3-3-5 timings at DDR400 in our tests. Comparing the Samsung 3700 results at DDR400 to any of the DDR400 2-2-2 results should give you a very good idea of the performance impact of 2.5-3-3 timings compared to 2-2-2.
All of the memory reviews have a Performance page that lists timings at each speed, and the Samsung performance chart is at http://www.anandtech.com/memory/showdoc.aspx?i=199... We used 4x256MB modules in testing the Samsung, but we have already shown in past memory articles that 4SS performs the same as 2DS dimms. The performance penalty comes with 2SS modules on an Intel platform.
We are planning a Value RAM shootout in the future which should directly address your question.
Forgot to mention it was tested on an A64 rig. If you've got an AXP rig, lower latency might mean more. I should go test it out...
Anyway, great article Wes! It's good to see Geil come out near the top. They make some great memory and are very under-recognized sometimes.
I forget where I saw it, but I did see latency compared in a Doom 3 article. Low CAS 2 latency benefitted over CAS 2.5 by giving you roughly 2-3% more framerates. Which added up to about 1 frame. Since I can't quite remember where I found it, take this with a large boulder of salt :P
#1 - Corrected. We are missing our Web Editor who is taking a well-deserved vacation.
#2 - You make an interesting point. I was leaning toward Samsung TCCD chips, particularly based on the poorer overclocks on Athlon 64, but I agree there are some timings at certain speeds that don't really fit. Hynix also exhibits poorer A64 overclocks compared to Intel, so it is a possibility.
Nice work Wes. To bad we don't know what chips they have...I'm leaning twards hand picked Hynix that can do 2-2-2 @ 200 by the way it mirrors it's brothers at higher bandwidth...in addtion to it's bandwidth.
Just a couple of things you might want to fix, on the bottom of page 2 it has the title "OCZ PC3200 Platinum Revision 2 Specifications" which should be "Geil PC3200 Ultra X Memory Specifications". Also on page 4 at 557DDR speed the Quake3 fps is missing a decimal and is listed as 43486 instead of 434.86 fps. Other than that great review, awesome looking ram man!
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
19 Comments
Back to Article
xeoph - Wednesday, March 9, 2005 - link
ISSUE:pricing at the top m-400-512x2gl1gb3200dc
= misleading.
you are looking for the m-400-512x2glx1gb3200dc model.
notice the X that was missing before, the difference between Ultra X and Ultra Platinum...
Stupid mistake I almost made.
Zebo - Sunday, August 22, 2004 - link
#5,6,12This expensive ram, all expensive ram has a horrendous price to performance ratio and I think that's what you're asking. You're talking about 3-5% performance increases from the very best ram to the cheapest Kingston Value Ram.. It's amazing people readily spend 100-200% more money for 3-5% more performance. In fact I can't think of a poorer cost/benefit ratio of parts inside computers than in expensive ram, though SCSI, EE's, FX's come close.
Anyway check out toms, he did such a review last year.
Aelius - Friday, August 20, 2004 - link
It seems RAM is still the Achilles heel of the A64.Do you guys plan to review the OCZ PC3700 Gold Rev 3 and OCZ EL PC4000 Gold Rev 2?
It might not bring anything to the table for A64 but from talking to OCZ and some minor early review leaks by them I get the impression that they think the new Hynix chips will perform much better then previous chips on A64.
Worth reviewing anyway. Especially since it includes the first PC4000 RAM that can run 2.5,3,3,6 at only 2.6 volt at stock speed.
Wesley Fink - Friday, August 20, 2004 - link
#11 - To keep results as consistent as possible we have always set BIOS on the P4C800-E with Performance Mode on Auto and Performance Acceleration on Auto. We do not test memory or overclocks with either of these options enabled.#14 - The manufacturers tell us heatspreaders cool memory, but I am not a believer. I find memory is often cooler without heatspreaders, but I have no objective proof for this claim. Heatspreaders do a good job of hiding the name of the memory chip maker. A high output heatsink fan nearby or a fan over the memory modules cools memory better than any heatspreader I have seen.
Wesley Fink - Friday, August 20, 2004 - link
#13 - We did Doom 3 week with Doom 3 benches on every CPU and video card we could find. We have no plans of using Doom 3 for memory testing. You will see Doom 3 benchmarks in future video card tests at AT. We are also considering using Doom 3 in future motherboard tests.Visual - Friday, August 20, 2004 - link
Wesley, thank you for your response.It seems then that anyone not about to overclock the ram much is better off saving some bucks going with value RAM, or getting larger capacity modules instead of lower-latency ones...
Due to lack of money I've already decided against the low-latency modules, I'd be looking forward to a possible Value RAM shootout at this site :)
Another question popped in my mind seeing comment #12... Do RAM heatspreaders actually help the cooling, when overvolting or not, or are they just for decoration?
Thanks a bunch!
Visual
manno - Friday, August 20, 2004 - link
why does Anandtech refuse to post Doom 3 benches? I feel like I'm taking crazy pills! :)PrinceGaz - Friday, August 20, 2004 - link
#5 and #6- you must have read my mind as I considered asking exactly the same having read the article this morning. We get to see how all these premium sticks perform at various speeds but I'd be very interested to know how the cheap ones compare.Wesley- although its possible to guesstimate the performance of various brand-name PC3200 ValueRAM / CAS3 or CAS2.5 modules, actually testing them is the only way to get a reliable figure. More importantly it should give some guide as to whether they are able to overclock to any useful extent, and at what timings and reasonable voltage increases. I don't think any of us would want to stick 3V into a stick of cheapo (but brand-name) PC3200 without a heat-speader if we wanted it to last for an extended period, but upping the voltage to 2.7 or even 2.75 isn't going to harm any module.
I'm sure a lot, maybe even the majority of AnandTech readers use budget or intermediate RAM for price/performance reasons rather than the premium sticks which are usually reviewed, so it would make sense including at least a few of the major brands.
TRDACH - Friday, August 20, 2004 - link
Wesley,The Asus P4C800-E is the motherboard you have been using in recent memory reviews.
Can you please tell me what bios setting you are using for the "Turbo" option under the JumperFree Configuration Menu and what setting is being used for the "Performance Acceleration Mode" option under the Advanced Chipset settings Menu.
Wesley Fink - Friday, August 20, 2004 - link
#5 & #6 -The info you are looking for is already in the charts but it takes a little digging. The Samsung PC3700 (DDR466) performed at 2.5-3-3-5 timings at DDR400 in our tests. Comparing the Samsung 3700 results at DDR400 to any of the DDR400 2-2-2 results should give you a very good idea of the performance impact of 2.5-3-3 timings compared to 2-2-2.
All of the memory reviews have a Performance page that lists timings at each speed, and the Samsung performance chart is at http://www.anandtech.com/memory/showdoc.aspx?i=199... We used 4x256MB modules in testing the Samsung, but we have already shown in past memory articles that 4SS performs the same as 2DS dimms. The performance penalty comes with 2SS modules on an Intel platform.
We are planning a Value RAM shootout in the future which should directly address your question.
qquizz - Friday, August 20, 2004 - link
excellent read thanks Anandtech for the useful infoAvalon - Friday, August 20, 2004 - link
Forgot to mention it was tested on an A64 rig. If you've got an AXP rig, lower latency might mean more. I should go test it out...Anyway, great article Wes! It's good to see Geil come out near the top. They make some great memory and are very under-recognized sometimes.
Avalon - Friday, August 20, 2004 - link
I forget where I saw it, but I did see latency compared in a Doom 3 article. Low CAS 2 latency benefitted over CAS 2.5 by giving you roughly 2-3% more framerates. Which added up to about 1 frame. Since I can't quite remember where I found it, take this with a large boulder of salt :Pciwell - Friday, August 20, 2004 - link
I am wondering the same thing as #5. It would be very interesting to see a comparison between the Value RAMs and these.Anyways, great article.
Visual - Friday, August 20, 2004 - link
This isn't really in the goals of this article, but I'm wondering, is the extra-low latency worth it at DDR400?Could you include a slow(Value RAM) memory from Kingston, Corsair or other, running at CAS 2.5 and 3 for comparison?
I really wanna know if the extra $100 or so for a low-latency ram would give me noticeable difference.
pookie69 - Friday, August 20, 2004 - link
Mr Fink!!!! Another GREAT READ!!!! I so LOVE these memory articles of yours. SO informative and give much food for thought.Thanks and keep up the gd work! ;)
Wesley Fink - Friday, August 20, 2004 - link
#1 - Corrected. We are missing our Web Editor who is taking a well-deserved vacation.#2 - You make an interesting point. I was leaning toward Samsung TCCD chips, particularly based on the poorer overclocks on Athlon 64, but I agree there are some timings at certain speeds that don't really fit. Hynix also exhibits poorer A64 overclocks compared to Intel, so it is a possibility.
Zebo - Friday, August 20, 2004 - link
Nice work Wes. To bad we don't know what chips they have...I'm leaning twards hand picked Hynix that can do 2-2-2 @ 200 by the way it mirrors it's brothers at higher bandwidth...in addtion to it's bandwidth.KingofCamelot - Friday, August 20, 2004 - link
Just a couple of things you might want to fix, on the bottom of page 2 it has the title "OCZ PC3200 Platinum Revision 2 Specifications" which should be "Geil PC3200 Ultra X Memory Specifications". Also on page 4 at 557DDR speed the Quake3 fps is missing a decimal and is listed as 43486 instead of 434.86 fps. Other than that great review, awesome looking ram man!