Best CPU for excel 2010?

[Arg Clin]

I'm doing some rather heavy work in excel. My current xeon W3503 feels like it's not quite up to the tasks as it takes forever each time it has to recalculate some of the larger projects I have.

I haven't been able to find any up to date excel specific benchmarks of current gen cpus online, so I'm hoping some of you guys might have some tips.

What is the best workstation class cpu for heavy excel? I am thinking maybe the W3680 (6-core, 3,3Ghz, 12Mb cache) but I'd very much appreciate a second opinion.

 

[Idontcare]

Anandtech Excel 2007 benchmark results (lower is better).

Anandtech Excel 2007 Monte Carlo benchmark results (lower is better).

 

[Arg Clin]

Thanks, I apreciate it

However I seem to recall from experience that Excel 2010 behaves very differently than 2007. Excel 2007 never went over 50% on my two cores, while 2010 goes all the way up to full advantage.

So I was really interested in specific results or experience with Excel 2010.

Or maybe I'm wrong and they do behave similar..?

 

[Tsavo]

http://www.anandtech.com/bench/CPU/2

Anand does Monte Carlo benches. Select what CPU you want to look at and go from there. He doesn't have Xeon in that list, so you can look at the 980X which will be just a hare faster than a W3680.

 

[Tsavo]

http://nl.hardware.info/reviews/236...s-microsoft-excel-2010---montecarlo-benchmark

 

[Ferzerp]

This isn't exactly what you asked, but have you considered porting your worksheets to a database engine? Once things reach the point you are describing, it often makes sense. However, if you don't understand DBs, you can very easily create something that works, but is no faster. If you do it well though, you should notice a world of difference. BTW, no, not Access.

 

[sm625]

lol if a W3680 isnt fast enough then you're basically SOL. There is nothing that will get you more a noticable improvement. Even a 2600k on LN2 probably wouldnt run noticably faster. You'd still be waiting. Just multitask.

 

[Arg Clin]

This isn't exactly what you asked, but have you considered porting your worksheets to a database engine? Once things reach the point you are describing, it often makes sense. However, if you don't understand DBs, you can very easily create something that works, but is no faster. If you do it well though, you should notice a world of difference. BTW, no, not Access.

Yes I'm very much considering this, but lack the skill. I'm not really a programmer by any standard, just have some basic knowledge of excel programming. So I wouldn't try to take on DB porting myself.

At some point I might have someone port it for me, but i'm not quite there yet. Also I'll still need excel for development.

Thanks for the input though.

 

[Arg Clin]

http://nl.hardware.info/reviews/236...s-microsoft-excel-2010---montecarlo-benchmark

Thank you. Much appereciated

 

[Tsavo]

I tried to find something that would approximate the lack of speed that 3503 has, but wasn't successful.

Suffice it to say, however, that a 3680 will absolutely blow your doors off.

 

[BlueBlazer]

At only 2.4GHz and 2C/2T, of course it would be slower than anything higher than that. Even the W3680 with 3.33GHz and 6C/12T would certainly be much faster. You may want to check your motherboard specifications and see if it can support a higher CPU though (and possibly a BIOS update may be required). :hmm:

 

[nyker96]

i think i3s do pretty well in excel benchmarks. but since no one has a very complicated sheet i'd bet most modern cpu can do excell quite well, even amd line up. but if have a extremely tough spreadsheet that takes forever to calc go w/ intel cpu. they do better in excel.

 

[dbcooper1]

I think 2007 was the first version that was multithreaded to any significant degree; does anyone have an example of a spreadsheet that takes long enough to recalc or run a macro that some of us could benchmark with various CPUs and report results?

 

[Arg Clin]

Here's something odd:

I took it home and tried it out on my PHII 4X 965. Definitely way faster than on the W3503, but it seems like it's only using 2 cores on the PHII. Doesn't make sense, since it's set to use all 4 cores.

I'm not sure if 2007 was properly multithreaded. On the W3503 it would run with 50% load on each core.

 

[Idontcare]

I'm not sure if 2007 was properly multithreaded. On the W3503 it would run with 50% load on each core.

Your concerns appear to be well-founded: http://forwardthinking.pcmag.com/so...ti-threaded-but-does-it-help#fbid=B8F9GuTEr1M

Indeed, the Excel team is very clear that their efforts on multithreading in Excel 2007 mostly centered on basic calculations, the kind you would do in a MonteCarlo simulation or a basic worksheet. They didn't particularly work on multithreading a number of the areas that seem to be behind most of the models I really find slow - data tables, pivot tables, external data links, and certain VBA scripts. While the team won't commit to specific improvements in the future, they were clear they saw the multi-threading in Excel 2007 as just the start of the road to taking advantage of multiple cores; they are looking to improve performance in future versions.

 

[BlueBlazer]

Then you may want to check this review. Its one the very few comparisons that pits top CPUs against lower end CPUs. :hmm:

That Core i3 2120 should be a lot faster than your Xeon W3503, core-to-core wise (also considering the former clocks are much higher). And there's also that A8-3850 which performs similarly to your Phenom II X4 965.

 

[dbcooper1]

One thing to keep in mind is that depending on the spreadsheet, Excel can spend a lot of time moving blocks of memory around which won't be influenced as much by number or cores or CPU speed.

 

[Arg Clin]

One thing to keep in mind is that depending on the spreadsheet, Excel can spend a lot of time moving blocks of memory around which won't be influenced as much by number or cores or CPU speed.

That would explain a lot of things. I think the W3503 is coupled with PC10600 so I'm out of luck if this is the case.

It's not that my calculations are very complex, it's more a case of very large datasets needing to be relatively simply processed in various ways.

 

[Broheim]

That would explain a lot of things. I think the W3503 is coupled with PC10600 so I'm out of luck if this is the case.

It's not that my calculations are very complex, it's more a case of very large datasets needing to be relatively simply processed in various ways.

just out of curiosity, how big are the files?

 

[Vdubchaos]

Are we really talking about this? Performance in Excel?

ROFL

 

[exar333]

Are we really talking about this? Performance in Excel?

ROFL

Yes. Lot's of people use Excel every day.

 

[dawks]

Are we really talking about this? Performance in Excel?

ROFL

You haven't spent much time in Excel hey? I've seen spreadsheets where if you change one number it can take up to 15 minutes to recalculate everything.. or try opening and editing a spreadsheet with several thousand items.

The Phenom will probably be faster for some things, just based on the fact that its clocked almost 1Ghz higher (in this case at least).

Beyond those benchmarks linked, I'd suggest getting the highest clock dual/quad core Intel CPU you can, and get one with turbo boost, that way if it ever has to run single threaded, you'll get a speed boost. The Core i5 2500 can turbo up to 3.7Ghz from the base of 3.3Ghz. Speedy little chip.

But as suggested, it may be in your best interest to switch to some more advanced software.

http://ark.intel.com/products/family/59134/2nd-Generation-Intel-Core-i5-Processors/desktop

 

[Arg Clin]

The largest I'm working with has something along the lines of 100.000 rows x 60 columns (on avg) x 9 sheets. So that's 54.000.000 cells that needs processing.

Ofcourse some of them are empty or passive - a rough estimate is I'm up to around 20% of those being actually filled with data or code at the moment.

So it's not looking too bright. Guess I might have to look into something else, just hate to lose the ability to just tweak the stuff myself whenever I need to.

 

[paperwastage]

The largest I'm working with has something along the lines of 100.000 rows x 60 columns (on avg) x 9 sheets. So that's 54.000.000 cells that needs processing.

Ofcourse some of them are empty or passive - a rough estimate is I'm up to around 20% of those being actually filled with data or code at the moment.

So it's not looking too bright. Guess I might have to look into something else, just hate to lose the ability to just tweak the stuff myself whenever I need to.

other bottlenecks? eg matched RAM sets might help/more ram, ssds etc

 

[Davegod]

Arg Clin, try this link, and the ones in the summary of the article.

BTW someone mentioned the i5... I think you'd want the i7 as listed in the benchmark. It has hyperthreading and more cache, both of which I suspect Excel can take advantage of.

ps. you've tried turning the background refresh off?