feipoa wrote on 2020-05-04, 00:51:

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.

Thanks. Unfortunately no change in access time for 1K/2K blocks 🙁
I don't know what else to do...

How do you detect the L1 cache on DLC chips? They also are 1K and are a lot more common.

Can whatever method CPU-Z is using to detect the L1 cache be adopted for HWiNFO?

Maybe it is not necessary to read the speed difference across read/write file sizes. Once you've determined the type of chip that is in a given system, you can take the information directly from the CPU registers for L1: CR0 to determine that L1 is enabled, CCR0 to determine the cache type set (2-way vs. direct-mapped), etc?

Detection of L1 on DLC is same as for DRx2 - measuring access latency for various blocks. The same method is used on all older CPUs which don't provide cache details via CPUID instruction.
L2 is detected the same way, only on later mainboards with PCI chipsets it's read from configuration registers.
So I'm really puzzled why it doesn't work here.. CCR0 won't tell the size of cache, just the type.
I'm not sure if CPU-Z is detecting the cache size, or it has it hard-coded per CPU type, will check with Franck.
I always prefer to detect things which are present in system rather than assuming something.

Do you need to know the size of the cache by measuring latency? Once you know what the CPU is, DLC or DRx2 (e.g. by using the method provided by TI in their DLC databook on page A.9), you can via table-lookup know that this CPU came in only a 1KB L1 size. I suppose if you can't figure out the latency issue, you could hard-code this particular CPU flavour for the time being.

Yup, it's always possible to hard-code things, I just wanted to do this as the last resort...

I have this 'TUBE2000' Set top box/thin client with a Cyrix MediaGX CPU/Motherboard. Neat little thing, even with an ISA or PCI slot 😀

- 166MHz Cyrix CPU, but speed is incorrectly reported as 175,4MHz (I think it also showed 175,7MHz another time, not sure though).
- One of those DiskOnChip flash drives. Is detected as a SCSI device.
- Realtek RTL8139B LAN (Onboard), Detected properly

I should have photos of the rebuild somewhere, but I can't find them yet. Thing is a little bit hard to disassemble, but if you need more info I will get you some pics.

The Cyrix parts seem to have drinks-related codenames: Kahlua, Grappa, etc 😀

Log attached.

Cyrix200+ wrote on 2020-05-04, 20:13:

I have this 'TUBE2000' Set top box/thin client with a Cyrix MediaGX CPU/Motherboard. Neat little thing, even with an ISA or PCI […]
Show full quote

I have this 'TUBE2000' Set top box/thin client with a Cyrix MediaGX CPU/Motherboard. Neat little thing, even with an ISA or PCI slot 😀

- 166MHz Cyrix CPU, but speed is incorrectly reported as 175,4MHz (I think it also showed 175,7MHz another time, not sure though).
- One of those DiskOnChip flash drives. Is detected as a SCSI device.
- Realtek RTL8139B LAN (Onboard), Detected properly

I should have photos of the rebuild somewhere, but I can't find them yet. Thing is a little bit hard to disassemble, but if you need more info I will get you some pics.

The Cyrix parts seem to have drinks-related codenames: Kahlua, Grappa, etc 😀

Log attached.

Thanks, this is an interesting system 😀
Could you please run the build posted here with the -d switch: Re: HWiNFO support of vintage hardware
That will produce a HWINFO.DBG file too, which should tell me more why the CPU clock isn't reported properly.

Mumak wrote on 2020-05-04, 20:40:

Thanks, this is an interesting system :) Could you please run the build posted here with the -d switch: Re: HWiNFO support of vi […]
Show full quote

Cyrix200+ wrote on 2020-05-04, 20:13:

I have this 'TUBE2000' Set top box/thin client with a Cyrix MediaGX CPU/Motherboard. Neat little thing, even with an ISA or PCI […]
Show full quote

I have this 'TUBE2000' Set top box/thin client with a Cyrix MediaGX CPU/Motherboard. Neat little thing, even with an ISA or PCI slot 😀

- 166MHz Cyrix CPU, but speed is incorrectly reported as 175,4MHz (I think it also showed 175,7MHz another time, not sure though).
- One of those DiskOnChip flash drives. Is detected as a SCSI device.
- Realtek RTL8139B LAN (Onboard), Detected properly

I should have photos of the rebuild somewhere, but I can't find them yet. Thing is a little bit hard to disassemble, but if you need more info I will get you some pics.

The Cyrix parts seem to have drinks-related codenames: Kahlua, Grappa, etc 😀

Log attached.

Thanks, this is an interesting system 😀
Could you please run the build posted here with the -d switch: Re: HWiNFO support of vintage hardware
That will produce a HWINFO.DBG file too, which should tell me more why the CPU clock isn't reported properly.

Here you go 😀

edit: I should attach both files duh

Cyrix200+ wrote on 2020-05-04, 20:49:

Mumak wrote on 2020-05-04, 20:40:

Thanks, this is an interesting system :) Could you please run the build posted here with the -d switch: Re: HWiNFO support of vi […]
Show full quote

Cyrix200+ wrote on 2020-05-04, 20:13:

I have this 'TUBE2000' Set top box/thin client with a Cyrix MediaGX CPU/Motherboard. Neat little thing, even with an ISA or PCI […]
Show full quote

I have this 'TUBE2000' Set top box/thin client with a Cyrix MediaGX CPU/Motherboard. Neat little thing, even with an ISA or PCI slot 😀

- 166MHz Cyrix CPU, but speed is incorrectly reported as 175,4MHz (I think it also showed 175,7MHz another time, not sure though).
- One of those DiskOnChip flash drives. Is detected as a SCSI device.
- Realtek RTL8139B LAN (Onboard), Detected properly

I should have photos of the rebuild somewhere, but I can't find them yet. Thing is a little bit hard to disassemble, but if you need more info I will get you some pics.

The Cyrix parts seem to have drinks-related codenames: Kahlua, Grappa, etc 😀

Log attached.

Thanks, this is an interesting system 😀
Could you please run the build posted here with the -d switch: Re: HWiNFO support of vintage hardware
That will produce a HWINFO.DBG file too, which should tell me more why the CPU clock isn't reported properly.

Here you go 😀

Thanks, but could you please attach the HWINFO.DBG file here too?

Mumak wrote on 2020-05-04, 20:58:

Cyrix200+ wrote on 2020-05-04, 20:49:

Mumak wrote on 2020-05-04, 20:40:

Thanks, this is an interesting system 😀
Could you please run the build posted here with the -d switch: Re: HWiNFO support of vintage hardware
That will produce a HWINFO.DBG file too, which should tell me more why the CPU clock isn't reported properly.

Here you go 😀

Thanks, but could you please attach the HWINFO.DBG file here too?

Done. It would not allow the DBG file extension so I missed it the first time.

Cyrix200+ wrote on 2020-05-04, 20:59:

Mumak wrote on 2020-05-04, 20:58:

Cyrix200+ wrote on 2020-05-04, 20:49:

Here you go 😀

Thanks, but could you please attach the HWINFO.DBG file here too?

Done. It would not allow the DBG file extension so I missed it the first time.

Thanks!
I checked the data and it all seems correct. HWINFO measures the Time-Stamp Counter clock, which should be running at the original clock and that really gives 175 MHz. Moreover, the CPU indicates the actual clock ratio is x7, which results in bus clock of 25 MHz. Are you sure the CPU is running at 166 MHz? That would also give a weird bus clock...

Mumak wrote on 2020-05-04, 21:12:

Cyrix200+ wrote on 2020-05-04, 20:59:

Mumak wrote on 2020-05-04, 20:58:

Thanks, but could you please attach the HWINFO.DBG file here too?

Done. It would not allow the DBG file extension so I missed it the first time.

Thanks!
I checked the data and it all seems correct. HWINFO measures the Time-Stamp Counter clock, which should be running at the original clock and that really gives 175 MHz. Moreover, the CPU indicates the actual clock ratio is x7, which results in bus clock of 25 MHz. Are you sure the CPU is running at 166 MHz? That would also give a weird bus clock...

There is no way to configure the speed, no jumpers and no BIOS options. There is a 50MHz crystal on the board, so the 25MHz bus speed seems more likely by the minute.

CACHECHK en CHKCPU also report 175MHz.

Seems like the darn BIOS is lying to us! 😀

Cyrix200+ wrote on 2020-05-04, 21:19:

Seems like the darn BIOS is lying to us! 😀

Yup! Never trust the BIOS 😁

Could not contain my curiosity, took it apart again a bit. The actual CPU is a 233MHz part, and there are some DIP switches on the board. Could not find any documentation in one quick Google search.

Cyrix200+ wrote on 2020-05-04, 21:47:

Could not contain my curiosity, took it apart again a bit. The actual CPU is a 233MHz part, and there are some DIP switches on the board. Could not find any documentation in one quick Google search.

So you should somehow switch the bus clock from 25 to 33 MHz.

Attached is the debug and report file for the DRx2-66. Shows 1 KB of L1 cache. I was wondering if you had any interest in reporting the caching scheme, e.g. 2-way set associative vs. direct-mapped, etc?

Does the Windows 9x version of HWiNFO also need to be run on the DRx2-66 again?

feipoa wrote on 2020-05-05, 10:50:

Attached is the debug and report file for the DRx2-66. Shows 1 KB of L1 cache. I was wondering if you had any interest in reporting the caching scheme, e.g. 2-way set associative vs. direct-mapped, etc?

Does the Windows 9x version of HWiNFO also need to be run on the DRx2-66 again?

Thanks. Unfortunately neither the last update did change cache latency times, so HWiNFO defaulted the size to a hard-coded number.
I'm totally puzzled.. this issue and the other one with DX-100 low clock can't get thru my brain...
No need to run the Windows test again.

Cyrix200+ wrote on 2020-05-04, 21:47:

The actual CPU is a 233MHz part, and there are some DIP switches on the board.

Nice board. They limited freq to fit in low TDP constrain, with better cooling you could push it higher, I guess the jumpers will do the job, there could be a PLL not directly clocked by xtal but who knows. Board design looks modern as there is SDRAM slot instead of SIMM....

Mumak wrote on 2020-04-29, 14:47:

That looks more like the KBC check is the problem here. But also wondering why the FDC37C663 wasn't detected, in the build above I added detection per its datasheet.

Hi Martin,
I am sorry for late answer, 30.4. you sent me PM about better detection LAN adapter and SIO in my HP Vectra. I can't sent you PM, because i am newly registered user here. If it is still actual, yes you can send me new version of HWINFO with improvements in detection and i test it for you.

hasat wrote on 2020-05-06, 09:38:

Mumak wrote on 2020-04-29, 14:47:

That looks more like the KBC check is the problem here. But also wondering why the FDC37C663 wasn't detected, in the build above I added detection per its datasheet.

Hi Martin,
I am sorry for late answer, 30.4. you sent me PM about better detection LAN adapter and SIO in my HP Vectra. I can't sent you PM, because i am newly registered user here. If it is still actual, yes you can send me new version of HWINFO with improvements in detection and i test it for you.

Yes, I'm still interested. Please use the version attached here: Re: HWiNFO support of vintage hardware