6 replies to this topic

[aionaddict]

I am thinking about boosting my trusty old e6750 from 3.2 to 3.4 or 3.5 ghz. For some reason when increasing the CPU speed the Ram copy speed goes down. The CPU tests stable at 3.4 ghz and interestingly if I go to 3.5 ghz the Ram copy speed goes back up a bit. Any ideas?

[Nagisan]

It could be due to the higher latency but that is pretty unusual, the memory bus speed goes up but the copy speed (and only the copy speed) drops. I'm guessing latency has something to do with it but I've never full understood RAM overclocking.

EDIT: You could try lowering your latencies back down to where they were (might have to give more voltage to the ram), or put your overclock back to 3.2 Ghz, and increase the latencies to where they are when you go to 3.4 Ghz....that should give you a better direct comparison of how latency affects copy speed (if it does at all).

EDIT2: From a quick google, latency is basically the wait time between operations (in cycles), so if your ram has a latency of 5 (such as your 3.2Ghz), it will wait 5 cycles between reading and writing. A latency of 6 would wait 6 cycles. Obviously this wouldn't affect read or write speeds, but when copying you are doing more than 1 operation so the longer the wait in-between, the slower the overall copy (when measured over the same amount of time or using the same amount of data).

[geekyscott]

Possibly something to do with the increased memory timings?

[Detection]

Possibly the RAM needs some extra juice

[aionaddict]

Nagisan, on 04 October 2012 - 14:29, said:

EDIT2: From a quick google, latency is basically the wait time between operations (in cycles), so if your ram has a latency of 5 (such as your 3.2Ghz), it will wait 5 cycles between reading and writing. A latency of 6 would wait 6 cycles. Obviously this wouldn't affect read or write speeds, but when copying you are doing more than 1 operation so the longer the wait in-between, the slower the overall copy (when measured over the same amount of time or using the same amount of data).

Hmm that sort of makes sense will try looking more into it and experimenting more.

[+Tempus]

I'd go with timing issues. When the clock is pushed up, the CAS cycles are most likely not toggling fast enough on all operations to keep up so you land up skipping cycles to move the data.In the copy process you need to read and write and if that may not always able to fit in a single cycle so you land up in a wait state for a cycle.

Just a guess however...

[+Jdawg683]

yeah it's absolutely the change in memory timings. I wouldnt care though - you'll benefit more from higher clocks than timings.