LinkExchange Member
  OS/2 Warp
  Benchmark Centre

Frequently Asked Questions

  Why do I get an error logged in sb_error.log about fonts?

With fixpack 12 for Warp 4 or with Warp Server for e-Business IBM helpfully replaced the sysmono.fon file in \os2\dll that contains the System Monospaced fonts. They neglected to ship the only one that looks any good (10x8). When run at 1600x1200 or higher or on DBCS versions of Warp, Sysbench looks for this font specifically and tries to use it. If it cannot find it then it will not run. I've reported this to IBM but I'm not holding my breath for a solution. In the meantime I recommend that you save a copy of \os2\dll\sysmono.fon before FP12+ and put it back afterwards!

What does all this mean?

Basically, it is a chance for OS/2 users around the world to compare their hardware with other users, based soley on the output of the Sysbench benchmark program, maintained by Trevor Hemsley. Now you can brag, in public, how well OS/2 v3.0 runs on your old 386-16. :) If you have suggestions for inclusion in the benchmark, let me know. You can mail me if you decode trevor-hemsley at dsl dot pipex dot com to a mail address!

OK, what do I need to participate?

A machine running OS/2 Warp 3.0 or higher, and the latest Sysbench program. There are links to get the program from the main page of this site here. After that, run the program, save the results, and use the Submissions page to send both the TXT and the HTML results files to me for inclusion on the site.

Note: There is also a replacement FLOPS test for Sysbench 0.9.2 and 0.9.3 ONLY, not 0.93a or later. This corrects a bug that I introduced in 0.9.2 and fixed in 0.9.3a. Grab that here as well.

Wow, I thought my machine was better than that!

Don't give it to your little brother or sister just yet! While it's true that the Sysbench results can be less than flattering sometimes, you have to keep them in perspective. All they really mean is that Machine X is faster at running Sysbench than Machine Y is. Nothing more, nothing less.

From the README.TXT ...
This is a freeware benchmark for OS/2. Like all such benchmark programs, its results can be read in many ways. Beware of making too many decisions based on its output - by their very nature, benchmark programs are *not* real world applications. By tuning your system with the aid of benchmarks, you may end up with a system that runs benchmark programs wonderfully well but runs applications badly! Just so you're aware....

So, don't rush out to buy the new fangled 3000MHz BrandX CPU, or the new BrandY 40GB SCSI-9 UUUWWW hardrive based on these scores alone! :)

OK, then why have a "King Of The Heap"?

Basically, so that you can see the top scoring hardware on one page, not spread out over all of them. It is only an indication of the highest scoring hardware, using the Sysbench results as the only criteria.... well, almost the only criteria. You won't find me listing any hardware here that is made by Diamond due to their lack of OS/2 support. Otherwise, price etc. is not taken into account. Should you base your next purchase on this week's King Of The Heap? Probably not, but looking at the scores, you would probably have a killer machine nonetheless!

I've found a bug in ...

Problems/bugs/etc with Sysbench should be sent to trevor-hemsley at dsl dot pipex dot com.. Problems/bugs/etc with these pages are sent to the same address as above. If you have any other questions, concerns, or suggestions, send them to me as well.

Why doesn't my CD drive show up in the list of drives to be tested?

Prior to version 0.9.4 there was a bug in my code that meant that CD drives without a disk inserted at the time that you started Sysbench were not listed. The solution is to make sure that there is a CD in the drive or upgrade to 0.9.4 when it's available.

Why does the Cache/Bus disk test score less than the Average xfer test?

Bottom line, I don't know. I don't really trust the Cache/Bus transfer test since it appears to return some very odd results. I've rewritten much of the code for the disk I/O tests in 0.9.4 and improved the results from this test but I still don't place much reliance on its results. My guess as to why many people see this anomaly is that some drives, usually the high performance ones, are optimised to discard the most recently read track from their internal cache RAM in order to make room for the next record that it thinks is most likely to be read. For the Cache/Bus test this has the effect of forcing a reread of the data from the disk surface and thus slowing everything down. On the other hand there might be a bug in my code that's survived the rewrite.

I just tried to run Sysbench on OS/2 2.1x and....

Yes, you're right, it doesn't work on OS/2 2.1x any more. Sysbench requires that DIVE.DLL be accessible and loadable and it doesn't exist for OS/2 2.1x. I've tried copying the DLL into \OS2\DLL but it then complains about the level of PMWIN.DLL and this was totally rewritten by IBM for Warp 3. Short of creating a separate version of Sysbench with no DIVE related code in it specifically for OS/2 2.1x I don't know how to bypass this problem. Since OS/2 2.1x is now drawing its old age pension and sitting by the fire drinking cocoa, I'm inclined not to do anything more about this. Sorry.

I tried to run Sysbench and get an error message in sb_error.log that says "Can't get the right font"

Sysbench requires that you have either the System Proportional or System Monospaced fonts installed. On non-DBCS versions of Warp the Proportional font is used unless your display is set at 1600x1200 or higher. On DBCS versions of Warp where the System Proportional font doesn't exist then the System Monospaced font is used instead. If you have a DBCS version of Warp and still see this message then you may not have the latest version of the code since I added support in about 0.9.3 for this.

All the CPU usage figures for the disk and CD tests show 100%....

Either you don't have busmastering drivers and/or hardware for those drives or you have another program running that is sucking up CPU time. Many CPU monitors work by starting a low priority thread in the background that sits there counting like mad. This thread knows how many times it can count in a certain time period when it's the only thing running and can thus work out how much CPU time is being used by everything else on your machine. Sysbench uses this method of calculating CPU time on machines that are running versions of Warp prior to V3.0 plus fixpack 29 when a new API was added that allows a program to query OS/2 for the amount of CPU time used. Whichever method of calculation Sysbench uses will be inaccurate if you have another load meter running that is gobbling CPU time.

I get a Sys3175 when running the File I/O tests....

If you are running HPFS386 this is a known problem in my code and I fixed it. Upgrade to the latest version and try again.

I have a hard disk listed that doesn't exist...

This is probably caused by having the EXT2-FS Linux File System IFS installed. It makes native Linux partitions appear as separate physical disks by some slight of hand. This is normal and nothing to worry about.

Running Sysbench causes my machine to hang...

It's most likely that you have a hardware problem. I've had several people report this and most were fixed by changing something in the hardware. One person changed a 486DX/2 66 for a DX4/100 and the hang went away. I've seen it on my own machine when one of the fans had fallen off my second Pentium Pro. It would freeze solid in the middle of the CPU tests.

Back to The OS/2 Benchmark Centre