Is n´t it rediculous? 1 cpu faster than 2 cpus?

Constantinos
Constantinos
Joined: 16 May 05
Posts: 39
Credit: 938513
RAC: 0

RE: I spent my afternoon on

Message 30450 in response to message 30449

Quote:

I spent my afternoon on controlled trials on the HT advantage/disadvantage then extended the testing as akosf requested for his "S41.07HT" code.

Where do we find this S41.07HT code?

Gravity increases significantly in Autumn, because apples fall in large numbers during that time!

anders n
anders n
Joined: 29 Aug 05
Posts: 123
Credit: 1656300
RAC: 0

RE: RE: I spent my

Message 30451 in response to message 30450

Quote:
Quote:

I spent my afternoon on controlled trials on the HT advantage/disadvantage then extended the testing as akosf requested for his "S41.07HT" code.

Where do we find this S41.07HT code?

this link.

Anders n

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6534
Credit: 284710168
RAC: 110462

RE: [pre]Version HT nHT

Message 30452 in response to message 30449

Quote:
[pre]Version HT nHT HT/nHT productivity ratio
dist 7584 4630 1.22
S40 1954 1159 1.19
S40.04 1928 1141 1.18
S41.07 2367 870 0.73
S41.07HT 1948 1145 1.18[/pre]


Well, yes. That's no surprise mostly .... :-)
The 'other' thread is taking resources - HT will never speed up a given single task if you halve the resources available to it!
What is really needed is whether two simultaneously running WU threads ( under HT ) finish in less than twice the time that the one WU does ( under non-HT ). ( All WU's identical ).
Cheers, Mike.

( edit ) Or putting it another way .... does the non-HT single thread configuarion take more than half the time of one ( of two ) HT threads? If so, then with HT enabled, and both threads used for WU's, then you can complete more WU's in a given time! A rough comparison with your data:

[pre]Version HT nHT * 2 Percentage Gain ( +ve towards HT )
dist 7584 1740 = 870 * 2 { (1740 - 2367)/1740 = - 627/1740 = - 36.0 % }
S41.07HT 1948 < 2290 = 1145 * 2 { (2290 - 1948)/2290 = 342/2290 = 14.9% )
[/pre]

So that's a good point archae86, work nicely done! So what's with S41.07? Any hints anybody? akosf?

( edit squared ) Makes Intel's 30% claim ( for servers? ) pretty likely to be best case. Seldom seen?

I have made this letter longer than usual because I lack the time to make it shorter ...

... and my other CPU is a Ryzen 5950X :-) Blaise Pascal

Akos Fekete
Akos Fekete
Joined: 13 Nov 05
Posts: 561
Credit: 4527270
RAC: 0

Hi! At first, thanks to

Message 30453 in response to message 30452

Hi!

At first, thanks to archae86 for the nice datas.

Quote:
So what's with S41.07?

S41.xx has an advanced overlapping technique, that means the FPU and the SSE instructions are executed in same time. Probably the HT control cannot do with this, because this technique is able to load the cpu totally. S41.07HT doesn't consist overlapping.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.