Excel speed test comparison against an FX-8350

[John Tauwhare]

Oh come on, I've had mine up to 1.63v now, and it's an Ivy Bridge. (that's what, 0.11 over Intels supposed recommended max?) You have to keep going till you beat me again.

Ok, WTF, I pushed a bit further. With 1.5V I got 5.3 and 5.4 and with 1.53V I got 5.45GHz (52 x 104.8). At 5.5 and 1.55V I just got lockups and BSODs so I'd need to change something else. I'm quite happy with 21.95 though!

 

[John Tauwhare]

Edit: chart update Feb 5, 2013 (v9). SB, amazing or what?

 

[John Tauwhare]

Funny how we get caught up in the OCing cycle

Yes, sort of addictive. Gee, I need to get a life...

 

[Vectronic]

Damn, that's gonna be a tough one to beat. I suspect about 5.1GHz on my CPU... I never got 5.2 stable, but other than 4.5GHz I usually do it in 200MHz increments, so didn't try 5.1.

 

[Makaveli]

Chart updates Feb 5, 2013. SB, amazing or what?

i'm very shocked it took a 2700k at 5.4Ghz to just barely beat an Ivy chip at 4.8 !!!

I will be interested in seeing what Haswell will do here!

 

[Vectronic]

Well, to be fair that's roughly a 4% increase over my 4.8GHz... his at 5.25 would probably "barely beat" mine.

Ivy ~9% > Sandy... that's kind of par for the course.

This is a tiny test, with with rather dramatic fluctuations of about 3%... amusing, but not an indication of much really.

There's a 2500K and a 3570K both at 4.5GHz, basically the same times, 2500K slightly in the lead. Another 2500K @ 4.5GHz, that's 1.5 seconds slower, a 3570K @ 4.5 that's 2 seconds slower... so in those cases, Sandy > Ivy.

 

[Dadofamunky]

Wow, my 2600K in my sig (which is now running at 4.35 GHz) posted a 28.57 in the initial run. That is a hella nice spreadsheet.

 

[dave_the_nerd]

So I'm guessing it's extremely single threaded?

Well, on my machine when I hit "go" Excel pegged to 100% CPU use and had 8 threads.

My time was piss-poor, but that happens sometimes to us Mac OS folks.

So it would probably benefit from a quad core anyway. Since the 8-core AMDs only have 4 FPUs, maybe that's an issue?

 

[SPBHM]

this uses little over 1 core here... forcing it to use 1 or 2 threads, and it's clear that there is not a big difference from 1 to more cores.

 

[John Tauwhare]

Wow, my 2600K in my sig (which is now running at 4.35 GHz) posted a 28.57 in the initial run. That is a hella nice spreadsheet.

Thanks Dadofamunky. The time for your 2600K is almost exactly the same as mine at that speed (interpolated). your time is in the chart (above).

 

[Face2Face]

3570K @ 4.8Ghz 25.43 - - Not sure why it's not faster..??

 

[John Tauwhare]

3570K @ 4.8Ghz 25.43 - - Not sure why it's not faster..??

Strange

, but good you've got a chip that's fit for purpose now.

 

[Vectronic]

3570K @ 4.8Ghz 25.43 - - Not sure why it's not faster..??

What version of Office, what version of Windows?

Windows 8 and/or Office 2013 makes a fairly big impact it seems. (so would a virus scanner/HIPS that's double-checking what Office is doing).

You might not also be completely stable so it's running slower because of error correction.

You have comparable RAM to mine, so that shouldn't be an effect.

 

[soccerballtux]

I recently built a new rig inside an old Dell XPS 710 case, converted from BTX to ATX, with an FX-8350 in a Gigabyte 990FXA motherboard and 8Gb RAM. This is my main PC and I use it for many things but mostly for spreadsheet work. And I had high hopes that the latest CPU from AMD would would be much faster at running Excel macros than the 5 year old QX6700 that used to be in my Dell case.

But, though extremely quick at graphics based applications, it didn't seem any quicker at Excel. So, to compare its performance I wrote a timing macro into an old spreadsheet and timed my new machine, at 4GHz, at 60 seconds. Then I overclocked it to 4200-4400. No change, still 60 seconds.

I timed all the laptops I could get hold of. My work Lifebook P770 (i7-640UM @1200/2266) did 102 seconds and my ultra-mobile Lenovo U260 (i5-470UM @1333/1866) did 172 seconds. No wonder they frustrate me so much; and it's not like the battery life is so great despite the low voltage. I now know to avoid anything with a "U" in it.

I timed the Lenovo E320 (i3-2330M @2200/2400) I gave my daughter for Christmas, which I thought would be nothing special, and it did 54.6 seconds. But the star was my Samsung R780 (i5-480M @2666/2933) at 49 seconds. Even my trusty Intel QX6700 at stock speed of 2666 in an ASUS P5K with 4Gb DDR2 was faster than the FX-8350, at 59 seconds, and overclocking it to 3200 dropped its time to 53 seconds.

I'd love to know what's going on here and why AMD's flagship Piledriver CPU is no faster at running Excel macros than a 5 year old Intel Extreme CPU and is considerably slower than a 2 year old mid-range Intel based laptop, both at much lower clock speeds.

The test spreadsheet is here:

http://dl.dropbox.com/u/87381454/EXCEL-SPEEDTEST-v1-(John-Tauwhare).xlsb

If anyone wants to have a go then click on the Stopwatch button and wait ~3 minutes. Please PM or post the elapsed time and I'll do a results table. And if anyone can beat 40 seconds then I'll know what to buy next.

can we do this in Open/LibreOffice instead?
I'd like to see the comparison between the CPUs using code that we know isn't compiled with Intel's math libraries.

 

[james1701]

980X 4.3ghz
Office 10
Windows 7

 

[Vectronic]

can we do this in Open/LibreOffice instead?

If you want to convert it, go for it... I made a small attempt at it, gave up.

Office is being used by the OP... it was originally just about figuring out which processor would work best with that... turns out... Intel. Problem solved so it kinda became a more generalized Intel benchmark for a few days.

 

[John Tauwhare]

can we do this in Open/LibreOffice instead?
I'd like to see the comparison between the CPUs using code that we know isn't compiled with Intel's math libraries.

I had a go. I'm not familiar with LibreOffice but I installed the Calc app to try it out. It seems to retain cell range names but doesn't recognise formula references to them. It would take ages to re-establish all the trashed formulae. I gave up, sorry.

 

[John Tauwhare]

Tried really hard today to get beyond 5.45G but no combination of anything I did would hold it at 5.47G. I don't know what to do next; I'm not running into error correction, or high temps (~60C), and Vcore is only 1.52V. I guess it's on the limit. But something went right because my times improved a bit. At 5.45G I got 21.38.

 

[John Tauwhare]

Updated chart (v13):

 

[Vectronic]

Oh you prick ...lol

If I have time after work, I'm gonna try and beat that.

 

[John Tauwhare]

Oh you prick ...lol

If I have time after work, I'm gonna try and beat that.

Good luck with the delidding....

 

[Vectronic]

Yeah pretty much... did a quick go for it at 4.9GHz... beat my previous score, however I know this is slower than it could be since 2 of my 4 cores throttled, and I was up around 1.58v.

 

[Face2Face]

What version of Office, what version of Windows?

Windows 8 and/or Office 2013 makes a fairly big impact it seems. (so would a virus scanner/HIPS that's double-checking what Office is doing).

You might not also be completely stable so it's running slower because of error correction.

You have comparable RAM to mine, so that shouldn't be an effect.

Window 7 Ultimate and Office 2007 - Does office 2010 run it better?

I would like to get this score down, because I know I can get 5.0Ghz at around 1.35 and it should be stable enough to run this test.

My current setup is very stable - Not sure how to check error correction. Looks like this test does not stress my cpu very much at all - Is it single threaded?

 

[John Tauwhare]

I tested a few different scenarios just to see what, if anything, impacted the performance of Excel running this app.

The Office 2010 vs Office 2007 results are as follows:

Office 2007 i7-3770K 4.0GHz DDR3-1866 33.41s, 33.08s, 33.33s (repeated 3-times) = 33.28s avg
Office 2010 i7-3770K 4.0GHz DDR3-1866 32.81s, 31.75s, 33.57s (repeated 3-times) = 32.71s avg


Office 2007 FX-8350 4.0GHz DDR3-1866 60.89s, 61.46s, 61.35s (repeated 3-times) = 61.23s avg
Office 2007 FX-8350 4.0GHz DDR3-1866 66.61s, 68.29s, 66.31s (repeated 3-times) = 67.07s avg

Conclusion: Office 2010 is neither faster nor slower with this macro when running on the Intel 3770k, but it takes 9.5% longer to finish on the AMD FX-8350 compared to running the same macro with Office 2007.

Seems to be no difference between Office 2007 and 2010. I'm using 2010 now. When I reinstalled 2007 after switching from the FX-8350 Microsoft told me I'd used up all my activations.

The test only uses about 40% CPU load. I assume that because the gradient of the GHz v Seconds chart is quite constant that I'm not running into error correction. It is very single threaded and Hyper Threading seems to be a burden (~6%) so you should disable it.

 

[John Tauwhare]

I'm a nervous wreck now but I finally but I got a run at 5.5GHz. Reckon that's the best this setup will do (21.33s)