Corsair 6400C3 at 1116MHz on AM2

[sierra_bound]

Running RAS# to CAS# Delay at 2 is very difficult to do with 1GB modules, at least on the Intel platform. The C3's work best at 3-4-3 or 3-3-3.

[_damien_]

Running RAS# to CAS# Delay at 2 is very difficult to do with 1GB modules, at least on the Intel platform.

Does the memory controller of current AM2 CPUs even support tRCD 2? I remember Tony had a thread about it shortly before AM2's release, stating that the controller doesn't support anything tighter than 3-3-3.

[sierra_bound]

I can't speak for AM2. 2X512MB kits will run 3-2-2 on the Intel platform.

I concluded a long time ago that memory speed is much more important than timings with DDR2. Given a choice between running 3-2-2 @ 350 vs. 4-3-3 @ 500, I would choose the latter any day of the week.

The reason why I bought the C3's is because they can run RAS# to CAS# at 3 at very high speeds, which is very rare for 1GB DDR2 modules. So you get the best of both worlds; speed and good timings. Many other DDR2 modules can't run X-3-X at DDR2-900 or higher.

[_damien_]

I concluded a long time ago that memory speed is much more important than timings with DDR2. Given a choice between running 3-2-2 @ 350 vs. 4-3-3 @ 500, I would choose the latter any day of the week.

Well, I'd say that's a characteristic of the P4 platform, not the memory technology

The reason why I bought the C3's is because they can run RAS# to CAS# at 3 at very high speeds, which is very rare for 1GB DDR2 modules. So you get the best of both worlds; speed and good timings. Many other DDR2 modules can't run X-3-X at DDR2-900 or higher.

Yeah, there's no denying that Corsair C3s are amazing. IMO Corsair is the leading brand in the DDR2 world ATM.

[ozzimark]

i speak from personal experience that AM2 does no support x-2-2-x
3-3-3 is the lowest it'll go it seems.. i wonder if amd will change that with revG's memory controller.

[_damien_]

i speak from personal experience that AM2 does no support x-2-2-x
3-3-3 is the lowest it'll go it seems.. i wonder if amd will change that with revG's memory controller.

Thanks, that's what I thought. But I still can't figure out what necessitates that limitation

[STEvil]

well, it might actually go lower once we get a tweaker like A64 tweaker for it.. it was possible to set TRCD and TRP(R or W) to 1 for DDR1 with certain bioses or with A64 tweaker for instance.. or it could just be a limit of the memory controller (which sucks but can be overcome).

[_damien_]

Yes, we've all seen many examples of unofficial timings that actually do work (even though many values do not really work). But as tRCD and tRP are internal DRAM operations that don't really stress the controller, I'm pretty optimistic about it.

[sierra_bound]

Well, I'd say that's a characteristic of the P4 platform, not the memory technology

Mmmm, not sure I buy that. But anyway, this discussion is straying from the subject of this thread.

Regarding the C3's, they are very expensive. But not many people have been disappointed by their performance.

[ozzimark]

well, it might actually go lower once we get a tweaker like A64 tweaker for it.. it was possible to set TRCD and TRP(R or W) to 1 for DDR1 with certain bioses or with A64 tweaker for instance.. or it could just be a limit of the memory controller (which sucks but can be overcome).

use systool



it cas tcl 2 oddly enough, but not trcd 2 or trp 2

[STEvil]

you cant set CAS latency from within windows anyways, so that doesnt matter.

Systool was built from a64 tweaker iirc..

edit - or heavily influenced by.

[_damien_]

you cant set CAS latency from within windows anyways, so that doesnt matter.

I don't think any DDR2-chips support CAS 2 anyway. Some of the internal DRAM operations would actually work at any setting, as long as the number of cycles provide sufficient time for signal stabilization. Many settings would work even @ 1, although in real life that would obviously require a pretty low clock frequency. However CAS latency is different, in that each setting must be build into the chips' I/O gating in the form of pipeline stages. If the chip lacks the CAS 2 stage, that setting simply won't work, regardless of voltage and clock frequency. CAS cannot be higher or lower than the supported settings - one example of that is the Winbond BH-design, which lacks the CAS 3 stage.

Mmmm, not sure I buy that. But anyway, this discussion is straying from the subject of this thread.

Certain CPU-features determine whether a platform responds well to memory timings or bandwidth - that's an established fact. But yeah, it's way off-topic.

[sierra_bound]

Back on topic. 6400C3's and Conroe.