12 minutes ago, SteveW said:

On that site said it was unglamorous game

Hey! It is Brad Pitt! Glamorous.

Edited February 10, 20206 yr by Dirk98

Yeah, they were not making out it was Brad or anyone like that.

Janelle Monae will be remained above anyone this time 

 

Forgot to double, fair enough, makes no odds that 19 is faster than 16 is the point I was making. So the 4800 slips out of that  just. Someone looking at CL rather than frequency may get something slower is the point I'm making.

Seems we almost always see memory performance is better with higher frequency than lower CL, DDR4 usually more than 15. No matter the unscrupulousness of memory dealers and what not else can be thrown in. Lets see what DDR5 brings.

😊

That's right Rob you can see 100% CPU and it's just counters, which is what I've said all along..

Edited February 10, 20206 yr by SteveW

Also what you posted Rob, is part of an explanation why HT enabled is better than HT disabled. "With hyperthreads, however, those stalled cycles can now be used by another thread".

And all that chest puffing you guys make about frequency and latency debates, which amount to very little, when there's nothing like shooting a hole in your performance like disable HT. Just not everyone knows how to measure it.

I think since you changed the subject again it is fair to mention. As far as I know, only Ideal Flight's experimental CPU monitor takes those things into account and can be seen to be recording a slightly different affair where the sim is concerned. So yes that's all fair to say. I've not changed my advice since I appeared on the scene. But others seem to be following. In the end it's what the sim performs like in real life that matters.

Edited February 10, 20206 yr by SteveW

I think that means that my DDR4-4266 RAM @ 3600Mhz has a latency of around 10.5ns.

That's the strobe timing, yes Chris. It's very difficult to determine the real performance.

Although Rob found a fairly accurate post on CPU performance counters not necessarily accurate, the post doesn't say the regular Task Manager graphs are useless. Far from it, they do allow an insight into why the fans are making a fuss, and give a pretty good idea of the consumed performance. Some programs and tools can hide those readings and can even show 100% all the time.

17 hours ago, Ron Attwood said:

You actually made sense of all that? I've got to be as thick as two short planks!

 it was as clear as mud Ron! I'm just trying to be courteous  

Some find it hard to be courteous but it costs nothing.

The way I see it is the public forums are for all-comers. So there's no reason to filter more difficult concepts based on the OP discussion, but I agree it can get a bit tedious and I stick to simplified examples as much as possible. We shouldn't plateau understanding though. Also, after-discussion, shooting the breeze, is often just as rewarding as the OP discussion.

The memory is accessed on a timely basis with a strobe. If it's not ready the system finds other things to do. At that data rate,  getting lower latency doesn't get so big a performance increase as you might think. But increase the frequency and then the accesses are noticeably faster. When doing the math, as Rob showed, the gap doesn't close much and at only the very highest rates the gap overlaps a tiny bit, with very expensive memory. Accesses are arranged pre ordered. The programs running are analysed and a more efficient use of the memory is made by pre ordering. When we enable HT we see more chances of hitting more efficient accesses.

"Accumulates wait time" OK I see.

Thanks Rob.

Steve