VOGONS


Reply 40 of 58, by darry

User metadata
Rank l33t++
Rank
l33t++
digicube wrote on 2021-01-15, 04:13:
I get this result between WINNT4 and WIN7 using this instruction https://www.dell.com/support/kbdoc/en-us/0001 … dth-using-iperf […]
Show full quote

I get this result between WINNT4 and WIN7 using this instruction https://www.dell.com/support/kbdoc/en-us/0001 … dth-using-iperf
Looks like IT IS the CF card/HDD bottleneck.
[ ID] Interval Transfer Bandwidth
[164] 0.0- 1.0 sec 13.0 MBytes 109 Mbits/sec
[164] 1.0- 2.0 sec 8.13 MBytes 68.2 Mbits/sec
[164] 2.0- 3.0 sec 10.3 MBytes 86.6 Mbits/sec
[164] 3.0- 4.0 sec 10.9 MBytes 91.1 Mbits/sec
[164] 4.0- 5.0 sec 10.9 MBytes 91.4 Mbits/sec
[164] 5.0- 6.0 sec 6.77 MBytes 56.8 Mbits/sec
[164] 6.0- 7.0 sec 8.33 MBytes 69.9 Mbits/sec
[164] 7.0- 8.0 sec 11.0 MBytes 91.9 Mbits/sec
[164] 8.0- 9.0 sec 10.9 MBytes 91.8 Mbits/sec
[164] 9.0-10.0 sec 11.0 MBytes 92.1 Mbits/sec
[164] 10.0-11.0 sec 10.9 MBytes 91.3 Mbits/sec
[164] 11.0-12.0 sec 8.66 MBytes 72.7 Mbits/sec
[164] 12.0-13.0 sec 10.8 MBytes 90.8 Mbits/sec
[164] 13.0-14.0 sec 10.1 MBytes 84.7 Mbits/sec
[164] 14.0-15.0 sec 9.52 MBytes 79.8 Mbits/sec
[164] 15.0-16.0 sec 10.6 MBytes 88.6 Mbits/sec
[164] 16.0-17.0 sec 8.96 MBytes 75.2 Mbits/sec
[164] 17.0-18.0 sec 10.8 MBytes 91.0 Mbits/sec
[164] 18.0-19.0 sec 10.8 MBytes 91.0 Mbits/sec
[164] 19.0-20.0 sec 8.70 MBytes 72.9 Mbits/sec
[ ID] Interval Transfer Bandwidth
[164] 20.0-21.0 sec 4.10 MBytes 34.4 Mbits/sec
[164] 21.0-22.0 sec 10.1 MBytes 85.1 Mbits/sec
[164] 22.0-23.0 sec 10.9 MBytes 91.4 Mbits/sec
[164] 23.0-24.0 sec 10.9 MBytes 91.7 Mbits/sec
[164] 24.0-25.0 sec 11.0 MBytes 92.1 Mbits/sec
[164] 25.0-26.0 sec 10.9 MBytes 91.6 Mbits/sec
[164] 26.0-27.0 sec 10.8 MBytes 90.4 Mbits/sec
[164] 27.0-28.0 sec 10.9 MBytes 91.3 Mbits/sec
[164] 28.0-29.0 sec 9.51 MBytes 79.8 Mbits/sec
[164] 29.0-30.0 sec 10.1 MBytes 84.4 Mbits/sec
[164] 0.0-30.2 sec 300 MBytes 83.4 Mbits/sec

That is good news . The next step is to benchmark your hard disk or CF card . Remember that IDE DMA is not enabled by default with OS default drivers on Windows 9x . On NT4, IDE DMA probably requires a service pack to be optionally enabled and may be enabled by default with service pack 4 or higher, AFAICR .

Some CF to IDE adapters do not support DMA mode and not all CF cards do either .

Reply 42 of 58, by digicube

User metadata
Rank Newbie
Rank
Newbie

What's a good NT4 software to benchmark my HDD and CF card? I have NT4 SP6 installed.

Edit: I used DMACHECK to enable DMA in WINNT4 but it says DMA not in use in Channel 0, DMA in use in Channel 1. I'm guessing my CF card which is master should be Channel 0 and my Zip drive is slave Channel 1 or it could be my DVD drive on secondary IDE.
HD Tach reports avg 3.2MB/s read speed at 95% CPU usage, unable to do write test because it requires a blank drive. Seems like CPU and CF card bottleneck. Should I install Universal ATA driver? http://alter.org.ua/soft/win/uni_ata/

Reply 43 of 58, by dionb

User metadata
Rank l33t++
Rank
l33t++
digicube wrote on 2021-01-15, 04:13:
I get this result between WINNT4 and WIN7 using this instruction https://www.dell.com/support/kbdoc/en-us/0001 … dth-using-iperf […]
Show full quote

I get this result between WINNT4 and WIN7 using this instruction https://www.dell.com/support/kbdoc/en-us/0001 … dth-using-iperf
Looks like IT IS the CF card/HDD bottleneck.
[ ID] Interval Transfer Bandwidth
[164] 0.0- 1.0 sec 13.0 MBytes 109 Mbits/sec
[164] 1.0- 2.0 sec 8.13 MBytes 68.2 Mbits/sec
[164] 2.0- 3.0 sec 10.3 MBytes 86.6 Mbits/sec
[164] 3.0- 4.0 sec 10.9 MBytes 91.1 Mbits/sec
[164] 4.0- 5.0 sec 10.9 MBytes 91.4 Mbits/sec
[164] 5.0- 6.0 sec 6.77 MBytes 56.8 Mbits/sec
[164] 6.0- 7.0 sec 8.33 MBytes 69.9 Mbits/sec
[164] 7.0- 8.0 sec 11.0 MBytes 91.9 Mbits/sec
[164] 8.0- 9.0 sec 10.9 MBytes 91.8 Mbits/sec
[164] 9.0-10.0 sec 11.0 MBytes 92.1 Mbits/sec
[164] 10.0-11.0 sec 10.9 MBytes 91.3 Mbits/sec
[164] 11.0-12.0 sec 8.66 MBytes 72.7 Mbits/sec
[164] 12.0-13.0 sec 10.8 MBytes 90.8 Mbits/sec
[164] 13.0-14.0 sec 10.1 MBytes 84.7 Mbits/sec
[164] 14.0-15.0 sec 9.52 MBytes 79.8 Mbits/sec
[164] 15.0-16.0 sec 10.6 MBytes 88.6 Mbits/sec
[164] 16.0-17.0 sec 8.96 MBytes 75.2 Mbits/sec
[164] 17.0-18.0 sec 10.8 MBytes 91.0 Mbits/sec
[164] 18.0-19.0 sec 10.8 MBytes 91.0 Mbits/sec
[164] 19.0-20.0 sec 8.70 MBytes 72.9 Mbits/sec
[ ID] Interval Transfer Bandwidth
[164] 20.0-21.0 sec 4.10 MBytes 34.4 Mbits/sec
[164] 21.0-22.0 sec 10.1 MBytes 85.1 Mbits/sec
[164] 22.0-23.0 sec 10.9 MBytes 91.4 Mbits/sec
[164] 23.0-24.0 sec 10.9 MBytes 91.7 Mbits/sec
[164] 24.0-25.0 sec 11.0 MBytes 92.1 Mbits/sec
[164] 25.0-26.0 sec 10.9 MBytes 91.6 Mbits/sec
[164] 26.0-27.0 sec 10.8 MBytes 90.4 Mbits/sec
[164] 27.0-28.0 sec 10.9 MBytes 91.3 Mbits/sec
[164] 28.0-29.0 sec 9.51 MBytes 79.8 Mbits/sec
[164] 29.0-30.0 sec 10.1 MBytes 84.4 Mbits/sec
[164] 0.0-30.2 sec 300 MBytes 83.4 Mbits/sec

That's looking a lot better. Shows that under NT4 you're approaching the limits of the NIC - if your CPU doesn't have to do anything else.

Given DMA isn't active on your CF card, it could still be a CPU bottleneck as without DMA all transfers need to go via CPU. Of course, getting DMA working would still be the best next step.

Worth taking a look at: what is CPU utilization when doing those iPerf tests? And what is it when trying file transfers?

Reply 44 of 58, by digicube

User metadata
Rank Newbie
Rank
Newbie

30-50% CPU usage during iperf test. Universal ATA gives blue screen. When I try to reload a backup SNAPSHOT image to the same CF card, it reports 3810MB instead of the expected 3847MB and failed. Most likely due to bad sectors, the CF card now is smaller. How do I get the image back to the same CF card? I tried copying the files and do fdisk /mbr but it didn't work.

Reply 45 of 58, by darry

User metadata
Rank l33t++
Rank
l33t++
digicube wrote on 2021-01-15, 20:33:

30-50% CPU usage during iperf test. Universal ATA gives blue screen. When I try to reload a backup SNAPSHOT image to the same CF card, it reports 3810MB instead of the expected 3847MB and failed. Most likely due to bad sectors, the CF card now is smaller. How do I get the image back to the same CF card? I tried copying the files and do fdisk /mbr but it didn't work.

If you have bad sectors on your CF card, I recommend that you replace it .

Reply 47 of 58, by darry

User metadata
Rank l33t++
Rank
l33t++
digicube wrote on 2021-01-15, 20:45:

It's common for old CF cards to have bad sectors, I don't want to throw away if most is still usable. I'll do a reinstall of WINNT4.

I have never had a CF card with bad sectors and I have some very old ones (8MB, 128MB, etc) and several dozens of them in total. Also, my understanding of flash based media is that if bad blocks start appearing, it means that the medium is out of spare cells to do a re-mapping which, in turn, means that the said medium has endured a lot of write cycles and is on its way out anyway .

EDIT : Corrected typos and added precisions

Reply 48 of 58, by darry

User metadata
Rank l33t++
Rank
l33t++

I was feeling bored today, so I dug up my 400MHz Pentium 2 (4x100MHz), installed it in my Asus P3B-F and ran at 266MHz (4x66MHz) to get something fairly comparable to your CPU and then ran it with a 1400MHz CPU for a best case scenario .
- The machine was running Windows 98 SE and was summarily TCP optimized using https://www.speedguide.net/files/sguide_tweak_9x.zip (TCP Optimizer set to optimum did not give optimal results) .
- CPU-Z was run after the FTP transfer had completed

First, a disk benchmark with the 266MHz CPU :

266mhz_atto.png
Filename
266mhz_atto.png
File size
30.98 KiB
Views
1112 views
File license
Public domain

This is running in DMA mode, obviously, off of a SIL3114 SATA controller

Second, the best performance I could get with a 266MHz CPU through FTP using a 3COM 3C905B (100Mb NIC):

266mhz.png
Filename
266mhz.png
File size
26.14 KiB
Views
1112 views
File license
Public domain

Third, the best performance I could get with a 266MHz CPU through FTP using an Intel PRO/1000 GT (Gigabit NIC) :

266mhz_gig.png
Filename
266mhz_gig.png
File size
24.71 KiB
Views
1112 views
File license
Public domain

Finally, the best performance I could get with a 1400MHz Tualatin (PowerLeap adapter) CPU through FTP using an Intel PRO/1000 GT (Gigabit NIC) :

1400mhz_gig.png
Filename
1400mhz_gig.png
File size
27.01 KiB
Views
1112 views
File license
Public domain

It should be obvious that a 266MHz Pentium II CPU is just not fast enough to drive a Gigabit NIC .

Conclusion :
Assuming my results are representative, running a Gigabit NIC rather than a 100Mbps one on a 266MHz Pentium II CPU is practically pointless.

EDIT : corrected a typo and an image
EDIT2 : Added some context

Reply 50 of 58, by darry

User metadata
Rank l33t++
Rank
l33t++
digicube wrote on 2021-01-17, 01:54:

Welp, that's a conclusive answer for future seekers.

The good news, in your case, is that practically any compatible CPUs that you can find for your P2B-D would likely be a worthwhile upgrade .
See https://www.asus.com/us/supportonly/P2B-D/HelpDesk_CPU/

As for storage, a (U)DMA capable CF to IDE adapter , like this https://www.startech.com/en-ca/hdd/ide2cf , with a high performance UDMA capable CF card (possibly one of the "industrial" models) would likely solve your apparent disk bottleneck issues .

Reply 52 of 58, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

I’m going to attempt to do gigabit on my socket 7 windows 98 build.

But honestly I don’t think the way I did it is going to work, I was mainly buying that hardware with newer windows versions in mind(triple operating system on this build)

But what I did is I bought the startech usb 3 to pci card then bought a gigabit usb adapter cause pci slots are at a premium.

If you follow my posts later on you can ask me if it works in 98 but I kind of expect the usb 3 card won’t work and kind of doubt the adapter will work under 98 on my usb 2 ports. So I will likely be also using a 10/100 on usb into a switch.

It’s worth noting that this dongle has only been doing about 40mb/sec on my other computers. When plugged into usb3 so not sure what’s up there.. but that is faster than 10/100

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 53 of 58, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

If you haven’t resolved the 1.6mb/sec issue yet. It sounds like you may be at 10 speeds. Check adapter settings.

Then start testing your cables. I just went through this at my place. Had to replace cables, put new ends on them, fix mis wired cables. You may think your cables are right but you don’t know. I had actually like 5 factory done cables that were wrong.

Best wiring standard is B with the blue in between the two greens. And the order matters. And so does the twisting keep em twisted as much as possible and buy the highest rated cables you can find cat 5e, 6, 6a, etc.

When I finally got all mine working at gigabit the last thing I found was one of my router ports was stuck at 100 where the other four were fine. Check all the equipment!

Also, ignore these guys who knock you for wanting the fastest card you can get. Just cause they would buy a 10/100 doesn’t mean you have to. This is your build and people should help you do what you want to do, or kindly suggest alternatives while respecting that you don’t have to do as they would do.

Last edited by Sphere478 on 2021-01-17, 09:05. Edited 2 times in total.

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 54 of 58, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

If interested in making your disk faster,
Check out my build thread “fastest socket 7 on a 430tx chipset” in it I’m going to be running SSDs on sata, and a m.2 nvme.

I think a windows 98 compatable sata card and a ssd would be a good fit for you.

Or possibly a ide to sata adapter on a ssd plugged into your motherboard ide.

Ps. As others have suggested, a cpu upgrade will help you. Ask around to see what the fastest cpu you can cram in that is. I think you may be able to get a slocket adapter and plug a tualitin adapter into that and maybe get up to a p3 1.4. But I’m not sure. P2s and p3s were never really my thing.

But if the 266 has nostalgic significance I understand 😀 maybe overclocking?

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 55 of 58, by darry

User metadata
Rank l33t++
Rank
l33t++
Sphere478 wrote on 2021-01-17, 08:47:
If you haven’t resolved the 1.6mb/sec issue yet. It sounds like you may be at 10 speeds. Check adapter settings. […]
Show full quote

If you haven’t resolved the 1.6mb/sec issue yet. It sounds like you may be at 10 speeds. Check adapter settings.

Then start testing your cables. I just went through this at my place. Had to replace cables, put new ends on them, fix mis wired cables. You may think your cables are right but you don’t know. I had actually like 5 factory done cables that were wrong.

Best wiring standard is B with the blue in between the two greens. And the order matters. And so does the twisting keep em twisted as much as possible and buy the highest rated cables you can find cat 5e, 6, 6a, etc.

When I finally got all mine working at gigabit the last thing I found was one of my router ports was stuck at 100 where the other four were fine. Check all the equipment!

Checking the rest of the equipment involved is a good general recommendation, though IMHO, in this case, the Iperf numbers the OP is getting here Re: Looking for Windows 95 PCI Gigabit 1000Mbps ethernet card. give the impression that cabling/switch/router is not the main cause the reported performance issue . Specifically, if the problem were due to a 10Mbps link speed issue, not only would the Iperf results be several times lower, but even 1.6MB per second would not even be possible . Just to be clear, my understanding is that the OP actually means 1.6MB, as in megabytes as opposed to 1.6mb, as in megabits.

Sphere478 wrote on 2021-01-17, 08:47:

Also, ignore these guys who knock you for wanting the fastest card you can get. Just cause they would buy a 10/100 doesn’t mean you have to. This is your build and people should help you do what you want to do, or kindly suggest alternatives while respecting that you don’t have to do as they would do.

Another good point, OP is free to do what he wants.

My comments' intent was to
a) provide realistic expectations as to the network throughput performance and potential performance gains (if any) based on upgrading specific hardware in the OP's build
b) make suggestions as to ways to improve network throughput performance
c) give my opinion as to whether using a gigabit NIC was worthwhile in the OP's build and back that up with empirical data .

If I seemed too forceful in the way I presented my opinion, I apologize to OP . My criteria/thresholds for what is worthwhile or not do not need to match up to the OP's .

Reply 56 of 58, by darry

User metadata
Rank l33t++
Rank
l33t++
digicube wrote on 2021-01-17, 08:14:

That's the CF card I have. I didn't know industrial CF cards are good. They don't advertised speed ratings so I avoid them.

Did you mean to say that you have the Startech CF IDE adapter or that you have an industrial CF card ?

The reason for suggesting an industrial card was to avoid compatibility issues (see Re: What CF-IDE adapter, CF brand/type, and CF max size for... , for example) that may or may not apply to your setup. That Startech adapter definitely supports UDMA, so with a decent CF card and UDMA enabled drivers in the operating system, you should expect up to 33 megabytes sequential transfer rates (EDIT : I don't remember how fast my industrial CF cards are )

I have little experience using CF cards as IDE replacement in PCs as I have mainly used them in cameras and my Roland VS-880EX . For retro PC use, I prefer SSDs and modern SATA drives either with an IDE/SATA adapter (if the onboard IDE controller is LBA/LBA48 capable) or with PCI SATA controllers or even with PCI IDE controllers with an IDE/SATA adapter .

There is nothing wrong with using CF drives as IDE replacement in retro PCs (many people are happily using them), it's just that they have not been an optimal choice for me in most scenarios (large capacity disks).

Reply 57 of 58, by digicube

User metadata
Rank Newbie
Rank
Newbie

I have Startech IDE CF adapter and Patriot 266x 4GB CF card. I don't use 256GB SATA drives as I heard 16bit retro OS has problems detecting anything bigger than 2GB/4GB.
I don't want to get bigger memory or faster CPU because I may want to install DOS in the future.

Reply 58 of 58, by darry

User metadata
Rank l33t++
Rank
l33t++
digicube wrote on 2021-01-17, 18:42:

I have Startech IDE CF adapter and Patriot 266x 4GB CF card. I don't use 256GB SATA drives as I heard 16bit retro OS has problems detecting anything bigger than 2GB/4GB.
I don't want to get bigger memory or faster CPU because I may want to install DOS in the future.

A faster CPU can indeed be a hindrance when trying to run older speed sensitive DOS software. However, by disabling L1/L2 caches, even a 1400 MHz Tualatin works great for the DOS games I run, including speed sensitive ones like Gateway . My era of interest starts around 1990 (with a few exceptions from 1987ish). I have not yet needed to use software slowdown utilities, but that is an option I intend to try if required . Additionally, if needed, I can always use the BIOS to downclock the FSB to 66MHz to get a 700MHz CPU clock which, with caches disabled, would be quite slow indeed .

DOS 6.22 will not let you use partitions greater than 2GB (FAT16 limit) and a maximum disk size of 8GB . See https://msfn.org/board/topic/159631-testing-m … os-limitations/ I use DOS 7.1 (bundled with Windows 98 SE) as it has not trouble with 127GB FAT32 partitions on hard drives up to at least the 1 TB range . Running Windows 98 SE on such partition and hard drive sizes requires a few easy to install patched, but using just DOS 7.1 works fine without any patching (I would avoid using it's included FDISK, however).

I personally do not see a point in running DOS 6.22 (or any older DOS), Windows 3.1, Windows 95, Windows NT 4 or Windows 2000 . Here is my reasoning :

- DOS 6.22 : too limited in disk size support and I have only ever heard of one DOS game having a (patchable) issue with DOS 7.1 DOS 6.22 and Windows 98 SE Multi Install
- Windows 3.1 : Though I am confident some probably exist, I have never used a Windows 3.1 application that did not work under Windows 9x, either recently or back in the day
- Windows 95 : Windows 98 SE does everything that Windows 95 and usually does it better . The only advantage Windows 95 presents, IMHO, is lower resource usage which would be potentially worth it if running on a 486 or first generation Pentium .
- Windows NT 4 : I don't know of any games that only run under NT 4 and due to lack of Direct3D hardware acceleration, most non-OpenGL 3D accelerated Windows 9x games won't work . Maybe some non gaming applications only run on NT 4, but I don't use any . Windows XP and even 2000 are better choices, IMHO .
- Windows 2000 : Windows XP has better driver support and features . It was a great OS in its time, but XP does everything as well or better . 2000 might be worth consideration if the fear of Microsoft terminating XP activation servers is a concern .

That is why I run Windows 98 SE on one machine (and also on a testbed machine) and Windows XP on another machine . This obviously does not take into account any nostalgia one might have for a given OS . I am more of a pragmatist and am happy as long as I can get my old software to run well on actual old hardware using a compatible operating system, not necessarily the OS that was in fashion when the said software was released .

Please don't take this as an attempt to discourage you from doing whatever you want to do . Experimenting is more than half the fun, IMHO .

Cheers!