dominusprog wrote on 2024-05-04, 10:56:

What’s the point of installing a DX4/586 processor without L2 cache?

Since there's no cache upgrading stick available, I just test as is.

I just bought a real Cyrix 5x86 120GP today, hopes Doom benchmark will run smoothly without a glitch.

385387386 wrote on 2024-05-15, 06:21:

dominusprog wrote on 2024-05-04, 10:56:

What’s the point of installing a DX4/586 processor without L2 cache?

Since there's no cache upgrading stick available, I just test as is.

I just bought a real Cyrix 5x86 120GP today, hopes Doom benchmark will run smoothly without a glitch.

40fps on my 586-120gp without the cyrix optimisations enabled, that's on a VLB based system with rather mediocre memory performance. It rises slightly to 41fps with the enhancements turned on (not worth it for the slight instability they bring to other titles).

https://www.target-earth.net/wiki/doku.php?id … pca6145#results

I found the AMD X5 133 to be almost exactly the same performance as the 586-100. At 120MHz the Cyrix jumps ahead in most areas. My board doesn't have any voltage options (3.3 or 5 only), so I haven't pushed it to 133. THe really good thing about the Cyrix is that the multiplier is software controlled, so you get the benefit of the one system running at 586-120/am5x86-133, 486dx2-66, 486dlc-40 and 386dx-40 speeds just by various combinations of the multiplier and the L1 cache.

megatron-uk wrote on 2024-05-15, 08:48:

40fps on my 586-120gp without the cyrix optimisations enabled, that's on a VLB based system with rather mediocre memory performa […]
Show full quote

385387386 wrote on 2024-05-15, 06:21:

dominusprog wrote on 2024-05-04, 10:56:

What’s the point of installing a DX4/586 processor without L2 cache?

Since there's no cache upgrading stick available, I just test as is.

I just bought a real Cyrix 5x86 120GP today, hopes Doom benchmark will run smoothly without a glitch.

40fps on my 586-120gp without the cyrix optimisations enabled, that's on a VLB based system with rather mediocre memory performance. It rises slightly to 41fps with the enhancements turned on (not worth it for the slight instability they bring to other titles).

https://www.target-earth.net/wiki/doku.php?id … pca6145#results

I found the AMD X5 133 to be almost exactly the same performance as the 586-100. At 120MHz the Cyrix jumps ahead in most areas. My board doesn't have any voltage options (3.3 or 5 only), so I haven't pushed it to 133. THe really good thing about the Cyrix is that the multiplier is software controlled, so you get the benefit of the one system running at 586-120/am5x86-133, 486dx2-66, 486dlc-40 and 386dx-40 speeds just by various combinations of the multiplier and the L1 cache.

Hi, what doom timedemo? because in phil benchmark doom timedemo i get 54fps on a cx5x86 120

Yes, the Phil benchmark pack timedemo.

54 seems awfully high.

megatron-uk wrote on 2024-05-15, 08:48:

I found the AMD X5 133 to be almost exactly the same performance as the 586-100. At 120MHz the Cyrix jumps ahead in most areas.

This matches my experience. AMD 5x86 at quadruple clock has a similar overall performance to a Cyrix 5x86 (with branch prediction enabled) at triple clock. So parity is reached at AMD@133 to Cyrix@100 or AMD@160 to Cyrix@120. As it is way easier to get AMD 5x86 processors than Cyrix 5x86 processors, and furthermore nearly every AMD 5x86-133 overclocks without any issues to 160, but you need to push the Cyrix 5x86 100 hard to get it stable at 120, this leaves the Cyrix processors to Cyrix afficionados and curiosity hunters. If you just want a fast socket 3 system, the AMD system is the way to go. On the other hand, if you want to have some stories to tell how you acquired and tweaked your system, building a Cyrix system is a nice starting point. I for myself enjoy both processors in their own way. The Cyrix@120 is a very interesting option if you can get your board to perform good at FSB60. The Cx5x86 clearly beats a clock-doubling DX2/DX4 when both processors run at 2*60, and getting 3*60 out of an AMD 5x86 is finicky. If you get 180MHz from your Am5x86 on the other hand, it will beat the Cx5x86 at 2*60 in basically every task, though.

mkarcher wrote on 2024-05-15, 17:55:

... and furthermore nearly every AMD 5x86-133 overclocks without any issues to 160...

Does this include the Am5x86-133 ADW?

analog_programmer wrote on 2024-05-15, 18:14:

mkarcher wrote on 2024-05-15, 17:55:

... and furthermore nearly every AMD 5x86-133 overclocks without any issues to 160...

Does this include the Am5x86-133 ADW?

It seems the chance to get to 160 is lower on the ADW: Re: AM5x86 @160 / @200 Stability Testing

On the other hand, the ADW is just specified to have a 30°C lower maximum temperature than the ADZ, so you might even get that processor to 160MHz given sufficient peltier cooling.

mkarcher wrote on 2024-05-15, 19:42:

It seems the chance to get to 160 is lower on the ADW: Re: AM5x86 @160 / @200 Stability Testing

On the other hand, the ADW is just specified to have a 30°C lower maximum temperature than the ADZ, so you might even get that processor to 160MHz given sufficient peltier cooling.

Yeah, because of its lower maximum junction(?) temperature, I think this widespread Am5x86-133 ADW (first revisions Am5x86) CPU is not appropriate for overclock.

Thanks for the conformation.

theelf wrote on 2024-05-04, 16:49:

dominusprog wrote on 2024-05-04, 10:56:

What’s the point of installing a DX4/586 processor without L2 cache?

I have a cyrix 5x86 without L2. After many test, i checked the L2 cache did not make almost any positive difference, in fact, oposite

Did you have more memory than the cacheable area for your chipset at the L2 size used? If so you will see a decrease in performance. I see retro youtubers make this mistake all the time.

kingcake wrote on 2024-05-15, 21:09:

theelf wrote on 2024-05-04, 16:49:

dominusprog wrote on 2024-05-04, 10:56:

What’s the point of installing a DX4/586 processor without L2 cache?

I have a cyrix 5x86 without L2. After many test, i checked the L2 cache did not make almost any positive difference, in fact, oposite

Did you have more memory than the cacheable area for your chipset at the L2 size used? If so you will see a decrease in performance. I see retro youtubers make this mistake all the time.

I think my ram and L2 are correct, 256kb L2 and 32MB ram

is not i see a decreace in performance, is that L2 make almost no performance gain at all, and because i need to lower to 3-1-1-1 from 2-1-1-1 with L2 disabled if i want to use L2 stable, in fact the system is slower

For example, with 256kb L2 enabled and 2-1-1-1 i can still boot, but most test fails, windows did not boot, doom is unstable, quake hangs... and gain is very low, 3dbenchc up from 101.2 to 103.1, 2% gain, Doom from 54.2 to 54.9 fps... 1% gain... with 3-1-1-1 system is like 10-15% slower

kingcake wrote on 2024-05-15, 21:09:

theelf wrote on 2024-05-04, 16:49:

dominusprog wrote on 2024-05-04, 10:56:

What’s the point of installing a DX4/586 processor without L2 cache?

I have a cyrix 5x86 without L2. After many test, i checked the L2 cache did not make almost any positive difference, in fact, oposite

Did you have more memory than the cacheable area for your chipset at the L2 size used? If so you will see a decrease in performance. I see retro youtubers make this mistake all the time.

No cache installed. Ordered 5 pieces cache PCB today.

Async cache on 486 boards can become a hurdle when high FSB is involved. At some point it will be slower than EDO memory, without handpicking chips that can work on tight timings.