Simple little favor if someone would be so willing...

J

Joe Average

Ad Blocker - Banned
Joined
Apr 6, 2008
Messages
15,459
I'm interested in perhaps doing an i5 2500K build and perhaps looking at some overclocking since it seems to be so relatively easy, but I'm curious about the actual performance. I know there's benchmarks all over the place, but if someone would be so kind - if they're running a 64 bit version of Windows, preferably Windows 7 - to run a very simple benchmark for me. Ok, two actually.

I've been using an older version of Geekbench32 and Geekbench64 for a few years now because a) it tests CPU and RAM primarily and also how quickly the OS itself can execute the instructions, and b) it takes like 15 seconds to run. It provides a "score" that I can use for comparison based on some builds I've done in the past, single cores, dual cores, triple cores, quad cores, etc.

I'm a big fan of doing x264 encoding these days so I'm interested in raw processing power, and the ease that people seem to be having by getting 4.2-4.4 GHz off the 2500K with stock cooling is pretty amazing.

Anyway, here's the archive that contains the two .exe files - I swear to a higher power these are not infected files, but just in case I just rescanned them using VirusTotal, an online virus scanning service that checks a given file against 42 current antivirus scanning products. You can find the test results of the scan I just ran (and get the hash checksums as well here):

32 bit .exe - http://www.virustotal.com/file-scan...8ddc681414533c5ff4da183034a14ccb96-1295618675

64 bit .exe - http://www.virustotal.com/file-scan...5b7e944896e2be22110537bf4c2ef24cfa-1295618671

Here's the Zip archive:

http://www.mediafire.com/?bpagbe1w3ka2gdl (890KB Zip archive, 2 .exe files inside)

Obviously if you can run this with as little else running in the background - like right after a reboot - that would provide the most solid results with nothing really interfering with the testing (and hopefully no background stuff like Folding@Home running sucking up the CPU resources).

It's simple: extract the files someplace, open a Command Prompt window (in Windows 7 you can hold Shift and right-click on a folder, it'll put the "Open command window here" option on that right-click context menu), and type:

geekbench32 (press Enter, wait for it to be done, press N to submit results)

then

geekbench64 (same process)

and then post the results for each. Since I don't have access to a current SB build I just wanted to see how well they score. Oh, oh, one more thing (no Steve Jobs cracks, please), if possible can you test at stock speed of the CPU/RAM and then your current overclocked settings as well (if you're overclocking, of course).

Thanks in advance to anyone that can help... and people that aren't overclocking at all are welcome to post their scores, too. Would love to see some scores and their associated builds (CPU speeds, RAM speeds and timings, that sort of thing).
 
2500k @ 4.4 ghz currently (seriously anyone can run at this speed)

32 bit - 952.6
64 bit - 1029.3

I didn't reboot, few things open browsers and rdp, nothing CPU intensive... thought I'd give you a ball park, how does it stack up?
 
Very nice, about what I expected in reality. On my old Q6600 box (stock 2.4 GHz) and it pulled about 408 or so consistently, and about 525 when pushed to a solid 3 GHz; the 25% greater speed resulting in an almost perfect scaling of score as well. It's another reason I like that little benchmark so much, the results are very repeatable, and it's properly multithreaded too.

Thanks for the input.
 
Nice, if no one else chimes in I can do some more testing later. Just don't have time right now, have some stuff open I can't close, etc.
 
Excellent, will look forward to it if you can manage it. Mind if I ask what cooling you're using (stock or aftermarket HSF) and what the temps are if you're pushing a full load on the 2500K? Thanks...
 
Using a CoolerMaster 212+ can pick on up for $20-$30. Currently at 1.3v @ 4.4 ghz. I'm playing around with voltages and overclocks looking for the sweet spot. I've ran prime overnight on this overclock twice now with both Small FFTs and In-place large FFTs. Passes just fine. Temp usually maxes out at this range in the low 60's. I only have a single fan on the 212+.

http://www.amazon.com/Cooler-Master...YPH0/ref=sr_1_1?ie=UTF8&qid=1295621560&sr=8-1
 
As an Amazon Associate, HardForum may earn from qualifying purchases.
Purely because the question of how does X processor compare to SB X processor comes up a lot now, I ran it on my i7 920 which is currently running 3980 Mhz.

32bit: 919
64bit: 1021.9
 
Wow... that's pretty damned impressive that the "old school" i7 can hang with the new kids on the block, so to speak. ;) Hell of an overclock as well, is that something you just recently did or has it been running at that speed for a long period of time, if I may ask?

Thanks for the input.
 
Ran with my 2600K @ 4.8 GHz:

64 bit: 1243

Didn't bother with the 32 bit.
 
Joe Average said:
Wow... that's pretty damned impressive that the "old school" i7 can hang with the new kids on the block, so to speak. ;) Hell of an overclock as well, is that something you just recently did or has it been running at that speed for a long period of time, if I may ask?

Thanks for the input.
Click to expand...

It always runs around that, I messed with it a bit to see if I could get it higher from 3.8 in my sig but stay at stock V, and thats what I hit. I think Im going to see how close I can get to 4.4Ghz now and check the score apples to apples, or as close as I can get :p
 
Luckily my girlfriend is not home, heat is off, window is wide open and its negative 10 degrees outside!
 
Doubl3KiLL said:
Purely because the question of how does X processor compare to SB X processor comes up a lot now, I ran it on my i7 920 which is currently running 3980 Mhz.

32bit: 919
64bit: 1021.9
Click to expand...

The bench seems to scale pretty linearly. Running a Core i7 920 at 3900 MHz on a 52-hour boot came up with the following:

32 bit: 943.3
64 bit: 1042.9


-R
 
2500k @ 4.5 (current 24/7 OC)
No reboot. Browser open, but doesn't seem to affect the benchmark much.

32 bit: 1032.2
64 bit: 1118.7
 
Well Im still pushing. doing some stability testing right now, I found that I can increase the overclock better with.... a decrease in VCORE..... o_O. Should be a fun day now!
 
Here ya go Joe...

2600k @4.5
Coolermaster 212+

32 bit: 1110.3
64 bit: 1262.6

This program really didn't seem to push the CPU as the the hottest core only reached 51c. Under Prime95 it will be high 60's to 72c depending on the room temp.
 
Well, it's a short benchmark and it's a limited set one also: it runs a specific task to completion and then it's done, similar to what something like SuperPi or wPrime does as well. It's short, fast, repeatable results, stable, and more useful to me for getting a very quick idea of how a given machine will perform strictly from the CPU/RAM perspective as those are the two components really tested.

And besides, most of the 2500K powered boxen doing this will finish it in about 15 seconds, I'd suspect, so not enough time to really have any serious alteration on the operating temps - with Prime95 you just let it run and run and run...

One thing I've noted is just how much RAM timings can affect Geekbench scores (both versions): I had a machine with 9-9-9-18 RAM in it in the recent past (just DDR2) and had a score someplace close to 250 with a Core 2 Duo running at 2.4 GHz. I modified the timings to 8-8-8-15 and the score jumped to 280... was pretty surprising that such a relatively minor change could alter it in such a drastic amount (~13% higher score).

But thanks for the input, everyone, it's very useful info on deciding what I'm going to put together. If only Sandy Bridge was rocking SMP capability yet... but I can wait. ;)
 
Joe Average said:
One thing I've noted is just how much RAM timings can affect Geekbench scores (both versions): I had a machine with 9-9-9-18 RAM in it in the recent past (just DDR2) and had a score someplace close to 250 with a Core 2 Duo running at 2.4 GHz. I modified the timings to 8-8-8-15 and the score jumped to 280... was pretty surprising that such a relatively minor change could alter it in such a drastic amount (~13% higher score).
Click to expand...

Win XP 32-Bit
Old E6600 Core 2 Duo OC @ 3.075Ghz with Kingston DDR2 800 Mhz memory 4-4-4-4-24-2T = 323.5

bg4dj4.jpg
 
Just to compare a cheap low end AMD quad core CPU, cost me 84.99 a month ago.
64-BIT Rana X3 Unlocked 4th Core. Stock Speed 3.3 Ghz Stock Cooler = 514.0
DDR3 1333 Mushkin SIlverline 2x2GB cost 39.99
89.99 Mobo http://www.newegg.com/Product/Product.aspx?Item=N82E16813157198
I'm satisfied with it. Wish I had SB level performance for benchmarks.
SB obliviously doubles the performance in certain tests,but it's close to 2.5x-4x the cost for me. 224.99 for the 2500k, 329.99 for the 2600k.
2i1plrc.png
 
Joe Average said:
One thing I've noted is just how much RAM timings can affect Geekbench scores (both versions): I had a machine with 9-9-9-18 RAM in it in the recent past (just DDR2) and had a score someplace close to 250 with a Core 2 Duo running at 2.4 GHz. I modified the timings to 8-8-8-15 and the score jumped to 280... was pretty surprising that such a relatively minor change could alter it in such a drastic amount (~13% higher score).
Click to expand...
If RAM timings have such a profound effect on the scores, then this benchmark is not really representative of real world performance.
 
Just for shits and giggles, I ran the 32-bit version on my Dell Laptop (T5500 Core Duo @ 1.66ghz).

Score: 175.1

I
 
TheInfernal said:
If RAM timings have such a profound effect on the scores, then this benchmark is not really representative of real world performance.
Click to expand...

So you're saying that adjusting the performance of your machine by altering the RAM timings - which is what people do to adjust the performance of their machines in the real-world - is meaningless in the real-world but at least this particular benchmark can prove the effect of such alterations of said timings quickly and easily, is that right?

/me trips over the loopy nature of that statement...

Ok, try this one on for size:

The machine I mentioned above was also used to test rendering performance using Cinebench 10 (this was about 1.5 years ago, Cinebench 11 is out now). With the 9-9-9-18 timings I had a score that was 13% slower than the same machine running the same test with the same OS but with 8-8-8-15 timings. Lower timings = measured with Geekbench as providing faster performance = measured with Cinebench as providing faster performance.

When dealing with incredibly complex math and number crunching - not just everyday computing tasks - RAM timings can have a rather dramatic effect on the potential performance capabilities because of the highly repetitive nature of such tasks, especially if the CPU has a decent amount of L2 cache going for it.

How's that for real-world performance?
 
Did mine as well just for giggles.
i7 - 950 @ Stock, 12GB 9-9-9-24 @ 1333

32bit: 768.4
64bit: 862.8


Also ran my old 775 system:
Q6600 @ Stock | 8GB DDR2 @ 800

32bit - 416.3
64bit - 420.8
 
Last edited:
Just wanted to say that TinyPic.com is free, plus you don't have to register to up(load) a Paint picture which takes about 10 seconds, and they give you the image code for this forums. (just paste it here where you type to see the image)

I'm interested to see a score from either a Q6600/Q8300/Q650 posted if anyone could.

Also anyone with a Phenom X6/1055T/1100T if they could do this benchmark.

Then we would have about all the current CPU's compared from 2006-2011 era for Desktops for this benchmark. But that's probably asking to much. :p
 
Joe Average said:
So you're saying that adjusting the performance of your machine by altering the RAM timings - which is what people do to adjust the performance of their machines in the real-world - is meaningless in the real-world but at least this particular benchmark can prove the effect of such alterations of said timings quickly and easily, is that right?

/me trips over the loopy nature of that statement...

Ok, try this one on for size:

The machine I mentioned above was also used to test rendering performance using Cinebench 10 (this was about 1.5 years ago, Cinebench 11 is out now). With the 9-9-9-18 timings I had a score that was 13% slower than the same machine running the same test with the same OS but with 8-8-8-15 timings. Lower timings = measured with Geekbench as providing faster performance = measured with Cinebench as providing faster performance.

When dealing with incredibly complex math and number crunching - not just everyday computing tasks - RAM timings can have a rather dramatic effect on the potential performance capabilities because of the highly repetitive nature of such tasks, especially if the CPU has a decent amount of L2 cache going for it.

How's that for real-world performance?
Click to expand...
I am surprised how such a small change in timings could result in a 13% performance gap. Even if this holds true to number crunching, I would still say this test is misleading because IIRC, most applications don't scale well with memory timings in C2D.
 
TheInfernal said:
I am surprised how such a small change in timings could result in a 13% performance gap. Even if this holds true to number crunching, I would still say this test is misleading because IIRC, most applications don't scale well with memory timings in C2D.
Click to expand...

I changed the DDR2 800 Mhz 4-4-4-4 timings, to 5-5-5-5 and received these scores.

4-4-4-4 = 323.5
5-5-5-5 = 327.4

3.9 points almost 4.0.

So tighter timings for a C2D @ 3.075Ghz = barely over a 1 percent total difference with very slight changes to memory latency depending on 4 or 5 Cas.

Hope that sheds some light. (Joes memory is probly a little off...He said it's been a few years) :p
20t2kw1.jpg
 
Ran my old machine:

Q6600 @ Stock | 8GB DDR2 800

32bit - 416.3
64bit - 420.8
 
You must log in or register to reply here.
Back
Top