C'mon guys...how about compiler tests!

BoB-O TiVo

Limp Gawd
Joined
Mar 25, 2004
Messages
147
Hey guys,

I love your reviews, but I'm a software developer. I'd love it if I could see comparisons on CPUs when compiling code. Pick your favorite open source project...Mozilla...Linux kernel...whatever. Just something that takes more than a minute or two should show the differences.

Thanks
BoB
 
I could do a full build of Paint.NET if you'd like :) Usually takes ~3 minutes to build and package up all the setup files on an X2 4800+. I have an X6800 here that seems quite happy at 3.46GHz.
 
rolo said:
I could do a full build of Paint.NET if you'd like :) Usually takes ~3 minutes to build and package up all the setup files on an X2 4800+. I have an X6800 here that seems quite happy at 3.46GHz.

Nice, I too would be curious to hear the results.
 
Full rebuild of Paint.NET v2.64:
Athlon X2 4800+ @ 2.4 GHz takes 2 minutes 40 seconds.

I'm in the process of copying everything over to the Conroe, and have to install VS2k5, etc. Give me a few minutes :)
 
BoB-O TiVo said:
Hey guys,

I love your reviews, but I'm a software developer. I'd love it if I could see comparisons on CPUs when compiling code. Pick your favorite open source project...Mozilla...Linux kernel...whatever. Just something that takes more than a minute or two should show the differences.

Thanks
BoB

QFT!

I'd love to see a freebsd make buildworld with -j3 -j4 -j5 etc and see what it scales up to!
 
Robstar said:
QFT!

I'd love to see a freebsd make buildworld with -j3 -j4 -j5 etc and see what it scales up to!
If I have time I'll throw BSD on a spare HD when I get my machine to see how it works.

But until then I just aquired a 133mhz p1 that I'm going to put 6.1-RELEASE on :)
 
Ok I finally got everything set up correctly, it took longer than I thought.

Athlon X2 4800+ @ 2.4 GHz takes 2 minutes 40 seconds
Core 2 Extreme X6800 @ 3.46 GHz: 2 minutes 10 seconds

A large portion of the process is compressing an 80 MB file down to 49 MB using LZMA compression. That part of compiling takes 1m:45s on the X2, and 1m:26s on the overclocked Conroe.
 
rolo said:
Ok I finally got everything set up correctly, it took longer than I thought.

Athlon X2 4800+ @ 2.4 GHz takes 2 minutes 40 seconds
Core 2 Extreme X6800 @ 3.46 GHz: 2 minutes 10 seconds

A large portion of the process is compressing an 80 MB file down to 49 MB using LZMA compression. That part of compiling takes 1m:45s on the X2, and 1m:26s on the overclocked Conroe.

Wow, uh, that doesn't quite seem to scale how I would expect. What gives? Bandwidth limitations?
 
Yeah, I can see where intel has most likely locked in their workstation and server market with the release of Core in the face of so many folks clamoring for Opteron solutions from large scale enterprise vendors (ie Dell).

Those are seriosuly jaw dropping results at the higher frequencies. But even at lower frequencies intel seems to have closed the gap that Xeon vs Opteron left wide open.
 
thedude42 said:
Wow, uh, that doesn't quite seem to scale how I would expect. What gives? Bandwidth limitations?
Let me do some more number crunching to make sure I got these right. I'll test at stock speed on the Conroe as well.
 
Now that I think about it, the timestamping could be a problem. Some of our binaries that we build are digitally signed, and then we timestamp them. Digital certificates expire but this allows whoever issued the certificate to put their mark on the signature to say "Yes this file was indeed signed before the certificate expired." This requires some communication to a network server somewhere, and the Athlon and the Conroe are on two separate networks (my Athlon is at home, the Conroe is here at work). Obviously various aspects of network latency could affect how fast this process completes. I'll rerun w/o the timestamping and see if that makes a difference.
 
Ok I get the same #'s without timestamping.

If I run w/o the large LZMA compression job, which is honestly something I don't believe is typical for a compile/build process, here are the numbers I get:

Athlon X2 4800+ @ 2.4 GHz: 45 seconds
Conroe @ 3.43 GHz: 44 seconds

(plus or minus 1/2 second or so, so ... basically they score the same)

If I do a build with the IDE, I get roughly the same results but I notice that the C++ parts of the compilation process are noticably faster on the Conroe. Please note that the Athlon has a Raptor 150 while the Conroe has some typical 7200RPM SATA2 drive, maybe that's the key difference. Both have 2 GB RAM.

I'm going to try and find a C++ project to compile, and see what I can come up with. Any suggestions? These are Windows XP x64 boxes, please don't ask me to compile some Linux stuff. A Visual Studio 2003 or 2005 .SLN file is what I'm looking for here.
 
Hmm I'm interested in this as well. I stress my computer with calculations more than anything (SPSS, Excel, and MATLAB). When you're dealing with a data file that's 100MB; sounds like the Conroe would be good!

Anyone care to sort a 50x66536 Excel file and report the results?
 
To add some more data: on the project I am paid to work on, this 3.46 GHz Conroe is roughly twice as fast at doing "full build" as my old system, a dual 3.06 GHz Xeon w/ HyperThreading.

"Full build" consists of doing a clean, full debug build, then a 2nd clean, then full retail build.
 
rolo, if you want to eliminate the disk bottleneck, put /usr/src on mfs under freebsd :)

then do a make -j3 buildworld KERNCONF=GENERIC in freebsd 6.1

replace 3 with 4, 5, 6, 7, 8 etc.
 
Compiling like above takes 44-45 seconds on the Conroe at 2.93ghz as well. Definitely not CPU bound at all, d'oh! :p
 
Can you create a large enough ramdrive under windows to get around disk access bottlenecking?
 
rolo said:
I doubt that would make for a good benchmark ;)

Well if both environments were under vmware, it would at least be equal :)
if someone sends me a conroe, ill do the benchmark & send it back :)
 
I find it quite hard to belive that a conroe at 3.4ghz would be less than a X2 at 2.4ghz

Everything has to be fair , the raptor is probably the problem , i wanr to see it used with a 7950 .

I am buying one definately with some ddr800 with tight timings and i canr wait .

Its still going to be a forthnight before i can get it together.

I want to see some gaming benchmarks with the conroe overclocked at 3.5ghz
 
Robstar said:
Well if both environments were under vmware, it would at least be equal :)
if someone sends me a conroe, ill do the benchmark & send it back :)


i doubt it would be fair... doesnt conroe have hardware virtualization?
 
nooh said:
I find it quite hard to belive that a conroe at 3.4ghz would be less than a X2 at 2.4ghz

Everything has to be fair , the raptor is probably the problem , i wanr to see it used with a 7950 .

I am buying one definately with some ddr800 with tight timings and i canr wait .

Its still going to be a forthnight before i can get it together.

I want to see some gaming benchmarks with the conroe overclocked at 3.5ghz
I was just publishing the numbers while at the same time scratching my head wondering why they were the same. Not trying to say Conroe isn't fast :) Compiling C# code like Paint.NET just seems to be disk bound more than anything, and the X2 has a pretty fast hard drive. If I had a RAM drive utility I'd probably do like thedude42 suggested and run the compile in there. Anybody got a link to one? It has to work on XP x64.

This X6800 has a Radeon X1900XTX in it, it runs games great. It plays Oblivion noticably smoother at 1920x1200 than the X2 4800+ / GeForce 7800 GTX 256MB.
 
How large a ramdrive would you need?

There's a free one from microsoft that is limited to making a 33MB drive.
http://download.microsoft.com/download/win2000ddk/sample01/1/NT5/EN-US/Ramdisk.exe

There's a better one from CENATEK that has a fully functional evaluation mode that lasts up to 30 uses. You can create a ramdrive up to 4 GB in size.
http://www.cenatek.com/product_ramdisk.cfm

Found the info in this article:http://www.cenatek.com/product_ramdisk.cfm

The second one is probably the nicer solution. Unrestricted for 30 uses so it should be sufficient for testing purposes.
 
Yeah I'd need probably 512 MB to do this. I'll see what I can dig up, I'll check out those links too.
 
Back
Top