Ups and Downs: Memory Timings Put to the Test

[pitsi]

After reading this <A HREF="http://www.tomshardware.com/motherboard/20040119/index.html" target="_new">review</A> I was wondering ... does anyone else feel as stupid as I do for buying those costly Corsair PC3200 Low Latency memory sticks, while I could settle for some cheap CL 2.5 module without taking a big hit in performance?

Or am I simply being paranoid here and the LL was the right choice? What do you think?

 

[endyen]

After reading that review, I had a hard time not throwing up. Glad to see the lads got cas to work on the A64 boards. Ofcourse somehow intel almighty still creamed amd. Yes the benches were mostly lame (ha) but a good control of mobo functionality was also evident.

 

[Vapor]

Yeah, it's weird that Intel still 'won,' especially with 4DIMMs (there was an article a few months back about the 875P chipset being hindered tremendously by using more than 2DIMMs). Oh well, it serves one good purpose: finding out whether or not it is worth the extra money to buy better RAM for whichever processor you plan on getting.

Maxtor disgraces the six letters that make Matrox.

 

[Frozen_Fallout]

Just wondering they why the hell does Tom, in the past, always tell you to get the lowest Cas possible as it is worth it for the extra cash that you will be spending. Now Things may have changed a bit since I read a review which stated that Cas was important but I don't think so. So my question is why in the past has Tom always wanted you to get extreamly low Latancy Ram (When I am building computers for people I have always went on the fact that the lower the latancy for ram the better because I read in some Review here on Toms about a year ago that you NEED low latancies.) Then why the hell is he now saying that it really doesn't matter. What didn't anyone look to see Timings ment anything before this review. If so I would like to see what they came up with.

-------------------------------------------------
Remember what your fighting for, Remember why you even started fighting, and Remember who you are

 

[Vapor]

I don't understand what you are saying...indecipherable inglish you have there.

Maxtor disgraces the six letters that make Matrox.

 

[Rachmaninoff]

There is another good article about memory timings here.

<A HREF="http://www.gamepc.com/labs/view_content.asp?id=ddr400myths&page=1" target="_new">http://www.gamepc.com/labs/view_content.asp?id=ddr400myths&page=1</A>

 

[Crashman]

There was a good article at Anandtech about the i875P chipset being SUPER ACCELERATED by using 4 double-sided DIMMs. The SLOWEST dual channel operation was with 2 single sided. 2 double sided was much better, and 4 double sided supreme. So I guess your take on it depends on your source of information, but Anandtech showed HUGE gains in performance using 4 double-sided DIMMs.

<font color=blue>Only a place as big as the internet could be home to a hero as big as Crashman!</font color=blue>
<font color=red>Only a place as big as the internet could be home to an ego as large as Crashman's!</font color=red>

 

[endyen]

So!!!! Thats why the different ram for the P4 setup. They sure know how to optimize for P4 at Tom's.

 

[Crashman]

Yes, they were well paid by their advertizers too.

<font color=blue>Only a place as big as the internet could be home to a hero as big as Crashman!</font color=blue>
<font color=red>Only a place as big as the internet could be home to an ego as large as Crashman's!</font color=red>

 

[InkSpot]

Yes and the P4 gains around 3-5% running at the fastest timings compared to 1-2% on AMD systems... (atleast for games).

 

[shadus]

No, not really. I like the corsair LL and in the machines I've built with it it perform quite well (Both axp and p4). The price difference generally between LL and C2 is negligiable and on par with the performance increase you see (a couple %). No one said it would boost your machine 20% or anything. However, there is a pretty nasty hit for using a single chip which is what alot of people tend to do. *shrug*

<A HREF="http://www.lochel.com/THGC/html/shadus.html" target="_new">Shadus</A>

 

[pitsi]

Actually I am not talking about the price difference between the LL and C2 but about the difference between the LL and C2.5. The way I saw it, the increase in performance at least on an AXP system like mine, is minimal when using the LL instead of C2.5.

TWINX1024-3200LL . . . $251 (2-3-2-6)
Corsair Value Select (Dual Pack) 184 Pin 1G(512MBx2) DDR PC-3200 . . . $153 (CAS 2.5)

ouch !!!

Btw, how come I can run just fine 2-2-2-5 on my system and the best THG could get was 2-3-2-6 ? That surprised me a bit.

 

[zak_mckraken]

So my question is why in the past has Tom always wanted you to get extreamly low Latancy Ram...

The answer is quite simple and is even included in the article. Older chipset offered better speed at lower latencies but newer chipset (and CPUs) are able to manage the memory better than before. Especially the Athlon64. In the past, memory modules with low latencies made a big difference in system performance but now the difference isn't worth the extra 100$!


<i><font color=blue>I take a whiskey drink. I take a chocolate drink. And when I need to pee, I use the kitchen sink! I sing a song that reminds me I'm a urinating guy!</font color=blue></i>

 

[endyen]

2-2-2-5 is reserved for Intel almighty. Du maybe they cant quite figure out how to set timings on amd boards?

 

[Vapor]

The one I'm referring to wasn't at Anandtech, nor was it as detailed as to show performance with concerns to double vs. single sided RAM. I do remember it showing that with 4DIMMs, though, that the i875P was slower than it was with only 2DIMMs. I've tried googling for it, no lucking finding it, maybe someone else knows/has bookmarked the one I'm talking about.

Maxtor disgraces the six letters that make Matrox.

 

[Ncogneto]

Until Tom's Verifies the memory timings externally take this with a grain of salt. Setting memory timing in bios does not always insure that your memory is actually running at those settings. Intel and Amd chipsets handle marginal memory differntly. when Intel chipsets detect marginal operation they use less agressive timings. When AMD chipsets detect marginal opertion they decrease the closck speed. Unless you verify the memory timing externally, it is a mistake to assume that what you set in the BIOS is what is actually being used.

It's not what they tell you, its what they don't tell you!

 

[taitertot]

Then why the heck would you even need the options in the BIOS if they dont actually mean thats the currents settings being applied???

-taitertot

If this post has attitude, seems to be overly aggressive, rude, distasteful to 99% of the users here, and shows a zealous defense of Intel... It’s probably Spud.

 

[addiarmadar]

Well I always knew that the timmings are irrelevant since the release of DDR but I still get them anyways. Every little bit of stability and no data loss is always a plus for me since they are generally a better quality of ram.

Barton 2500+ @ 2200mhz (10x220 vcore @ 1.8)
Asus A7N8X Dlx 440 FSB
1gb Geil GD pc3500 Dual Channel (2-3-3-6)
Segata 80gb SATA 8.5ms seek
ATI Radeon 9800 Pro(420/720)

 

[Ncogneto]

It all depends on how its implemented. In some cases it means that it will try to run at those settings unless the chipset decides the memory is not capable of that. also bear in mind that the new K8's have the memory controller on the processer itself, not a seperate north bridge. This configuration incurrs less of a latency period than the north bridge configuration of the P4 chipset. so while a piece of memory might be able to run at Cas 2-2-2-5 on a p4 with the additional latency of the northbridge, When placed in the K8 with less inherent latency it might not. Meaning chipset latency + memory latency = aggregate latency.

It's not what they tell you, its what they don't tell you!

 

[Crashman]

Don't get me wrong, I know the article you speak of! And I'm saying that a source I find more reliable showed just the opposite of the article you mentioned!

Yep, I read both articles. I read lots of articles. Keeps me on top of things. Besides, you have to get your information from multiple sources if you want to make a sound judgement.

<font color=blue>Only a place as big as the internet could be home to a hero as big as Crashman!</font color=blue>
<font color=red>Only a place as big as the internet could be home to an ego as large as Crashman's!</font color=red>

 

[Vapor]

Yeah, I haven't seen either test backed up by a second place...don't know where to trust (I remember the place was less reliable than anandtech--so I'll believe anandtech for now).

Maxtor disgraces the six letters that make Matrox.

 

[taitertot]

Show me a white paper to support this untill then I will live in my sheltered world of where when you set the BIOS for settings its really the settings.

-taitertot

If this post has attitude, seems to be overly aggressive, rude, distasteful to 99% of the users here, and shows a zealous defense of Intel... It’s probably Spud.

 

[juin]

http://www.lostcircuits.com/memory/2gb/

I dont like french test

 

[Ncogneto]

Juin beat me to it.

It's not what they tell you, its what they don't tell you!