how do cas timings REALLY effect performance.

[digitalemperor]

When speaking in terms of gaming and frames per second, will tigher timings increase performace a noticeable amount in games like unreal tournament 2003 (a game that is more cpu limited then it is graphics card limited) ?

How much of a difference does 2-2-2-11 vs. 3-4-4-11 make?

have there been any benchmarks done recently to compare this? i am intrested in 1:1 ratio and for both tests to be constant except for the timings (same multi/fsb/etc).

maybe someone with good memory and a good setup could test this out?

i am looking to get some performance and I am wondering if better ram (capable of tigher timings) is going to yield anything worth while.

 

[Soulkeeper]

the pc4200 and pc4400 stuff seems to wipe the floor with it's bandwidth (hynix chips) even when timings are slow like you mention

if you are getting pc2700 to pc3700 look good timings
in my experience they can prove atleast 5-25% increase depending on what you are running when using the tightest timings like 2-2-2-6 (best in my experience with my bh-5 stuff)

 

[jhurst]

I have seen that MHZ beats timings. If you can run your RAM at DDR500/533, and actually utilize it fully (250FSB 1:1 for DDR500), running at 3-8-4-4, will beat any PC3200 @ 2-2-2-5. The extra bandwidth you can get from PC4000+ RAM really does make a difference.

 

[klah]

http://www20.tomshardware.com/motherboard/20040119/index.html

http://www.techwarelabs.com/reviews/memory/memory_timings/

 

[AWhackWhiteBoy]

as DDR ram speeds increase the timings matter less and less.

 

[Avalon]

I don't know if you count 3dmark2001SE as meaning much, but I tried benching my gig of PC3200 geil memory with CAS 2 and CAS 2.5 latency, and I only had a change of about 50-100 points or so. I tried to bench my memory at CAS 3, but for some odd reason my geil memory cannot run at CAS 3 no matter what voltage or memory speed I run my sticks at...very weird

 

[PWNettle]

I checked out those articles but they didn't answer a question I was going to ask today...

If I were building a system with an AMD Athlon 64 3000/3200 an a typical motherboard (Asus KV8, for ex) - would there be any difference in performance between, say, Kingston HyperX PC3200 and PC4000? Or would both operate the same due to bandwidth limitations. (I would not be overclocking which seems to be a factor in this issue.). Would there be any harm in running the PC4000 (if one could, say, get it at a steal at BBf?)

I'm very confused on this issue.

If someone could answer this in utter noob friendly terms I'd appreciate it. Or if you can link me to something that explains it and gives clear answers. I'm finding that a lot of reviewsand articles tend to include a lot of benchmarking and stuff but the reviewers will never come out and say XXX is the best or whatever - compared to this forum where people will give you clear answers.

 

[Avalon]

There would be no harm in running PC4000 memory in your system...I know you said you won't be overclocking, but will you run the memory as PC4000 or let it default to PC3200 and keep it that way? Running it as PC3200 will be the same as using a true PC3200 stick performance wise. You could get tighter timings with the PC4000 stick running at PC3200 if you chose to tweak your settings in BIOS. Either way, if you could get PC4000 memory cheaply enough where you consider it a steal, then that sounds like the best thing to do.

 

[PWNettle]

Thanks.

 

[Shimmishim]

the new ocz EB ram is breaking the notion of low cas = better performance..

they are able to get cas 2 performance from settings at cas 2.5 to cas 3 even...

to answer your question.. these days it doesn't matter as much...

but in the past, yeah cas 2 seem to be the way to go...

thugsrook had some tests where he ran llow fsbs tight timing cas2226 and compared it to loose timing higher fsb's..

 

[DAPUNISHER]

What Shim said. Plus, for gaming on A64 timings mean little. The best argument for fast timing ram on A64 is it tends to overclock better, but if you can get looser timing ram rated at a significantly higher DDR speed for approx. the same money that's a btter option since the speed is guaranteed.

 

[Avalon]

What I'm wondering about the EB series from OCZ is...is it their memory design that is providing their BH-5 comparable specs, or is it the configuration of their memory?
I tried to run at 3-2-8 but my memory groaned at me and gave me a nice black screen

 

[Pauli]

Originally posted by: PWNettle
I checked out those articles but they didn't answer a question I was going to ask today...

If I were building a system with an AMD Athlon 64 3000/3200 an a typical motherboard (Asus KV8, for ex) - would there be any difference in performance between, say, Kingston HyperX PC3200 and PC4000? Or would both operate the same due to bandwidth limitations. (I would not be overclocking which seems to be a factor in this issue.). Would there be any harm in running the PC4000 (if one could, say, get it at a steal at BBf?)

I'm very confused on this issue.

If someone could answer this in utter noob friendly terms I'd appreciate it. Or if you can link me to something that explains it and gives clear answers. I'm finding that a lot of reviewsand articles tend to include a lot of benchmarking and stuff but the reviewers will never come out and say XXX is the best or whatever - compared to this forum where people will give you clear answers.

Look, if you simply replace PC3200 in a running system with PC4000 without changing the timings or FSB speed settings, it should give you identical performance. The RAM itself doesn't control the speed of the system -- the Motherboard settings do. The only benefit of going from PC3200 to PC4000 is that you will then be able to change the Motherboard settings to either overclock the system and/or tighten up the RAM timings. If you don't plan on overclocking, PC4000 won't give you any better performance unless you were able to run it with tighter timings than the PC3200. Since almost all PC4000 sticks are spec'd to run at CAS3, then you'd be wasting your money.

 

[JSSheridan]

I ran some some test like this on my AXP 2400+, and the improvement from 2.5-3-3-7 to 2-2-2-4 was between 0 and 3%. Benchmarks included 3DMark2k1, Splinter Cell, UT2k2 Flyby and Botmatch, and Halo. I don't know how they affect a A64, A FX, or a P4 system's performance. Peace.

 

[edmundoab]

I don't really notice much difference from CAS 3 to CAS 2
but benchmarking definately showed the difference,

as for gaming.. thus far not at all.
At least I would say not visible to my point of view

 

[Dman877]

With my A64 I notice no difference switching from 2/2/2/5 to 2.5/3/3/6.