Here's the two runs for CPUtest.

JohnnyB 

Here are the Task Manager screens on my workstation. HT off was faster. CPU usage still averaged on 40%.

Thanks.

I guess bottom line, NT7 is not as efficient as it could be in terms of optimization and backtesting. So we'll hope NT8 does better.

I wonder if someone could also comment on the memory scaling question I had in post #1. I haven't had a chance to do my own tests yet (but will) --- question is: Does the amount of memory exponentially scale with the number of cores/threads?

Mike

JohnnyB 

Here's the two runs for CPUtest.

Interesting. If you happened to run it on the same settings as me (ES, January 1 to present, 60 minute time frame) I would like to know how long the test took to complete on run #2 or after (run #1 is needed to cache the barseries I think).

I ask because your machine is obviously much faster than my single X3450 and I'd like to compare.

Today Intel released the new Xeon E5's, and I am drooling.... must... resist...

Mike

I'll rerun you test lataer today with your settings.

Yes, the new Intel chips look great. I have several HP DL 380 G8s with the new Intel chips on order for testing. Unfortunately, it's for real work...

JB

i use Intel QuadCore Q6600 with 3GB memory 64 bit windows ultimate

heres my screenshot

jkpdon 

i use Intel QuadCore Q6600 with 3GB memory 64 bit windows ultimate heres my screenshot

Now that looks ok to me - decent utilization. COULD be higher, but getting it there can be a programmer task (i.e. require quite some tuning). Looks quite ok.

It seems really that Ninja just has a scalabilityp roblem in scheduling - it is ok for what was a decent workstation some years ago when the started working on 7, but it is not ok for what is now a higher end workstation (which has had 12 cores , 24 threads already last year - HP Z800 for example).

Interesting enough pretty much generic - as I said, when we move to brute force things got nicely used up.

Big Mike 

Quick test: With HT enabled = 54 seconds With HT disabled = 42 seconds, a 25% improvement in speed Mike

Brute Force with HT = 44 secs
Brute Force without HT (NO HT) = 34 secs

Genetic with and without was 3 and 2 seconds respectively. Great inprovement.

JohnnyB 

Brute Force with HT = 44 secs Brute Force without HT (NO HT) = 34 secs Genetic with and without was 3 and 2 seconds respectively. Great inprovement.

Very nice.

I just bought a Xeon E5-1650, 32GB, 2U Supermicro box. Will be interesting to see how it compares as well. Three week delay before I get it.

It will replace some servers I have at Steadfast in Chicago. I am still trying to decide what to buy to replace my i7 920 home machine, but leaning towards a 3930k (equiv of the E5-1650).

Mike

Big Mike 

Very nice. I just bought a Xeon E5-1650, 32GB, 2U Supermicro box. Will be interesting to see how it compares as well. Three week delay before I get it. It will replace some servers I have at Steadfast in Chicago. I am still trying to decide what to buy to replace my i7 920 home machine, but leaning towards a 3930k (equiv of the E5-1650). Mike

With a decent mothebord (there are none that are not - that is a x70 high end chipset) and a decent CPU cooler that is a nice 4 to 4.5 ghz per core under full load Remember to set that up - that chip is MADE for mild overclocking.

Home system will be water cooled, 4.5 should be walk in the park.

Mike