Artex wrote:

ODwilly wrote:

Hey Artex, is that a HP dc5700 on the bottom left? I have the same exact system waiting for a fresh install of XP and donation to a friend xD with a xw4600 sitting on top of it waiting for a power supply!

Actually a dc7800 loaded up to the brim with memory and a Core2Quad, Windows 8.1 and a SSD. I use it when I work remotely via Citrix. Seems to run pretty well for being as old as it is. I actually inherited it for free when my father's business upgraded all their machines.

My local School District uses a huge workforce of 7800's and 7900's, pretty nice systems. Free Core2 era machines are awesome, my two came from a local business that recycles their old systems once a single part dies. Got around 30 wiped hard drives (mostly IDE 40-80gb and two Raptors), ddr2, and a quadro fx 560 from their recycle pile free of charge!

sunaiac wrote:

All 486 before the DX4 should perform around the same at same clock speed. What you might have experienced, is a bad motherboard which divides PCI speed with 40MHz FSB, making everything but the CPU slower.

Not only that, I remember that I have read something like Cyrix DX2 80 internal cache speed was not doubled unlike Intel DX2 cpus, and that really crippled it's speed. On a fake cache pci motherboard its speed in games were more like a 386 DX 40 or so 😵 I think I tested it again some years after on my best 486 VLB motherboard (486SV2GX4) and even with it's 80 mhz speed and WB cache it was still no match vs an Intel DX2 66 with WB 😒 (Or I just dreamed all that time 🤣 )

In all case, for me It doesn't worth the effort to try some tests with it again (its place is on a showcase, not in a retro gaming computer 😁 )

Finally received the memory chips necessary to upgrade my ET4000/W32P to 2MB. I wanted to do this since there have been lots of talk of performance increases of 64-bit memory interleaving on these cards but I've failed to see any benchmarks on this before (like this post). Me and Artex previous published results was kind of overlooked because of this (even though Artex tested 1MB S3 vs 2MB ET4000 and I did the opposite and we got the same results)

Here are my results again with 2MB benches:

Hardware:
Pentium Overdrive 83@100mhz
ASUS VL/I-486SV2GX4 Rev. 2.0
32MB RAM (FPM)
1024kb L2 cache
Samsung IDE harddrive

ET4000/W32P 1MB
3DBench2: 95,8
PCPBench: 28 FPS
Doom: 1327 realticks
Quake: 23,6 FPS

ET4000/W32P 2MB
3DBench2: 95,8
PCPBench: 28 FPS
Doom: 1316 realticks
Quake: 23,6 FPS

S3 Vision 864 2MB (#9 GXE64)
3DBench2: 96,6
PCPBench: 27,8 FPS
Doom: 1325 realticks, 56.37 FPS
Quake: 23,6 FPS

Basically there are minimal changes, only a slight increase in DOOM can be noticed on the VGA benchmarks. Maybe the memory interleaving feature has more to it when it comes to Windows GUI acceleration.

I've asked Artex to do a similar benchmark with a ET4000 at 1mb. Artex's results can be seen in this thread: Build 486's And They Will Come! Suggestions please!

Interesting thing, remenber that upgrading my Cirrus logic GD5428 and S3 805 to 2 MB doesn't changed anything either or maybe in the vesa 2.0 res but the cpu is the real bottleneck here 😒

vetz wrote:

No, I didn't. I've tested them in two different boards and there are no problems. I don't think I've been super lucky, but I believe the failure rate is somewhere between your 10% and mine 0%. Lets see what Artex experience 😀

Well, after waiting for a few weeks now with no updates to the tracking information I was provided, I contacted AliXpress again this morning and found that although they issued a tracking # for the shipment of my (incorrect) 15ns chips, they actually never shipped any parcel because the -10ns chips I wanted were not in stock. Of course they never communicated this to me until today...

Now the onus is on me to 'apply for a refund' because I no longer want to deal with this company. Unfortunately, this is one scenario where the language barrier and crappy customer service are to blame. Unless I can source these chips from an English-speaking/writing location for a decent price, I think this L2 cache upgrade is dead in the water.

darksheer wrote:

sunaiac wrote:

All 486 before the DX4 should perform around the same at same clock speed. What you might have experienced, is a bad motherboard which divides PCI speed with 40MHz FSB, making everything but the CPU slower.

Not only that, I remember that I have read something like Cyrix DX2 80 internal cache speed was not doubled unlike Intel DX2 cpus, and that really crippled it's speed. On a fake cache pci motherboard its speed in games were more like a 386 DX 40 or so 😵 I think I tested it again some years after on my best 486 VLB motherboard (486SV2GX4) and even with it's 80 mhz speed and WB cache it was still no match vs an Intel DX2 66 with WB 😒 (Or I just dreamed all that time 🤣 )

In all case, for me It doesn't worth the effort to try some tests with it again (its place is on a showcase, not in a retro gaming computer 😁 )

Fake cache PCI motherboard? There is a strong chance you were using a PC Chips board. The 486-based PC Chips board I have implements a hidden and automatic 2/3 PCI multiplier when the FSB is set to 40 MHz or higher. I discovered this after running the ultimate 486 benchmark comparison. If your board was like mine, your FSB would be running at 27 MHz instead of 33 MHz. Read the U4BC thread to determine which benchmarks were effected. The Cyrix DX2/DX4 was indeed inferior to the Intel/AMD DX2/DX4 chips of the same clock frequency. This inferiority combined with a slow PCI bus and fake L2 cache likely accounts for your negative experience.

The Cyrix DX4-100 performs approx. 13% slower than an Intel DX4-WB.

The Cyrix DX2-66 performs approx. 5% worse than an Intel DX2-66-WT.

Do you actually own an Intel DX2-66-WB? Those did not seem very common among consumers. I have the gold-capped chip. There also exists a black-capped chip. Which do you have?

Artex wrote:

vetz wrote:

No, I didn't. I've tested them in two different boards and there are no problems. I don't think I've been super lucky, but I believe the failure rate is somewhere between your 10% and mine 0%. Lets see what Artex experience 😀

Well, after waiting for a few weeks now with no updates to the tracking information I was provided, I contacted AliXpress again this morning and found that although they issued a tracking # for the shipment of my (incorrect) 15ns chips, they actually never shipped any parcel because the -10ns chips I wanted were not in stock. Of course they never communicated this to me until today...

Now the onus is on me to 'apply for a refund' because I no longer want to deal with this company. Unfortunately, this is one scenario where the language barrier and crappy customer service are to blame. Unless I can source these chips from an English-speaking/writing location for a decent price, I think this L2 cache upgrade is dead in the water.

Did you pay via Paypal or escrow? And did you get your dough back? Why not utilise the same seller that vetz used? His source seems to have a 0% failure rate 😉

Paypal wasn't accepted - only CC and I did use the same seller.. Supposedly the refund was processed and should be back to my account in a few days.

http://www.aliexpress.com/item/IS61C1024-10N- … 1832174608.html

Same seller yes, but I didn't know I would receive 15ns chips which I warned Artex about. Artex made a case for the 10ns ones. I'm pretty sure if you just accepted the 15ns ones you would've had them by now.

vetz wrote:

Same seller yes, but I didn't know I would receive 15ns chips which I warned Artex about. Artex made a case for the 10ns ones. I'm pretty sure if you just accepted the 15ns ones you would've had them by now.

Yes, and that's where the language barrier comes into play. And really the crappy customer service too where he over-promised the -10ns, completed the order, gave me a shipping/tracking number and then told me he didn't have any -10ns chips in the warehouse.. but of course ONLY after I inquired about the shipment status.

Leave poor feedback on aliexpress?

feipoa wrote:

Leave poor feedback on aliexpress?

Yeah, unfortunately in this case I think I'll have to. Hate doing that....

Were you able to solder on the jumper and establish 40 MHz operation? How is the progress with the Saturn-II board?

feipoa wrote:

Fake cache PCI motherboard? There is a strong chance you were using a PC Chips board. The 486-based PC Chips board I have impl […]
Show full quote

Fake cache PCI motherboard? There is a strong chance you were using a PC Chips board. The 486-based PC Chips board I have implements a hidden and automatic 2/3 PCI multiplier when the FSB is set to 40 MHz or higher. I discovered this after running the ultimate 486 benchmark comparison. If your board was like mine, your FSB would be running at 27 MHz instead of 33 MHz. Read the U4BC thread to determine which benchmarks were effected. The Cyrix DX2/DX4 was indeed inferior to the Intel/AMD DX2/DX4 chips of the same clock frequency. This inferiority combined with a slow PCI bus and fake L2 cache likely accounts for your negative experience.

The Cyrix DX4-100 performs approx. 13% slower than an Intel DX4-WB.

The Cyrix DX2-66 performs approx. 5% worse than an Intel DX2-66-WT.

Do you actually own an Intel DX2-66-WB? Those did not seem very common among consumers. I have the gold-capped chip. There also exists a black-capped chip. Which do you have?

I don't really know whose the motherboard manufacturer and model in question (editing its bios show the letters AP => could be AMPTRON but theirs 486 PCI do not look exactly the same, even if it's also based on the sis496/497). I found similar models but not the exact one (tried differents bios with unlocked settings from similar boards but doesn't work), the motherboard itself is not as shitty as we can think of, because even without L2 cache I remeber it has some decent mem speed and cpu score on speedsys and a good amount of bios and jumper settings (glad the manual was with it, but the motherboard is only referenced as PCI-BUS 😵 ). I tried true SRAMS on it (128KB) and it seems to work because L2 cache is shown as 128 KB instead of 256 KB and it emits some heat (show that the traces are good) but with the default bios forcing the L2 cache as disable or f*cking things up, it hang at the boot sequence 😢 .

I remenber that with a Cirrus Logic pci card and a WT DX4-100 It could run Duke3d smooth by reducing the screen size a bit (by using 5/6 of the screen, or something like that...), even without L2 😀 But at that time I was like "meh want to play duke3d fullscreen with music and sound effects set to max 🤣 "

My DX2 66 WB is also a gold-capped chip (SX955) with 94301616NB serial number (that's a pitty I only own one WB cpu capable motherboard with VLB that I already use with an Intel DX4 100, the other is the 486 PCI with fake cache, elsewhere I would build a 486 DX2 66 WB rig with my GUS MAX that I don't use ATM 🤣 )

How does a gold capped 486 look like, you don't mean the backside? cuz they almost all are gold capped there...

You could try this util, should be able to enable/disable L1, L2 cache in dos. not sure if it works or not, never tried.

JaNoZ wrote:

How does a gold capped 486 look like, you don't mean the backside? cuz they almost all are gold capped there...

You could try this util, should be able to enable/disable L1, L2 cache in dos. not sure if it works or not, never tried.

I thought he was speaking about the backside cause I never heard of gold capped die for Intel 486 DX2 before 😵
Thanks for the utility but I can't even boot on a floppy disk... it fails just after updating ESCD with L2 cache present 😢

I beleive Amptron was PC Chips. Aside from them using a fake cache scheme, PC Chips 486 boards were not as bad as some people have made them out to be. The M919 v3.4 performs OK. It has some SCSI bus mastering issues that are tolerable as they are mostly an issue with W2K installations. The v3.4 board is limited to 256K cache due to the COAST-like module required. The auto implementation of the 2/3 PCI bus multiplier is a large annoyance as well.

I was indeed refering to the dye cap. Some SX911 and SX955 were black. I don't know why Intel made a few black caps - perhaps to save on gold?

feipoa wrote:

I beleive Amptron was PC Chips. Aside from them using a fake cache scheme, PC Chips 486 boards were not as bad as some people have made them out to be. The M919 v3.4 performs OK. It has some SCSI bus mastering issues that are tolerable as they are mostly an issue with W2K installations. The v3.4 board is limited to 256K cache due to the COAST-like module required. The auto implementation of the 2/3 PCI bus multiplier is a large annoyance as well.

I was indeed refering to the dye cap. Some SX911 and SX955 were black. I don't know why Intel made a few black caps - perhaps to save on gold?

The motherboard I own don't look like the pchips and similar models because it has 4x simm 30 + 1x simm 72 and don't have any coast module socket, just 5 fake cache chips (DIP 28) directly soldered on DIP 32 emplacements on the motherboard. I never found a model that perfectly match the same pcb anywhere on the web 😵

I recall there being at least one other "company" which produced fake soldered cache motherboards.

feipoa wrote:

Were you able to solder on the jumper and establish 40 MHz operation? How is the progress with the Saturn-II board?

Sorry I missed this in the thread replies...

Short answer is... not yet. I've been spending some time de-yellowing some of the yellowing plastic on these AT cases and I haven't had much time for further motherboard work. I hope to get back to it soon though.