Yes, the L1 cache is definately enabled. In DOOM, I see 12.2 fps with L1 enabled and 6.9 fps with L1 disabled.

CHKCPU also shows L1 as enabled.

Weird.. actually the test HWiNFO does to detect cache size is both read/write. Would be interesting to see more detailed cache latency results on that system, there are tools for that.

How about move? I'd have to play with various cache testing software to see if anything can pickup the L1.

feipoa wrote on 2020-04-30, 18:53:

How about move? I'd have to play with various cache testing software to see if anything can pickup the L1.

No problem, go ahead 😀

I wonder, do you thing that is possible on modern NVIDIA/ATI/AMD cards in detect somehow actual GPU and Memory clocks? Because it would help my dos benchmarking project a lot - Old+Modern videocards pure DOS benchmarking- which one is fastest?, need your numbers + analysis, 320x200 to 1600x1200!

Because these cards are have some basic/boot clocks until drivers are loaded, which are used for Dos too, so DOS performance is often related to it, or at least that is my guess why performance is far from their raw horsepower in 2D.

ruthan wrote on 2020-04-30, 19:49:

I wonder, do you thing that is possible on modern NVIDIA/ATI/AMD cards in detect somehow actual GPU and Memory clocks? Because it would help my dos benchmarking project a lot - Old+Modern videocards pure DOS benchmarking- which one is fastest?, need your numbers + analysis, 320x200 to 1600x1200!

Because these cards are have some basic/boot clocks until drivers are loaded, which are used for Dos too, so DOS performance is often related to it, or at least that is my guess why performance is far from their raw horsepower in 2D.

In DOS? For later NVIDIA GPUs this won't be possible as the register-level documentation is impossible to get; in Windows all tools rely on NVAPI.
For AMD it might be possible, but quite difficult to implement. Access to some resources would require access to memory mapped registers, so probably protected mode.

Yes in Dos.

I ported the NVclock utility to DOS years ago (available for download on my website) that can read and set timings of some older nvidia cards (pre-G80) but no way for modern cards as there's no documentation (there are thousands of memory mapped registers...).

Mumak wrote on 2020-04-30, 19:09:

feipoa wrote on 2020-04-30, 18:53:

How about move? I'd have to play with various cache testing software to see if anything can pickup the L1.

No problem, go ahead :)

Attached are two programs which can measure the speed difference between L1 and L2 cache when using a Cyrix DRx2. Not much change for write speed, but read and move show about twice the speed for L1 vs. L2 cache. Were you able to update HWiNFO to detect 1K of L1 cache on this chip?

feipoa wrote on 2020-05-01, 10:46:

Mumak wrote on 2020-04-30, 19:09:

feipoa wrote on 2020-04-30, 18:53:

How about move? I'd have to play with various cache testing software to see if anything can pickup the L1.

No problem, go ahead 😀

Attached are two programs which can measure the speed difference between L1 and L2 cache when using a Cyrix DRx2. Not much change for write speed, but read and move show about twice the speed for L1 vs. L2 cache. Were you able to update HWiNFO to detect 1K of L1 cache on this chip?

Not yet 🙁 I'm still looking into that but can't figure out why the heck it's giving me numbers as if there would be no performance improvement for 1K..

feipoa wrote on 2020-05-01, 10:46:

Mumak wrote on 2020-04-30, 19:09:

feipoa wrote on 2020-04-30, 18:53:

How about move? I'd have to play with various cache testing software to see if anything can pickup the L1.

No problem, go ahead 😀

Attached are two programs which can measure the speed difference between L1 and L2 cache when using a Cyrix DRx2. Not much change for write speed, but read and move show about twice the speed for L1 vs. L2 cache. Were you able to update HWiNFO to detect 1K of L1 cache on this chip?

Please try this build whether it gives different numbers for cache access:
[build pulled, newer available]

RayeR wrote on 2020-04-30, 20:57:

I ported the NVclock utility to DOS years ago (available for download on my website) that can read and set timings of some older nvidia cards (pre-G80) but no way for modern cards as there's no documentation (there are thousands of memory mapped registers...).

Fine, G80 = Geforce 8800, that is old. BTW which is first Nvidia card which are using 2 or more clocks, for boot / 2D / 3D?

Uh, I don't remmeber. There also was a tool Nibitor (nvidia BIOS editor) that could set some clock rates but also for old cards...

Hiya,

just got around to trying the latest version on the Compaq Presario 2232 again, all seems to be well now, no lock up on soundcard detection and CPU clocks seem to be stable too. Great stuff! Only anomaly I can see is clock speed of the FPU, but hey I'm not nitpicking!

BSA Starfire wrote on 2020-05-02, 14:43:

Hiya, […]
Show full quote

Hiya,

just got around to trying the latest version on the Compaq Presario 2232 again, all seems to be well now, no lock up on soundcard detection and CPU clocks seem to be stable too. Great stuff! Only anomaly I can see is clock speed of the FPU, but hey I'm not nitpicking!
DSC_0221.JPG
DSC_0224.JPG
DSC_0222.JPG
DSC_0223.JPG

Thanks for the feedback! Could you please create a full report file and attach?

DRx2-66 running on the PANDA board.

I finally did Navratils system info vs HWinfo comparison
https://docs.google.com/spreadsheets/d/1SSGKk … dit?usp=sharing // Comment are enabled.

I was right in the past, Navratil has much more options.. of course that we can discuss if such options make sense.. but i didnt find any obliviously useless ones.
Also i mapping can be wrong, because i simply could miss some HWinfo info option.

For me personally Navratil has also better user experience - easier navigation, works in all memory modes, its more PCem, virtuals friendly.

Its not some competition in normal sense, so i will gladly use both. I did this comparison to help, show where is Navratil additional value so far, not to complain too much.. but you know me, im goal oriented:)

ruthan wrote on 2020-05-03, 13:06:

I finally did Navratils system info vs HWinfo comparison https://docs.google.com/spreadsheets/d/1SSGKk … dit?usp=sharing // Co […]
Show full quote

I finally did Navratils system info vs HWinfo comparison
https://docs.google.com/spreadsheets/d/1SSGKk … dit?usp=sharing // Comment are enabled.

I was right in the past, Navratil has much more options.. of course that we can discuss if such options make sense.. but i didnt find any obliviously useless ones.
Also i mapping can be wrong, because i simply could miss some HWinfo info option.

For me personally Navratil has also better user experience - easier navigation, works in all memory modes, its more PCem, virtuals friendly.

Its not some competition in normal sense, so i will gladly use both. I did this comparison to help, show where is Navratil additional value so far, not to complain too much.. but you know me, im goal oriented:)

Thanks for the comparison, I will check it in more detail. For now just a question - which sound card was in that machine where HWiNFO reported PS1 only?

Latest debug and log file from HWiNFO DOS on DRx2-66 w/PANDA motherboard.

HWiNFO is reporting only 64K of L2 cache. Should be 256K. L1 still absent.

Mumak wrote on 2020-05-03, 13:34:

Which sound card was in that machine where HWiNFO reported PS1 only?

Emulated SB16, i started with testing on real HW, but i quickly realized that all menus are the same.. so if some detection is not working, there are is at least item with unknown value etc.. i retested few options on real HW, but it was same (but more time consuming) - from list of parameters. Almost all my Navratil tests are from PCem, Vmware and Dosbox, there it just works, i did later same for HWinfo.
If sure that for some items are is more informations for real HW, i can retest it.. but only difference which i saw so really emulated SB16 and drivers detection - which simply can work in DosBox, PCem should be on 99% same.