Is CAS2 memory really worth the premium?

I

It_The_Cow

Gawd
Joined
Aug 10, 2001
Messages
882
Does 2-2-2-5 really have a significant difference from something like 3-3-3-7? I'm in need of two 512MB sticks that wil run at least 200MHz, and I'm wondering if Mushkin's premium price is really worth it. Corsair's Value Select CAS2.5 DDR400 is about half of Mushkin's price. What kind of gains will a lower timing result in?

Edit: The memory will be complementing and Asus A7N8X and 1800+ XP that wil be overclocked to about 12*200
 
I don't think it makes a big difference. I tested my CAS2 CorsairXMS against my Kingston PC4000 which has CAS3 timings. I tested them in the same board, both at 400Mhz hard-coded to their rated timings. My memory scores weren't real different. I'm sure others will disagree, but those are my personal findings.
 
Originally posted by djnes
I don't think it makes a big difference. I tested my CAS2 CorsairXMS against my Kingston PC4000 which has CAS3 timings. I tested them in the same board, both at 400Mhz hard-coded to their rated timings. My memory scores weren't real different. I'm sure others will disagree, but those are my personal findings.
Click to expand...

so the difference of cas2 and cas3 at 400mhz isnt noticeable to you at all?
 
Yes it is faster however only you can decide wether the price perfomance ratio is worth it. I think so but I can afford to think so

PS, In the context of overclocking the relaxed timings is offset by higher bus speeds and subsequently more memory bandwidth so you get a push there.

All things being equal . lower latency adds performance.
 
Ras-to-cas delay matters most for performance, second to command rate but consequently most memory chips have problems with this below 3 at over 200MHz. WinBond BH-5, BH-6, and in some cases CH-5/6 are known for their ability to run tRCD2 at high speed (with lots of juice :D).

Sometimes you can't go with CL or stock timings alone. For example, my TwinMos PC3200 stick w/WinBond BH-5 is rated at CL 2.5 but it does 2-2-2-5/6 @ 208MHz w/2.6v (default). Do as much research on the sticks you're interested in as possible, you can't go wrong with brands such as Kingston Hyper-X, Corsair XMS, Mushkin, KingMax, Geil & OCZ (these last two seem to be mixed bags of results though).

When just looking at a single timing factor, you may not notice the performance increase, but collectively they can make a large difference in performance. Also, stock timings are just tested and confirmed working by the manufacturer. Usually they will run tighter even at a higher speed but may need more voltage. 2.9v and below are safe to run at extended lengths, but Winbond chips love voltage, a lot of folks run them at 3.2v for extended periods. Happy hunting :cool:
 
Originally posted by Jason711
so the difference of cas2 and cas3 at 400mhz isnt noticeable to you at all?
Click to expand...

It was noticeable if I ran benchmarks and compared scores. The difference wasn't very big. However in actual PC usage and gaming, I couldn't tell a difference.
 
Originally posted by djnes
It was noticeable if I ran benchmarks and compared scores. The difference wasn't very big. However in actual PC usage and gaming, I couldn't tell a difference.
Click to expand...

cool.. thats useful info.
 
Originally posted by PliotronX
Ras-to-cas delay matters most for performance, second to command rate but consequently most memory chips have problems with this below 3 at over 200MHz. WinBond BH-5, BH-6, and in some cases CH-5/6 are known for their ability to run tRCD2 at high speed (with lots of juice :D).

Sometimes you can't go with CL or stock timings alone. For example, my TwinMos PC3200 stick w/WinBond BH-5 is rated at CL 2.5 but it does 2-2-2-5/6 @ 208MHz w/2.6v (default). Do as much research on the sticks you're interested in as possible, you can't go wrong with brands such as Kingston Hyper-X, Corsair XMS, Mushkin, KingMax, Geil & OCZ (these last two seem to be mixed bags of results though).

When just looking at a single timing factor, you may not notice the performance increase, but collectively they can make a large difference in performance. Also, stock timings are just tested and confirmed working by the manufacturer. Usually they will run tighter even at a higher speed but may need more voltage. 2.9v and below are safe to run at extended lengths, but Winbond chips love voltage, a lot of folks run them at 3.2v for extended periods. Happy hunting :cool:
Click to expand...

Very Concise............You have a talent
 
For an AMD setup ... they matter a lot ... not for P4's because of their high FSB ...
 
I get better benchmark results at 2-2-2-6 than I did at 2-3-3-7 at the same FSB(231). During game play and normal use, I can't notice any difference.
 
Originally posted by primea
For an AMD setup ... they matter a lot ... not for P4's because of their high FSB ...
Click to expand...
Actually, I find that in terms what I can actually NOTICE without benchmarks, CAS latency doesn't make much of a difference at all on either Intel or AMD platforms.

Of course, RAM that can handle lower latencies than other RAM at a given clock speed will probably be able to overclock higher if you relax said timings. ;)
 
Tight timings are almost like a minor overclock in a way. If you bump up your CPU 100 MHz you may not feel a great deal of difference but you could probably see a slight gain in benchmark numbers. Same with memory timings. It comes into play more if you are stuggling to get performance in a game or application. When I had a system that was stuggling to get consistent 30 FPS in some games, just tightening up the timeings was enough of a bump to get things comfortably smooth. But, if your system is already blazing you probably won't see much.
 
Originally posted by primea
For an AMD setup ... they matter a lot ... not for P4's because of their high FSB ...
Click to expand...

Latency matters on both setups. The only difference is AMD systems seem to run better in 1:1 than asynchronous ratios. Or thats the theory anyways. But as far as latency goes, you cannot dismiss the affects of latency just because you crank up the FSB and processor speed. Some benchmarks and applications are only affected by timings very little or not at all. Those kinds of applications load large amounts of raw data and can benefit from having more bandwidth. 3D applications and games on the other do not require alot of bandwidth and perform the best with tighter timings for quick access.

The lower the timings are, the faster the computer is able to get data from the memory, and the faster the rest of the PC will be.

All the RAM setting are important. The CAS latency is just part of it. The RAS->CAS latency affects performance the most and is the hardest to run at the optimal setting of 2.

Lower timings will affect especially gaming performance anywhere from 3% or more. Considering most of the time the A64 only beats a Pentium 4 3-5% in the majority of games out there then yes lower timings are important.

And considering its cheaper to buy PC3200 or PC3500 there's no reason to spend all that money on PC4000+ unless your running specific programs that will benefit from more bandwidth and you aren't building a gaming machine.

If your not worried about getting every last drop of performance then i wouldn't worry about spending a ton of money on memory as long as you buy a quality brand like Mushkin, Kingston, or Corsair. But just remember you'll get a few extra FPS in games and better overall system performance from the lower latency RAM.
 
Originally posted by burningrave101
And considering its cheaper to buy PC3200 or PC3500 there's no reason to spend all that money on PC4000+ unless your running specific programs that will benefit from more bandwidth and you aren't building a gaming machine.
Click to expand...

Or unless you got it for $100 a stick at Best Buy.
 
Originally posted by djnes
Or unless you got it for $100 a stick at Best Buy.
Click to expand...

Yea thats about the only good reason i think of for someone buying PC4000 or higher is getting it dirt cheap. But even then i'd spend a few extra bucks and got the Mushkin PC3200 lvl 1 for like $210 instead or some Kingston HyperX PC3200/PC3500.
 
I'd say it's worth it if you've already got a killer rig and you're going out to get the last little bit of performance, but, for most people they probably have a bigger bottleneck.

What video card are you running? Also, you might consider investing the money saved by going with the Value Ram into a 2400+/2500+ Mobile XP processor :D In fact, if you're only going to run at 200MHz FSB, then I'd wager the Value Ram can run at CAS 2 anyway! (don't take that to the bank, but with a bit of a voltage bump, I'm sure it could do it :))
 
Originally posted by burningrave101
Yea thats about the only good reason i think of for someone buying PC4000 or higher is getting it dirt cheap. But even then i'd spend a few extra bucks and got the Mushkin PC3200 lvl 1 for like $210 instead or some Kingston HyperX PC3200/PC3500.
Click to expand...

I can personally tell the difference with my GEIL running at PC3700/CAS 2 than with it at PC3200/CAS 2.5. Granted, it's very small, but noticeable. It's also nice to have the headroom while overclocking.
 
Just got my bh-5 hyperx in today and noticed a difference. I had the 3500 CH-5 modules and had to over-volt them to get good timings: 2-6-2-2, now with the BH-5 I can get 2-5-2-2 @ stock (2.6v). Great memory, get it while they last!
 
What kind of benchmarks improvements did you guys see when going from a higher CL rating to a lower one? So far I'm having trouble justifying the premium cost. It will probably yield about an extra 5 FPS in games, right?
 
Originally posted by DiseasedCow
Just got my bh-5 hyperx in today and noticed a difference. I had the 3500 CH-5 modules and had to over-volt them to get good timings: 2-6-2-2, now with the BH-5 I can get 2-5-2-2 @ stock (2.6v). Great memory, get it while they last!
Click to expand...

You're saying you see the difference between 2-6-2-2 and 2-5-2-2? I was under the impression that the 5/6 number was nearly inconsequential, but the other 3 numbers could have a marginally noticeable affect.
 
I think he notices a difference between bh5 and ch5... not rcd 5 or 6...
 
Originally posted by afong
You're saying you see the difference between 2-6-2-2 and 2-5-2-2? I was under the impression that the 5/6 number was nearly inconsequential, but the other 3 numbers could have a marginally noticeable affect.
Click to expand...

All the settings are equally important, some a little more then others but not alot. You'll get the maximum performance out of a 2-2-2-5 setup. You dont want to lower the tRAS any more then 5 though because you'll cause data corruption on your hard drive.

What kind of benchmarks improvements did you guys see when going from a higher CL rating to a lower one? So far I'm having trouble justifying the premium cost. It will probably yield about an extra 5 FPS in games, right?
Click to expand...

First off CL rating doesn't mean anything as far as the overall performance. You have to look at all the ratings. Just lowering one isn't going to give you better performance. You have to tweak them all. Like i said optimum is 2-2-2-5 but your only going to get that with like BH-5 chips. CH-5 max out with 2-3-2-5 which is just about as good. You'll get around a 3% or more performance increase from tighter timings. If you can't justify paying more for that thats fine and its your choice. But like i said, if your building a gaming rig and have a Pentium 4, the performance difference between the Pentium 4 and A64 the majority of the time is only 1-5% if they are running at equal relative speeds so thats something to take into consideration before you stick cheap RAM in ither one of them. And RAM with CH-5 chips that can run 2-3-2-5 isn't THAT expensive. I dont really call 1GB at $200-$240 much of a premium compared to well over $300-$400 for the PC4000+.
 
lol, I did not notice the difference between 2225 and 2226! I did notice the difference between CH-5 and BH-5 though.
 
Originally posted by DiseasedCow
lol, I did not notice the difference between 2225 and 2226! I did notice the difference between CH-5 and BH-5 though.
Click to expand...

Well or course your not going to notice it in real world performance. We're talking about benchmark scores here or at least i am. You'll get the top score with 2-2-2-5 in any benchmark that simulates something that takes timings into account. Otherwise i would set it at 2-2-2-6 because the tRAS is actually suppost to be CAS + tRCD + 2 for DDR. Trying to set the tRAS too night and not allowing enough clock cycles to finish everything can cause some serious data corruption on your hard drive.
 
Ah, yeah, you'd notice the difference between the CH-5 and BH-5 (as you mentioned that you could run the BH-5's at stock voltage). Had me confused there hehe.
 
I'm ordering new ram this weekend in an attempt to watercool my rig and overclock it a bit. As for the RAM, is there a thought or truth in that I don't need as much head-room to overclock 1:1 on a 3.2 P4 800fsb and should stay with lower latency?

I was under an impression (yes im a noob) that with say a 2.4 or 2.6 to do some strong overclocking I would need higher numbers, say DDR500 vs. 400 but with the 3.2 I can get as much out of the 400 or 433?

I'd like to run a matched pair or 512mb on the DFI LanParty 865 with the 9800XT. Gonna water cool the chip with a dangerden kit as soon as my new case shows up. Looking for some advice?

Cheers and Thanks
 
You must log in or register to reply here.
Back
Top