How much do RAM timings make a difference?

[Johnbear007]

I was thinking of picking up some crucial pc3200 but its CL3

How much worse is this than 2.5 or 2? and what speed difference will I see?

 

[wicktron]

It makes a huge difference on an AMD platform. According to your sig, it shows Barton there, so I assume this is going to be put in that rig. In that case, latencies will affect performance.

 

[Johnbear007]

Originally posted by: wicktron
It makes a huge difference on an AMD platform. According to your sig, it shows Barton there, so I assume this is going to be put in that rig. In that case, latencies will affect performance.

Yep... hmmm so A CL3 piece is a bad buy I would assume then? I just sold my crucial PC2700 and was going to buy a stick on FS for about 85 of 3200 so I could run my FSB at 200 with ram and CPU because right now they are at different FSB

 

[chocoruacal]

Originally posted by: wicktron
It makes a huge difference on an AMD platform.

Sure about that?

Here's one of a billion reviews that compare various timings. In terms of case 2 vs 2.5, I defy anyone to "feel" a 200 point 3dMark difference I have HyperX in all my nForce2 boards (2-2-2-6 I think) and the lower timings aren't apparant to me in real world use, gaming, encoding, etc.

I've never run DDR at cas 3, but your Crucial should have no problem with at least 2.5.

 

[Johnbear007]

Originally posted by: chocoruacal

Originally posted by: wicktron
It makes a huge difference on an AMD platform.

Sure about that?

Here's one of a billion reviews that compare various timings. In terms of case 2 vs 2.5, I defy anyone to "feel" a 200 point 3dMark difference I have HyperX in all my nForce2 boards (2-2-2-6 I think) and the lower timings aren't apparant to me in real world use, gaming, encoding, etc.

I've never run DDR at cas 3, but your Crucial should have no problem with at least 2.5.

Am I safe teo assume then, that getting the RAM and Cpu at 200mhz FS together, is probably more important than a small difference in timings?

 

[chocoruacal]

Originally posted by: Johnbear007
Am I safe teo assume then, that getting the RAM and Cpu at 200mhz FS together, is probably more important than a small difference in timings?

If the choice is higher FSB or timings, I'd go with the former.

 

[Regs]

Performance comes with a good balance of bus speed and timings. The value select corsair ram I had ran at 2.5 cas which was fine. When I switched to XMS 3200 I didn't notice a damn difference while the bus speed was still set at 333Mhz. The CAS latency on XMS is 2.0, which hardly makes a difference between 2.5 on the pc2700. The other timings hardly play much of a performance factor, not unless they are raised in their timings significantly enough.

Now if I had a mobo that supported a 400 bus, then I'm sure my 3200 LL would be faster than my value select 2700, since 2700 only supports a 333 Bus.

 

[jhites]

Running 1:1 is more important that the timings. Bandwidth will increase with improved timings but this would only show up as overall improvement in very few instances.

 

[hurrikaane]

And yet a lot of tests show that running 5:4 at 2-3-2 or 2-3-3 is equivalent to running 1:1 at 3-4-4.

 

[madcow235]

Thats only true on p4 systems hurrikaane. On an AMD system you need 1:1 to get decent performance. ANd yes tighter timings are good, very good.

 

[Johnbear007]

Alright, then I will go with the 1:1 over really tight timings

 

[ScrewFace]

I've noticed a drop of 400+ points in 3DMark2001SE (330) from when I had just 2x256MB of GEIL PC3200. Here I had timings of 5-2-2-2.0 but when I got a 512MB stick of Corsiar PC3200 I had to use timings of 5-3-3-2.0 so there is significant performance decrease for me at least. Also, my memory bandwidth dropped from 3.08MB/sec to 2.8GB/sec.:beer:

 

[Johnbear007]

Originally posted by: ScrewFace
I've noticed a drop of 400+ points in 3DMark2001SE (330) from when I had just 2x256MB of GEIL PC3200. Here I had timings of 5-2-2-2.0 but when I got a 512MB stick of Corsiar PC3200 I had to use timings of 5-3-3-2.0 so there is significant performance decrease for me at least. Also, my memory bandwidth dropped from 3.08MB/sec to 2.8GB/sec.:beer:

hmmm that sux but 300 pts on 3d mark doesnt concern me all that much.

 

[JustAnAverageGuy]

Originally posted by: ScrewFace
I've noticed a drop of 400+ points in 3DMark2001SE (330) from when I had just 2x256MB of GEIL PC3200. Here I had timings of 5-2-2-2.0 but when I got a 512MB stick of Corsiar PC3200 I had to use timings of 5-3-3-2.0 so there is significant performance decrease for me at least. Also, my memory bandwidth dropped from 3.08MB/sec to 2.8GB/sec.:beer:

*snicker*

Try 2-2-2-11, or 2-2-3-11 You should if nothing be able to get 2-3-3-11

 

[wicktron]

Assuming AMD, I've seen on a few benchmarks where, at any given speed, 2-2-2 timings will give the same performance as RAM that have 2.5-3-3 timings and are 11-14MHz higher, depending on benched application. The higher end typically being synthetic benchmarks, the lower end being real-world benchmarks. So, I'd say tight timings on AMD give a tangible performance boost.

 

[InlineFive]

Originally posted by: wicktron
Assuming AMD, I've seen on a few benchmarks where, at any given speed, 2-2-2 timings will give the same performance as RAM that have 2.5-3-3 timings and are 11-14MHz higher, depending on benched application. The higher end typically being synthetic benchmarks, the lower end being real-world benchmarks. So, I'd say tight timings on AMD give a tangible performance boost.

I can testify to this. Acutually PC2700 2.5-3-3 gives me 2GB/s (Athlon XP Palomino) where as PC2100 2-2-2 took it down to 1.9GB/s for some reason.

-Por