AnsweredAssumed Answered

Was wondering how cpu cache played into SW

Question asked by bob hern on Feb 3, 2008
Latest reply on Feb 5, 2008 by bob hern

Like • Show 0 Likes0
Comment • 5

went ahead and built a machine to see how the effects of cpu cache had on the benchmark vs pure clock speed. decent for people looking for a budget setup or people who use small parts/assemblies

here are the machine specs:

OS: XP PRO SP2
SW: 2008 SP0
Motherboard: GIGABYTE GA-P35-DS3L (F7 Bios)
CPU: INTEL E2160 @ 3.0GHz (dual core, 65nm, conroe, 1MB cache, using stock cooler)
Power Supply: Antec EarthWatts 380W
GPU: 3dfx voodoo 3 2000 PCI (just a proof of concept/test build)
RAM: 4GB MICRON DDR2-667 (D9 chips)
HDD: 7.2K PATA

cost - i had a lot of parts already, and picked up everything else through watching the hot deals sections on anandtech or fatwallet - so my out of pocket for the hardware was ~$170

time to finish the ctrl-q rebuild on the punch holder - 150seconds

a bit slow compared to the 6750/6800 and other higher end cpus, but when you consider this is a ~$70 cpu, not bad. this is a stable and from what i have read easy o/c (stock is 1.8GHz) w/ thes x21xx series of cpus. there is still more headroom in it, but this showed me what i wanted to see w/ comparisons to anna's spreadsheet.

looking for someone w/ a 4GHz+ 45nm cpu to do the benchmark

No one else has this question

Outcomes

Simon Oldfield Feb 4, 2008 7:46 AM

I'm not convinced on your method, surely to test this out you need a variety of chips running at the same clock speed with differing ammounts of Cache - otherwise all you are proving is that overclocking a cheap low end CPU works quicker than it would if not overclocked.

Also during the overclocking you are not accounting for the increased FSB or other factors like PCI bus speed increase (unless you locked it).

Actions

bob hern @ Simon Oldfield on Feb 4, 2008 4:46 PM

Simon Oldfield wrote:

I'm not convinced on your method, surely to test this out you need a variety of chips running at the same clock speed with differing ammounts of Cache - otherwise all you are proving is that overclocking a cheap low end CPU works quicker than it would if not overclocked.

Also during the overclocking you are not accounting for the increased FSB or other factors like PCI bus speed increase (unless you locked it).

that is what anna's spreadsheet is for - it has all the other data. when o/c'd, the chip was running at a 1333fsb/667 so it can be compared to the 6850 for clock speeds or 6750 (slower at 2.66GHz) for pure fsb comparison only.

also, pci/pci-e bus speed is locked

Actions

Dave Solly Feb 4, 2008 7:19 PM

bob hern wrote:

CPU: INTEL E2160 @ 3.0GHz (dual core, 65nm, conroe, 1MB cache, using stock cooler)

That is a pretty impressive OC on a E2*** chip, particularly on stock cooling
The problem with that series of chip is that it is essentially nothing more than a dual core celeron,
what was the result of the test at stock speeds?

But as Simon has said... its not convincing. TBH you need to take that CPU off and replace it with a 6850 and run it with all the same components, otherwise you are not drawing a direct comparison.
There are alot of different components out there, and everyone probably has a slightly different setup of their system when they ran Anna's benchmark

Actions

bob hern @ Dave Solly on Feb 4, 2008 8:45 PM

honestly, i didn't benchmark it at stock and now it is apart going into another case as i make it my main rig (still gets the job done a bit over 1min than current rig).

i can understand what you are saying about getting a 6850 to compare. unfortunately i feel atm the 6850 is not a good investment and personally wouldn't buy one, but you can look at anna's spreadsheet - the 6th or 7th machine down is another i spec'd for a friend - running the same m/b and the ram timings are the same and he has a 6750 getting 108 seconds for the same benchmark, atm that is as close as i can get to a direct comparison. i really thought i would be in the 120s or so....

once i get it into the other case, i will then redo the benchmark at stock speeds to see how well it scales.

and thanks for the kudos on the o/c, it gets a bit warm but still under the "hot" point for these cpus. once in the new case a new hsf will go on. i have read in many different forums that these cpus seem to top out @ 3.0-3.4GHz, so getting 3.0 right out of the box was nice - not much messing w/ V, in fact i was able to lower it a bit from what "auto" had it set on and it is still prime95 stable and dropped 2C

Actions

bob hern Feb 5, 2008 7:27 AM

the same setup @ stock speed (1.8GHz) runs the benchmark in 225seconds

Actions

5 Replies

Simon Oldfield Feb 4, 2008 7:46 AM
Mark Correct
Correct Answer
I'm not convinced on your method, surely to test this out you need a variety of chips running at the same clock speed with differing ammounts of Cache - otherwise all you are proving is that overclocking a cheap low end CPU works quicker than it would if not overclocked.

Also during the overclocking you are not accounting for the increased FSB or other factors like PCI bus speed increase (unless you locked it).
Like • Show 0 Likes0
Actions
- bob hern @ Simon Oldfield on Feb 4, 2008 4:46 PM
  Mark Correct
  Correct Answer
  
  Simon Oldfield wrote:
  
  I'm not convinced on your method, surely to test this out you need a variety of chips running at the same clock speed with differing ammounts of Cache - otherwise all you are proving is that overclocking a cheap low end CPU works quicker than it would if not overclocked.
  
  Also during the overclocking you are not accounting for the increased FSB or other factors like PCI bus speed increase (unless you locked it).
  
  that is what anna's spreadsheet is for - it has all the other data. when o/c'd, the chip was running at a 1333fsb/667 so it can be compared to the 6850 for clock speeds or 6750 (slower at 2.66GHz) for pure fsb comparison only.
  
  also, pci/pci-e bus speed is locked
  Like • Show 0 Likes0
  Actions
Dave Solly Feb 4, 2008 7:19 PM
Mark Correct
Correct Answer

bob hern wrote:

CPU: INTEL E2160 @ 3.0GHz (dual core, 65nm, conroe, 1MB cache, using stock cooler)

That is a pretty impressive OC on a E2*** chip, particularly on stock cooling
The problem with that series of chip is that it is essentially nothing more than a dual core celeron,
what was the result of the test at stock speeds?

But as Simon has said... its not convincing. TBH you need to take that CPU off and replace it with a 6850 and run it with all the same components, otherwise you are not drawing a direct comparison.
There are alot of different components out there, and everyone probably has a slightly different setup of their system when they ran Anna's benchmark
Like • Show 0 Likes0
Actions
- bob hern @ Dave Solly on Feb 4, 2008 8:45 PM
  Mark Correct
  Correct Answer
  honestly, i didn't benchmark it at stock and now it is apart going into another case as i make it my main rig (still gets the job done a bit over 1min than current rig).
  
  i can understand what you are saying about getting a 6850 to compare. unfortunately i feel atm the 6850 is not a good investment and personally wouldn't buy one, but you can look at anna's spreadsheet - the 6th or 7th machine down is another i spec'd for a friend - running the same m/b and the ram timings are the same and he has a 6750 getting 108 seconds for the same benchmark, atm that is as close as i can get to a direct comparison. i really thought i would be in the 120s or so....
  
  once i get it into the other case, i will then redo the benchmark at stock speeds to see how well it scales.
  
  and thanks for the kudos on the o/c, it gets a bit warm but still under the "hot" point for these cpus. once in the new case a new hsf will go on. i have read in many different forums that these cpus seem to top out @ 3.0-3.4GHz, so getting 3.0 right out of the box was nice - not much messing w/ V, in fact i was able to lower it a bit from what "auto" had it set on and it is still prime95 stable and dropped 2C
  Like • Show 0 Likes0
  Actions
bob hern Feb 5, 2008 7:27 AM
Mark Correct
Correct Answer
the same setup @ stock speed (1.8GHz) runs the benchmark in 225seconds
Like • Show 0 Likes0
Actions

Was wondering how cpu cache played into SW

Outcomes

Related Content

Recommended Content