curggles@gmail.com wrote on 2023-04-07, 22:32:

I have had the clock multiplier on the 3x setting and clk at 50mhz, 150MHZ cpu and haven 't even seen any signs of stability issues.

Wow, but to my thinking 150 MHz seems too good to be true. I have an Am486 DX2 for which my SIV utility reported 28 Dhrystones, what do you see @ 150 MHz?

red-ray wrote on 2023-04-07, 23:10:

curggles@gmail.com wrote on 2023-04-07, 22:32:

I have had the clock multiplier on the 3x setting and clk at 50mhz, 150MHZ cpu and haven 't even seen any signs of stability issues.

Wow, but to my thinking 150 MHz seems too good to be true.

150MHz seems quite possible. The V16BGC variant contains the second revision of the 5x86 core, which was targeted to hit 133MHz on every non-defective unit. It's likely that AMD stopped producing the old SV8B variant by that time, and just repackaged 5x86 processors strapped to support x2 instead of x4 to meet market demands for the DX2-66 processors.

red-ray wrote on 2023-04-07, 23:10:

Wow, but to my thinking 150 MHz seems too good to be true.

Yes, it looks like a declassed 133 unit, so no wow at all.
There have been rumors that AMD hasn't gone past 133 MHz with their 486's in numbers to not cannibalize their not so famous 5K86 and K5 predecessor CPUs.
Not few of the 5x86 (486 DX5) should work with 150 (3x50) and 160 (4x40) MHz but are sold as 133 MHz units. Not enough of this, it seems that at this time there have been high requests for additional DX2-66 CPUs at this time. Since there were reasons to not produce new charges of old DX2 dies, AMD put a large amount of 5x86 dies that easily could run 133 MHz and more into a package of a DX2 and labelled them Enhanced DX2-66 and let the doubled amount of 16 KB L2 cache enabled and operating in write back mode. Well, they in fact did not have much to do with a DX2-66 except than marketing - but luckily for overclockers you easily could get them to 50 MHz FSB clock, and they still allowed to raise the multiplicator from 2 to 3 too.
However, that kind of declassed CPU will probably run fine without any heatsink at all at 66 MHz, but I do not recommend this with higher clocks.

Operating a 486 board with 50 MHz FSB is, of course, another thing.

mkarcher wrote on 2023-04-08, 07:09:

The V16BGC variant

Thank you for the clarification, I had not picked up on the 16KB L1 cache and now suspect the V16BGC will support CPUID, curggles@gmail.com please will you post the Menu->Hardware->CPUID-CPU-0 panel so I can check?

Below you can see the [_]CPUID is not checked and will be checked if your DX2 has CPUID support.

Update: I wonder if SIV get's the correct L1 cache size, on the initial screen does SIV report 8KB or 16KB for the L1 Unified Cache, what does [ Cache-0 Latency] report?

Interesting. But how do read that contents in the EAX, EBX, ECX and EDX register then, if that CPU does not have CPUID? Have I missed something?
Is EFLAGS bit 21 ("ID") set on this CPU?

I don't own one of these, but I've been aware of them for some time. I think in 2016 I convinced Feipoa to try one at 2x66, which it did without any issues. Perhaps 2x66 is the only reason you might want to own one versus a regular 5x86...or maybe 3x60 if you need some extra B1 steppings to attempt 180.

Disruptor wrote on 2023-04-08, 07:26:

and labelled them Enhanced DX2-66 and let the doubled amount of 16 KB L2 cache enabled and operating in write back mode

They've also sold 16Kb CPUs as 8Kb CPUs.

AMD AM486DX4-100SV8B and AM486DX2-66: Interesting observation / CPU relabeling

Sure. But I'm not so sure about the overclocking potential of the 8 kB WB-Cache 486 DX4's from AMD.

However, we use an Am486 DX4 120 SV8B as DX2 120 (2x60 MHz). It houses in a 25398 package.

Disruptor wrote on 2023-04-08, 07:26:

.....
Not few of the 5x86 (486 DX5) should work with 150 (3x50) and 160 (4x40) MHz but are sold as 133 MHz units.
......

I know what you mean, but I still disagree with this statement.
I have a lot of 5x86 133 CPUs. ALL of them work at 160MHz. Without proper cooling they will crash after a while, but cooling an AMD 5x86 is not really a challenge.
Beyond 160MHz (180MHz) things change, then (more or less) none of it seem to work properly with air cooling.

red-ray wrote on 2023-04-08, 08:10:

Thank you for the clarification, I had not picked up on the 16KB L1 cache and now suspect the V16BGC will support CPUID, curggle […]
Show full quote

mkarcher wrote on 2023-04-08, 07:09:

The V16BGC variant

Thank you for the clarification, I had not picked up on the 16KB L1 cache and now suspect the V16BGC will support CPUID, curggles@gmail.com please will you post the Menu->Hardware->CPUID-CPU-0 panel so I can check?

Below you can see the [_]CPUID is not checked and will be checked if your DX2 has CPUID support.

Update: I wonder if SIV get's the correct L1 cache size, on the initial screen does SIV report 8KB or 16KB for the L1 Unified Cache, what does [ Cache-0 Latency] report?

What program is this you are using? If I can get it I will post the info. I have just been using Phil's dosbenchmark collection. Other than that I have ran the system for hours gaming and such and haven't had any odd things yet. I was kind of expecting some issue but so far haven't had one while overclocked.

curggles@gmail.com wrote on 2023-04-14, 14:43:

red-ray wrote on 2023-04-08, 08:10:

Thank you for the clarification, I had not picked up on the 16KB L1 cache and now suspect the V16BGC will support CPUID, curggle […]
Show full quote

mkarcher wrote on 2023-04-08, 07:09:

The V16BGC variant

Thank you for the clarification, I had not picked up on the 16KB L1 cache and now suspect the V16BGC will support CPUID, curggles@gmail.com please will you post the Menu->Hardware->CPUID-CPU-0 panel so I can check?

Below you can see the [_]CPUID is not checked and will be checked if your DX2 has CPUID support.

Update: I wonder if SIV get's the correct L1 cache size, on the initial screen does SIV report 8KB or 16KB for the L1 Unified Cache, what does [ Cache-0 Latency] report?

What program is this you are using? If I can get it I will post the info. I have just been using Phil's dosbenchmark collection. Other than that I have ran the system for hours gaming and such and haven't had any odd things yet. I was kind of expecting some issue but so far haven't had one while overclocked.

You can find red-ray's SIV program in the first post of this thread. SIV support for 386/486/586 class + Alpha CPUs and 3dfx + S3 + SiS + Matrox + XGI + old ATI + NVidia GPUs - Testing Help

red-ray wrote on 2023-04-08, 08:10:

Thank you for the clarification, I had not picked up on the 16KB L1 cache and now suspect the V16BGC will support CPUID, curggle […]
Show full quote

mkarcher wrote on 2023-04-08, 07:09:

The V16BGC variant

Thank you for the clarification, I had not picked up on the 16KB L1 cache and now suspect the V16BGC will support CPUID, curggles@gmail.com please will you post the Menu->Hardware->CPUID-CPU-0 panel so I can check?

Below you can see the [_]CPUID is not checked and will be checked if your DX2 has CPUID support.

Update: I wonder if SIV get's the correct L1 cache size, on the initial screen does SIV report 8KB or 16KB for the L1 Unified Cache, what does [ Cache-0 Latency] report?

siv32l win 95 doesn't give me an option to check cpu Id. This program is also only reporting 8kb of catch but if I use other tools it reports 16kb and even watchingemory performance the first drob is after 16kb l1.