Cavemanthe0ne Posted September 28, 2015 Posted September 28, 2015 I am very interested to find out how far back in versions of Windows benchmarks and CPUz will work on. I am doing a test of some ancient hardware come this spring, and it would be nice to be able to know if CPUz (and anything else, like Hwbot Prime, or GPUPI, etc...) would even work on some of the hardware. And when I say ancient, I mean systems from 1990 and up... So, from running Windows 3.x, to Win 95, and onwards. If you were wondering, I am going to be testing if Moore's Law has also had an effect an CPU performance over the years, and not just transistor count. Thanks! Quote
GENiEBEN Posted September 28, 2015 Posted September 28, 2015 Anything older than W2K is going to be a problem. Pifast and Wprime should run. HWBOT Prime might, since there are java builds for 98/ME. GPUPI is a no go. SPI works, but not the version we're using. Quote
Cavemanthe0ne Posted September 28, 2015 Author Posted September 28, 2015 Ok, that is helpful. Most of the hardware I'll be testing *could* run W2k, even if slowly. Thanks for the info! Now to see where the legacy builds are for Java. YAY for testing old hardware, the stuff will probably take months to run Pifast and/or Wprime lol . In the name of research! Other thing- Will I be able to validate the hardware with CPUz or will I have to run an outdated version to get it to work? I see some people are having issues with old hardware (like socket A rigs) ... Quote
GENiEBEN Posted September 28, 2015 Posted September 28, 2015 (edited) Ok, that is helpful. Most of the hardware I'll be testing *could* run W2k, even if slowly. Thanks for the info! Now to see where the legacy builds are for Java. YAY for testing old hardware, the stuff will probably take months to run Pifast and/or Wprime lol . In the name of research! Other thing- Will I be able to validate the hardware with CPUz or will I have to run an outdated version to get it to work? I see some people are having issues with old hardware (like socket A rigs) ... Java builds: WIN95 www.oldapps.com/java.php?old_java=7 WIN98/ME www.oldapps.com/java.php?old_java=52 WIN2K http://www.oldapps.com/java.php?old_java=6281 Java.com also has a downloads archive but requires an account (free?), but it's bloody well hidden. Depends on the hardware tested, but yes, running Spi32/W1024 on Pentium 1 will set you for 30 hours+. Do remember Wprime has an issue when a run takes over 24hrs, make sure you revert the OS time back every 23ish hours. Yes, you can't validate frequencies with CPU-Z on stuff older than XP, but for screenshots only you could use any version. Socket A was particularly buggy with CPU-Z on *some* boards, feel free to browse this forum for solutions, I believe there's even special builds of cpu-z for some older sockets. Edited September 28, 2015 by GENiEBEN Quote
lanbonden Posted September 28, 2015 Posted September 28, 2015 Anything older than W2K is going to be a problem. Pifast and Wprime should run. HWBOT Prime might, since there are java builds for 98/ME. GPUPI is a no go. SPI works, but not the version we're using. About GPUPi, it should be possible to run on W2K with the 12.1 Catalyst driver for cpu suport. Quote
GENiEBEN Posted September 28, 2015 Posted September 28, 2015 About GPUPi, it should be possible to run on W2K with the 12.1 Catalyst driver for cpu suport. The problem is the Visual Studio Redis, the legacy version I think uses VS2012, which does not work on W2K (or XP pre-SP2) afaik. You could try and patch it, or find one patched. Quote
lanbonden Posted September 28, 2015 Posted September 28, 2015 The problem is the Visual Studio Redis, the legacy version I think uses VS2012, which does not work on W2K (or XP pre-SP2) afaik. You could try and patch it, or find one patched. Good to know now that Im working on figuring out what tests can be run on what (old school) hardware. Quote
Cavemanthe0ne Posted September 28, 2015 Author Posted September 28, 2015 Thanks for the links! Certainly helps. And I'll be running a couple of 386 and 486 systems (Maybe more than a couple, time will tell on how many I can find), so it might take a while on those rigs to finish the benches. Ok, will take that into account when running benches(About the CPUz & wprime). Also, on older Windows versions could I use syscheck for validation? I browsed around a bit and some other people with really old hardware used that for submitting to Hwbot. Quote
GENiEBEN Posted September 28, 2015 Posted September 28, 2015 Yes, we can make reasonable exceptions for such old hardware, but try to provide as much information and pictures, so that no one has complaints. EDIT: http://forum.hwbot.org/showthread.php?t=108247 Quote
Cavemanthe0ne Posted September 29, 2015 Author Posted September 29, 2015 OK, thanks! I'll likely run at stock for the most part, but will include Syscheck or an appropriate CPUz tab or three when submitting. This is still a decent amount of time away, but again thanks for the info, and links to old Java versions! Heck, i can probably mention you if the research I have in mind goes well enough. (In other words, if I find enough hardware to make a good comparison & then write up an article or paper). Quote
trodas Posted September 29, 2015 Posted September 29, 2015 Windows 2000 run even the latest CPU-Z, so validation should not be a problem. For Win98se you need to dial back to CPU-Z 1.51. For WinNT 4.0 SP1 you need to dial back to CPU-Z 1.30. Also you can include programs like TSCWin and TestCPU, like I did: http://hwbot.org/submission/2978646_ Pifast and Wprime should run PiFast does run on WinNT 4.0 SP1 w/o a glitch, but wPrime does not. It require the msvbvm60.dll library and that library, even in the oldest version, require the WinNT 4.0 SP3. Then it should work, but I did not yet get around to test that, but it is likely. W/O ActiveX (on for example highly optimized WinXP) you cannot install the CPU-Z libraries, but that it is. However, for a i386 / i486 systems I'm affraid I have bad news for wPrime. Current version 1.55 have a nasty bug, that prevent longer that 24h calculations to take place. It check for time and when the time "move back", then it write "Cheat detected" and stop the calculation. There is no known way for me to get around it. I just set the time to 0:01 and run the calculation. Unless it exceeds 24h (1024M on AMD K5 PR75 @ 90 is getting close with 17h 48min: http://hwbot.org/submission/2986787_ ), then all is fine. But it ignore the day, when it check if the time is not dialed back during the run, so unless your tests with wPrime is under 24h - OR - there is a fix for that, you have to strike wPrime from the test list, because I'm affraid that on i386 even 32M run could be over 24h... So, this is just something to be carefull off. Quote
Cavemanthe0ne Posted March 13, 2016 Author Posted March 13, 2016 Well, now I have a Wyse 386/25 PC that is running Dos-Under-Unix. (I think? It uses an obscure VGA connector, one with 9 pins, not 15, so I don't really know... but that is what a lot of the other Wyse multi-user PC's ran on so...) Would it be that hard to run benchmarks on a 25 Mhz CPU? I'm not even sure if the super-stripped-down version of Win XP works on the old thing . Quote
Mr.Scott Posted March 13, 2016 Posted March 13, 2016 Well, now I have a Wyse 386/25 PC that is running Dos-Under-Unix. (I think? It uses an obscure VGA connector, one with 9 pins, not 15, so I don't really know... but that is what a lot of the other Wyse multi-user PC's ran on so...) Would it be that hard to run benchmarks on a 25 Mhz CPU? I'm not even sure if the super-stripped-down version of Win XP works on the old thing . WinNT. Quote
Cavemanthe0ne Posted March 13, 2016 Author Posted March 13, 2016 WinNT. Cool, thanks. Now to venture through the internets to find the 9-pin VGA to 15-pin VGA adapter lol- Then I might actually be able to use the thing... Quote
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.