Sign in with
Sign up | Sign in
Your question

I5-2500 vs i7-2600 for Solidworks

Last response: in CPUs
Share
June 11, 2011 9:59:21 PM

I am planning to (in about a month) build myself a computer for gaming, solidworks and matlab, and I will probably not OC past what I can do without changing voltages. It is clear that for gaming the i5-2500 (or k version) is the best performance for the price, but I wanted to figure out if the i7-2600 had anything substantial to offer for solidworks or matlab. I haven't found a good source of matlab's benchmark results yet (only this article http://www.mathworks.com/matlabcentral/newsreader/view_thread/296005) so this thread will be about the solidworks benchmarks.

A searchable list of solidworks benchmarks results can be found here:
http://www.solidworks.com/sw/support/shareyourscore.htm

I have organized them by processor and made some graphs showing results.
All scores are in seconds (lower is better).
The markers correspond to average scores.
The error bars with end caps are one standard deviation from average.
The error bars without end caps are max and min scores.
Each graph is labeled by what the solidworks website called the benchmark.



I expected to see (especially on the cpu scores) tighter groupings for the non-K version of each CPU in all of these benchmarks, as varying amounts of overclocking on the K versions should spread out those results. However, this only really happened on the "Render" benchmark. However, I was surprised by how tight groupings were on the 2500K, not even overlaping 2500 results on "CPU" or "I/O".

In cases where the averages are far apart the 2500K wins over the 2500 by amounts that don't seem reasonable (unless everyone OC'd the crap out of their 2500K). If I didn't realize the K version of each chip was so similar to the normal version, these results would lead me to buy the i5-2500K, but those differences do not seem reasonable, so I figured I should look for an explanation. Also the 2600K has an extremely large spread in all results.

Maybe better results came from computers in which better parts went in overall. I don't have very much information about the computers these cpu's were in, or how much OC'ing was done, so the best I can really do is a histogram of GPU's based on price (widely varying ages and some being professional made comparing benchmarks unreasonable).

I'm going to assume that more expensive GPU's not only would do better against cheaper GPU's but also correspond to the other parts of the computer being better.
A price of 0 means something like "VNC Mirror Driver" was written in the GPU column.
With the exception of the 2500 (non K) all histograms are scaled to have the same total area filled in with bars, so height corresponds to percent in a comparable sense.
Here are the histograms:


Note that for the i7-2600K the expensive graphics card, the ATI FirePro V7800 (FireGL) corresponded to the best results where on the i5-2500K it corresponded to very mixed results (even though the guy who used it in the 2500K was kind enough to write in that he OC'd his to 4.6 GHz.

These histograms do seem like enough to imply that the 2500K's were on average put into better computer's than the 2500 and explain the huge differences in benchmarks. They also explain the large spreads on the 2600K benchmarks.

It looks like on average the 2600 (k or not) isn't better by enough to buy it and large spreads from average seem to be explained by GPU.

I'd be interested in hearing more ways of interpreting these results as well as opinions on whether the 2600 has anything to offer.

More about : 2500 2600 solidworks

June 11, 2011 10:00:19 PM

Dammit, my links and pictures didn't work, I'll have to fix that.

Edit: fixed them
a c 309 à CPUs
June 12, 2011 3:36:42 AM

The 2500 and 2500K are exactly the same, except that the K can be overclocked.
If no overclocking is involved, the benchmarks should have been precisely the same.
I would assume that there was some level of OC involved to show such improved results.

Ditto, the 2600 and 2600K. The 2600 and 2600K start at one multiplier higher, and they have more cache. They also support hyperthreading.
the 2600K costs about $100 more than the 2500K.

Each chip can be different, anf the OC you can expect will vary. To my mind, you can expect 4.0 as a conservative oc using the voltage on auto,
and 4.5 as an aggressive oc. A good cooler will let you do better. Both the 2500k and 2600k should oc to about the same levels.
Some think the 2600K is bin selected, and will OC higher, but I have not seen anything definitive on that theory.

It is not clear to me how to value the added cache on the 2600's. It may depend on the application.
If your app can use more than 4 threads, then the hyperthreading of the 2600/k will be helpful.
Few games use more than 2 cores, let alone 4. That is why the 2500K is more attractive to gamers.
I estimate each hyperthread acts like 1/3 or 1/4 of a full thread.

If I recall correctly, matlab was single threaded. There may be a multi thread implementation of some sorts available now.
The difference you saw between the 2500 and 2600 was more than one higher multiplier would account for.
I suspect that the benchmark fit in the 2600 cache better than the smaller 2500 cache.

---------------bottom line--------------------------
If the $100 does not mean that much to you, then by all means get the 2600K.
Related resources
a b à CPUs
June 12, 2011 4:03:43 AM

Hi Geofelt!
The i7 2600k also has 2mb extra cache.
a b à CPUs
June 12, 2011 2:52:16 PM

I think Solidworks, and games too, will definitely be limited by your graphics card, no matter which of the K series you go with. Therefore, you should get the 2500K and put the extra $100 into a better GPU, or just save it. Or give it to me.
You MUST get a K and overclock like crazy.
June 27, 2011 1:30:38 PM

I'm a little late to this discussion, as someone on the SolidWorks forum linked here.

This is a very interesting comparison, and it doesn't quite match up with the results seen before. Thank you for compiling this data. To note, there is another SolidWorks benchmark that only calculates "part rebuild times", and rendering separately. There are four of them, and they are available here: http://www.solidmuse.com/benchmarks/

To specifically reply to the last post by kajabla: SolidWorks is not heavily dependent on the video card. It is typically much better to spend the money on the CPU, and skimp on the video card. SolidWorks models are nowhere near as complex as a high-end 3D video game; thus they don't need as beefy of a video card. I typically suggest the ATI FirePro v4800, which is above and beyond what most users need.

So I don't actually have any additional insight, as I haven't heard of many users trying to use the i5. I am a professional, and most other professionals (and their companies) are willing to spend the $100 to get the added bonus. But Ekid2k, if you did buy the i5, please run the SolidWorks benchmark (and the solidmuse.com benchmark above), and let us know what score you got. I will be interested to hear more.
June 28, 2011 5:22:51 AM

charles1culp said:
I'm a little late to this discussion, as someone on the SolidWorks forum linked here.

This is a very interesting comparison, and it doesn't quite match up with the results seen before. Thank you for compiling this data. To note, there is another SolidWorks benchmark that only calculates "part rebuild times", and rendering separately. There are four of them, and they are available here: http://www.solidmuse.com/benchmarks/

To specifically reply to the last post by kajabla: SolidWorks is not heavily dependent on the video card. It is typically much better to spend the money on the CPU, and skimp on the video card. SolidWorks models are nowhere near as complex as a high-end 3D video game; thus they don't need as beefy of a video card. I typically suggest the ATI FirePro v4800, which is above and beyond what most users need.

So I don't actually have any additional insight, as I haven't heard of many users trying to use the i5. I am a professional, and most other professionals (and their companies) are willing to spend the $100 to get the added bonus. But Ekid2k, if you did buy the i5, please run the SolidWorks benchmark (and the solidmuse.com benchmark above), and let us know what score you got. I will be interested to hear more.


I'll make sure to run the benchmarks on my computer once it's built (only buying the pieces that go on sale till I can afford the whole thing mid july).

Also, I am very interested in seeing what on the solidworks forums linked you here.
!