The big memory dilemma

Trizik

Senior member
Jun 17, 2005
362
0
0
We've seen it before, we're seeing it again... I apologize in advance.

I currently have 2 x 512MB DDR 400 (2.5-3-3-7) running @ 1T with an Athlon 64 3200+

In the long run, would I be better off with 4 x 512MB DDR 400 (2.5-3-3-7) @ 2T or 2 x 1GB DDR 400 (3-4-4-8) @ 1T?

I don't want to give up that 0.5 CAS (and other lower timings) by switching to 2 x 1GB, but from what I've read the A64 will make up for it.

THE BIG QUESTION: Is 1T speed with higher latency (3-4-4-8) better than 2T speed with lower latency (2.5-3-3-7) on an A64 setup?
 

Vegito

Diamond Member
Oct 16, 1999
8,329
0
0
im using 2x1gb on 1 system and 4x1gb on another..

the way i see it, you'll be needing more memory as time goes, if you dont upgrade as often as me, then getting the biggest ram available.. else you'll need to sell 2x512 and get 2x1gb later on...

 

JBT

Lifer
Nov 28, 2001
12,094
1
81
CAS 2 instead of 2.5 is going to improve performance like 1 to 2% maybe.

At any rate got for the GB's.
 

Trizik

Senior member
Jun 17, 2005
362
0
0
I'll have CAS 3 if I go with 1GB sticks... CAS 2.5 if I go with 512MB sticks.
 

cryptonomicon

Senior member
Oct 20, 2004
467
0
0
Originally posted by: Trizik
THE BIG QUESTION: Is 1T speed with higher latency (3-4-4-8) better than 2T speed with lower latency (2.5-3-3-7) on an A64 setup?

absolutely. 1T vs 2T has the greatest effect of any timing setting

 

MobiusPizza

Platinum Member
Apr 23, 2004
2,001
0
0
Originally posted by: cryptonomicon
Originally posted by: Trizik
THE BIG QUESTION: Is 1T speed with higher latency (3-4-4-8) better than 2T speed with lower latency (2.5-3-3-7) on an A64 setup?

absolutely. 1T vs 2T has the greatest effect of any timing setting

While that's true, the maximum performance impact is 5%
If you get 512MB ram with lower latency and higher overclockerbility, the performance difference narrows to less than 1-2%

For gaming, 4*512MB is more fruitful
4*512MB would overclock better because of the nature of RAM
By that for games it's more than offset the 2T timing
2T timing does not affect game framrate by more than 2 fps

Read here for direct performance comparison

For gaming, you'll probably want to go with 4x512 of memory that is known to hit good speeds (240MHz or so) and keep decent 2.5-2-2 or 2-2-2 timings. As you can see, you don't have to worry about the 2T performance hit because the memory speed will more than compensate that. Crucial Ballistix fits this bill and is much cheaper than the alternative of GSkill TCCD. Or, you could try your luck with 2 gigs of TwinMOS SP or Mushkin Redline.

For basically all else, it would probably be quite a bit cheaper to get 2x1024mb sticks. Some, such as Crucial or Patriot, can have very tight timings at between 200 and 220mhz. So those would be a good choice. Another advantage of 2x1024 is that it leaves some slots empty for future expansion. I hope that you are now a more informed forum in this regard.


Also, from here

WELL! there you have it. From this, and baron's results, it seems that the typical 'benchmarks' are the only programs that see any difference from memory speed and latency.
Games show a small difference, but nothing really amazing.
Music conversion.. yeah
Rendering programs are entirely CPU limited.
 

Trizik

Senior member
Jun 17, 2005
362
0
0
I'm not interested in overclocking, and the RAM I bought can't really be overclocked anyway.

I am a definite gamer and am very concerned about performance, but I can deal with a ~3 fps drop.

Won't my A64 3200+ compensate for the 2T speed because of its ability to process a lot of data per clock cycle?

EDIT NOTE: Your signature freaked me out for a sec. ;)
 

Diasper

Senior member
Mar 7, 2005
709
0
0
Originally posted by: AnnihilatorX
Originally posted by: cryptonomicon
Originally posted by: Trizik
THE BIG QUESTION: Is 1T speed with higher latency (3-4-4-8) better than 2T speed with lower latency (2.5-3-3-7) on an A64 setup?

absolutely. 1T vs 2T has the greatest effect of any timing setting

While that's true, the maximum performance impact is 5%
If you get 512MB ram with lower latency and higher overclockerbility, the performance difference narrows to less than 1-2%

For gaming, 4*512MB is more fruitful
4*512MB would overclock better because of the nature of RAM
By that for games it's more than offset the 2T timing
2T timing does not affect game framrate by more than 2 fps

Read here for direct performance comparison

For gaming, you'll probably want to go with 4x512 of memory that is known to hit good speeds (240MHz or so) and keep decent 2.5-2-2 or 2-2-2 timings. As you can see, you don't have to worry about the 2T performance hit because the memory speed will more than compensate that. Crucial Ballistix fits this bill and is much cheaper than the alternative of GSkill TCCD. Or, you could try your luck with 2 gigs of TwinMOS SP or Mushkin Redline.

For basically all else, it would probably be quite a bit cheaper to get 2x1024mb sticks. Some, such as Crucial or Patriot, can have very tight timings at between 200 and 220mhz. So those would be a good choice. Another advantage of 2x1024 is that it leaves some slots empty for future expansion. I hope that you are now a more informed forum in this regard.


Also, from here

WELL! there you have it. From this, and baron's results, it seems that the typical 'benchmarks' are the only programs that see any difference from memory speed and latency.
Games show a small difference, but nothing really amazing.
Music conversion.. yeah
Rendering programs are entirely CPU limited.




Wrong about the 4x512MB being better for overclocking - the memory controller simply won't be able to handle four full dimms running at 250mhz at timings like 2-2-2. Heck I don't know if it can even run at that speed at any timing speed.

It would be much better to go for 2 x1GB sticks. Your best option IMO is to go for Crucial Value 1GB sicks since they're speed binned Ballistic - some Ballistic have been able to hit 2.5-2-2 at 250mhz @1T @2.8V (though people's experiences have varied) while many people getting value sticks have been reporting 3-3-3 at around 240-250mhz.

Not a bd deal at all IMO - and with the overclockabale bandwidth at 1T you'll be set for X2 which might suffer from running 4 dimms at only 200mhz and at 2T - they do need more bandwidth
 

MobiusPizza

Platinum Member
Apr 23, 2004
2,001
0
0
Originally posted by: Diasper
Wrong about the 4x512MB being better for overclocking - the memory controller simply won't be able to handle four full dimms running at 250mhz at timings like 2-2-2. Heck I don't know if it can even run at that speed at any timing speed.

That's the exact reason why 1T timing cannot be issued.
However I do believe 250mhz and 2-2-2 timing is feasible

nevertheless I'd have a definite answer when I try the configuration in the summer
3-3-3 timings hurt performance quite bad
Keeping latency below 2.5 is recommended

 

cryptonomicon

Senior member
Oct 20, 2004
467
0
0
Originally posted by: AnnihilatorX
4*512MB would overclock better because of the nature of RAM
[/quote]

...your credibility just got destroyed. running 4x512 w/good timings and overclocking too is next to impossible on a64 platform.
 

rise

Diamond Member
Dec 13, 2004
9,116
46
91
Originally posted by: cryptonomicon
...your credibility just got destroyed. running 4x512 w/good timings and overclocking too is next to impossible on a64 platform.

go visit xs or dfi.street. a guy named thunda is running his tcc5 chips 4x512 with a 3700+sd at 2.5-3-3-8 264mhz. others can do the same. it depends on the chips and the board as well as the cpu.

edit-and from everything i've seen low latency is far better than 1t outside of synthetic bench bs.
 

Trizik

Senior member
Jun 17, 2005
362
0
0
Just reading the replies to this thread and other threads, as well as a couple of web sites, I can tell that the debate between 1T vs. 2T is far from over. Some say higher latencies @ 1T is better than lower latencies @ 2T, others say lower latencies @ 2T are better than higher latencies @ 1T. Everyone seems to agree that 1T performs 2-3% better in benchmarks than 2T at the same timings, but some say those benchmarks aren't noticeable in the real world while others say the boost definitely makes a difference even if the 1T latencies are a bit higher.

I can't find a thorough explanation about the differences between 1T and 2T. I know 1T is faster, but why?
 

cryptonomicon

Senior member
Oct 20, 2004
467
0
0
lol, i've read thunda's threads. what he did was a pure anomaly, he was incredibly lucky. NO WAY would i EVER recommend over 2 dimms, it's so unpredictabale.