clock frequancy v's work done ?

IanBoyle

IanBoyle

IanBoyle

Associate
Joined
1 Apr 2006
Posts
117
Location
Central Scotland
hi,

i just got a i7 920 and for my first clock pushed it up to a very stable cool 3.4 gig
what i want to know is can you overclock a chip so that although its reading all sorts of high frequancy's it actually does less work ?

I do climate modeling on my machine and there are others with far higher overclocks who dont seem to get the work done that my rig does

Ian

best regards
Ian
 
If I understand that correctly, then no. It does the same processing, just faster.

The other people with higher clocks will just not have their modelling running as much as you methinks.
 
The megahertz is only really a good way to compare the same family of CPU, it has never been a good way to measure different brands/models.

i5/i7's are clock for clock faster than anything else out there and plus the i7's have hyperthreading.
 
Last edited:
Yeah you can. If you use a lower FSB/BCLK and a higher multiplier the performance will be slightly worse than if you use a higher FSB/BCLK and a lower multiplier. But I'm not sure how much difference there is, or how this might relate to your question.
 
hi,

climate modeling needs millions of calculations done so has a huge work load
all cores on continuously 24/7
a bit like running prime 95 for a year or more
so can you get to a glass ceiling where no matter how much you overclock the machine hits a brick wall and wont work any faster

its a bit like having a video card you overclock you may get some nice numbers but if the thing cant work properly you wont get a lot done on the video front

hope that explains
Ian
 
Much as mmj_uk says. To a first approximation, a given processor at 4ghz will do twice as much "work" in a given time period as the same processor would at 2ghz.

However two different types of processor, both at 2ghz, would get different amounts of work done in a given time period.

I say to a first approximation, as in practice overclocking the cpu tends to feature slackening off on some other components (ram timings, northbridge strap, unclock multi etc), so the performance increase isn't quite as good as proportional to the main clock frequency. Software will probably also make a difference, though I'm ill placed to suggest what. Lower multiplier for the same clock speed to run the ram faster is an example of tightening the sub timings for better performance.

IanBoyle said:
so can you get to a glass ceiling where no matter how much you overclock the machine hits a brick wall and wont work any faster
Click to expand...
This is unlikely to be true. At best, you're finding you're limited by something other than the processor speed, possibly how fast it can shove data to/from ram (less likely if on a triple channel board) or (my guess) to/from hard drive. Computing is all about bottlenecks, if you're not cpu limited then more ghz won't help.
 
Last edited:
hi,

sorry was posting along side you all
i am using the same cpu etc as a friend of mine so i'm comparing like for like
his is set for 3.8 gig although cpuz says its push 3.98
mines is set at 3.4 gig 21x160 so if what you say is correct i should be getting even worse performance

bit baffling though

Ian
 
IanBoyle said:
hi,

sorry was posting along side you all
i am using the same cpu etc as a friend of mine so i'm comparing like for like
his is set for 3.8 gig although cpuz says its push 3.98
mines is set at 3.4 gig 21x160 so if what you say is correct i should be getting even worse performance

bit baffling though

Ian
Click to expand...

his will still have turbo on im guessing, but as your multi is 21x yours cant. think im right in saying that.
but why yours is working faster with lower settings i dont know.
 
the reason i ask is if i'm getting all there is at 3.4 gig why bother going to 4 gig to get the same throughput or workdone figures ?

Ian
 
Nope, both have turbo on. Otherwise the highest multiplier would be x20.

Can you list the complete specification of each machine, and preferably what speed/timings the ram is running at? If they're essentially the same yet the slower clocked one is doing more work then this may be an interesting thread. A really obvious thing to do is to change the clock speed on your machine to see if it makes things better or worse, as it may just be that your mate's set his computer up badly.That way you know you're testing on the same hardware, rather than just similar.

I assume you're doing the same calculations? Points awarded in folding@home are heavily dependent on which work unit you're working on, if this is a similar distributed effort then the same idea may apply.

What sort of difference in performance are you seeing?
 
IanBoyle said:
the reason i ask is if i'm getting all there is at 3.4 gig why bother going to 4 gig to get the same throughput or workdone figures ?

Ian
Click to expand...

maybe the software you use doesnt utilise the cpus full potential or something. i know in rendering terms there was a difference in performance on my i7 from 3.4, 3.7 and 4GHz so would say there is possibly something else other than just the cpu speed.
 
hi,

things get a bit weird on the comparrison front here because the set up he uses was an overclocked 3.8 gig i7 920 gigabyte bundle from here ( overclockers )

i am running the same chip but the latest gigabyte GA-X58-UD7
and i dont really know what his memory is sorry but mines is patriot 1600

we both do the same modeling however with the same software

i've only had this cobled together system for a couple of days now so maybe i've been really really lucky and found a super sweet spot for my rig that happens to be 3.4 gig ?

i know when i got the amd fx57 ( yes before all these multi cores )
if i overclocked it too much the thing actally slowed down
now the amd had no thermal protection so would just burst into flames

but i heard at the time intel's chips were almost MOSFET in their design so would throttle back if overclock too much

wondered if the same still maybe happened slightly even today

Ian
 
im running mine at 19x211 (4GHz) and must say it feels a lot nippier when opening apps and stuff than my 3.7GHz oc did, but it does take an age before it posts :(
 
The obvious answer would be hard disk speeds.
Yours might write/read data quicker than his and so be able to complete the calc's quicker.
But this would be to simple an answer!

Get your friend to join and post his spec so we can compare!
 
IanBoyle said:
i've only had this cobled together system for a couple of days now so maybe i've been really really lucky and found a super sweet spot for my rig that happens to be 3.4 gig ?
Click to expand...

No.

IanBoyle said:
but i heard at the time intel's chips were almost MOSFET in their design so would throttle back if overclock too much
Click to expand...
Intel chips will throttle if running over something like 1.45V, 80 centigrade, but they only lose the turbo multiplier (drop down to x20). You can disable this in the bios. I don't use turbo for this reason, currently at 200x20. They also turn off if they go over 100, mine did this when I got the keyboard cable caught in the stock cpu fan while folding. No harm done whatsoever. Processors are hard as nails these days.
 
hi,

thats interesting i never thought of the hard drive as i have a ssd now i dont know what he uses
but i would have thought with 6 gig of tri memory having a faster read/write performance on your hard drive wouldnt make a differance ?

Ian
 
JonJ678 said:
Much as mmj_uk says. To a first approximation, a given processor at 4ghz will do twice as much "work" in a given time period as the same processor would at 2ghz.
Click to expand...

Almost but not quite, there is bottle neck in other components, memory bus, pci bus etc it would be more like 95% upwards, but not 100% increase.
 
@IanBoyle The hard drive is often a bottleneck. I'm not sure if it is for your software or not, but if your friend is using a hard drive and you an ssd, it would fit.

@Jason indeed, but the directly proportional relationship is good enough for most purposes. Especially when described as a first approximation :)
 
You must log in or register to reply here.
Back
Top Bottom