It depends. The Pentium Overdrive is a full Pentium CPU (so yes, it will run all Pentium software). Because of the 32-bit 486 socket/chipset it will be bottlenecked somewhat, so in practice it will perform somewhere around a P75.

The 5x86 is just a 486-class CPU, so just a single pipeline rather than the superscalar two-pipeline Pentium. It compensates for that by being clocked much higher. In best-case scenarios it can perform even better than the P75. Worst-case it is much slower than a P75. Mostly when FPU is involved, such as in Quake. Quake is unplayable on a 5x86, but runs fine on a P83OD or a real P75.

So my choice would be the P83OD.

Refer to this topic.
Should provide you with the answers you're looking for.
The Ultimate 486 Benchmark Comparison
Also of interest is the Cyrix 5x86 CPU. But, availability could be a problem.
However, if you want to play games and run software intended for a Pentium, then I would just obtain a Pentium motherboard and CPU.

jesolo wrote:

Refer to this topic.
Should provide you with the answers you're looking for.
The Ultimate 486 Benchmark Comparison

It seems that it only does some synthetic ALU/FPU tests, which doesn't mean a whole lot for real-world performance.
ALU-performance comparisons between 486 and Pentium are very difficult, because they rely heavily on how well-optimized the code is for the Pentium's dual-issue pipeline.
In a synthetic test it's quite easy to craft code that runs on only one of the Pentium's two pipelines, resulting in 486s with higher clockspeeds easily outperforming a Pentium.
Likewise, you can craft code that makes optimal use of both of the Pentium's pipelines, and also taking advantage of other Pentium features, such as the larger cache or better branch prediction.

In practice, games aimed at P75-level performance will generally be optimized for Pentium.

Correct.
If you refer to the detail of the various benchmarks that were performed, it also included benchmark results on games like Quake and Doom.
So, if you're looking at it from a gaming perspective, then you would look more at those results.
But, then again, I wouldn't even bother running a game, optimised for Pentium class CPU's, on a 486 based PC.

3D games typically run much better on P5 architecture than a 486. I'm not sure if Am5x86 can compete in games. They would have to be integer-based games. When I messed around with a POD, my motherboard would disable the L2 cache. It was however solid at 100 MHz. It was dramatically faster than the Am5x86 160 in games like Quake and Jedi Knight, but Windows seemed quicker with the working L2 and Am5x86 160.

Motherboard compatibility is a major issue with POD. Moreso than Am5x86.

POD has a 32K L1 cache like the Pentium MMX and that probably gives it some interesting performance behavior compared to regular P5/P54.

I must admit, I've always wanted to try out a POD 83 MHz to see how it stacks up against other CPU's but, I haven't been able to find one yet.

The AMD 5x86 also supports L1 write through cache (in case of incompatibilities).
However, there will then be a slight performance hit.

Some motherboards like Asus PVI-486SP3 with old BIOS flash chips that dosnt support the newer BIOS version needed for 5x86 support wont work correctly with the 4x multiplier (it hangs at HDD initialization when you do a soft reboot). The 5x86s 16KB write back cache works fine with the Asus board though as the earlier BIOS supports both the 8KB and the 16KB WB cache versions of the AMD 486DX4-120.

If your hardware isnt stable at 50 MHz FSB then running the AMD 5x86 at 3x40 MHz is the only option when the 4x multiplier dosnt work correctly, at this speed however there is a better alternative and that is the Intel DX4-100 WB which normally also works fine at 120MHz and is about 3-5% faster than the AMD 5x86 at the same clock frequency.

Scali wrote:

jesolo wrote:

Refer to this topic.
Should provide you with the answers you're looking for.
The Ultimate 486 Benchmark Comparison

It seems that it only does some synthetic ALU/FPU tests, which doesn't mean a whole lot for real-world performance.

Aside from Doom and Quake, which games, or real-world tests, would you like me to compare which return some sort of benchmark value and which will run on a 486/POD?

GLQuake is perfectly playable on an X5-160 or Cyrix 5x86 provided you use a Voodoo or GeForce. Reference, A brief comparison of Voodoo-Quake results on a 486

feipoa wrote:

GLQuake is perfectly playable on an X5-160 or Cyrix 5x86 provided you use a Voodoo or GeForce. Reference, A brief comparison of Voodoo-Quake results on a 486

Obviously, since the bottleneck is mainly in the polygon rasterizer, which is optimized for Pentium.
Replace the polygon rasterizer with hardware acceleration, and suddenly things don't matter anymore.
Doesn't say anything about the CPU of course.

I've always wished for something like an Age of Empires or some other strategy game benchmark. Something that's not 3D graphics but hits the CPU hard in other ways like unit pathfinding and AI scripting. Though I have not looked particularly deep into opportunities for that.

Hopefully something will turn up between now and the second iteration of the 486 benchmark comparison, which I intend to include every possible socket3 CPU speed, the UMC U5S, as well as 386-486 chips, like DLC, SXL, DRx, BL3, etc.

swaaye wrote:

I've always wished for something like an Age of Empires or some other strategy game benchmark. Something that's not 3D graphics but hits the CPU hard in other ways like unit pathfinding and AI scripting. Though I have not looked particularly deep into opportunities for that.

I've noticed that the game runs well on Pentium 75mhz with 16mb ram. (bit below the 90mhz requirement)
That's the oldest system I have to test. I tried that game on a Pentium II 333MHz with 256MB ram and surprisingly it still did not have enough cpu power to max out the game! I was playing ROR with the largest map and 8 players and it hit a slowdown sometime during the game.

My best guess to max AOE is with a 100MHz fsb Pentium II 400 or faster. I haven't really played attention to performance on the faster cpus, but probably the slowest Pentium III could max out the game..

computergeek92 wrote:

Which upgrade is faster? And will the Pentium Overdrive 83MHz allow me to play games that require the P5 Pentium architecture? (Such as AOE ROR)

if you motherboard supports both, you'll get that (MB8433):

If it does not, the 5x86 might not run and the POD be slower than a DX4 (MB1433):

Thanks for the very useful info! 😀

I finally managed to obtain a Pentium Overdrive 83MHz CPU for a resonable price and thought I'd compare its performance to my AMD 5x86 133MHz CPU.
I've attached the results from a couple of benchmark programs:

I tested both CPU's on the same hardware, with their L1 cache in write back mode.
System specs:
Motherboard: ASUS PVI-486SP3 (rev 1.2) with BIOS revision 0306 (128KB L2 cache in write back mode)
Memory: 32MB Fast Page Mode RAM
Graphics: ATI 3D Rage II+DVD with 2MB memory (PCI). I also installed a Creative Labs 3D Blaster Banshee 16MB PCI, but didn't notice any difference in the benchmark results
OS: MS-DOS 7.10 (from Windows 95 OSR2.5 - I booted up straight into Real Mode MS-DOS, without loading EMM386.EXE, bypassing the Windows GUI)

My BIOS settings were at the default, but pushing it to the optimal settings caused (in some cases) a hung system. Fortunately, the difference in speed wasn't that significant as most of the BIOS settings are already at their optimal values with the default setting.

My benchmark results were lower than those from the Ultimate 486 Benchmark test, but I suspect that the PCI/VLB combination of this motherboard (refer this topic: Performance comparison of 486 motherboards with VLB-only, PCI-only, and PCI+VLB) and/or a lower L2 cache (128KB instead of 256KB) could be a contributing factor in the final results.

For purposes of this topic and based on the attached results, the AMD 5x86 does prove to be quite a good performer compared to the Pentium Overdrive 83MHz.
As expected, the Pentium Overdrive does provide better results in the Quake timedemo test, but I think either CPU does a good job in a 486 class motherboard and will provide very good performance for any games that were designed to be played on high end 486 systems.
Both CPU's also works quite well in "de-turbo" mode and slows down to more or less the CPU speed of somewhere between an Intel 486DX 33Mhz & an AMD 486DX 40Mhz (not always that visible in the benchmark results, but can be more or less "guessed" by the slow down performance of Doom & Quake).