Just tested it last week on an i430FX (slightly slower, but same ballpark). Difference was just over 5% in RAMspeed FLOATmem an INTmem.

On Intel 430FX (Intel board with no tuning) and 200MHz MMX:

Doom framerate: +5% (low res)
Quake framerate: +6% (320x200)

On the HX both FPM and EDO bursts are one cycle faster compared to the FX so the proportion will be about the same.

Nobody with an HX board?

Just benchmarked this on my GA-586HX. Comparing EDO x-2-2-2 and FPM x-3-3-3, the EDO returns:

Doom: +5% (low res) / +2% (high res)
Quake: +5%

In absolute numbers the HX with FPM is still a tiny bit faster than FX with EDO...

EDO x-2-2-2 and FPM x-3-3-3

Is this what the BIOSes' AUTO feature sets it to? If FPM and EDO use the same BIOS timings, is there any difference in benchmark results?

On my Proserva V Plus system, which is running dual P233MMX chips, the BIOS options for memory are only 60ns or 70ns, and 3 options: ECC, Parity, None. Both EDO and FPM work on the 60 ns setting. I ran some tests without parity or ECC at 60 ns.

FPM
Quake @ 320x200 = 50.5, 50.4, 50.5, 50.4, 50.5 = Average of 50.46 fps
Quake @ 640x480 = 18.0, 18.0, 17.9, 18.0 = Average of 17.98 fps
DOOM = 751, 751 = Average of 99.45 fps

EDO
Quake @320x200 = 50.0, 50.1, 50.1, 50, 50.1 = Average of 50.06 fps
Quake @ 640x480 = 17.9, 17.9, 17.9, 17.9 = Average of 17.9 fps
DOOM = 753, 752 = Average of 99.32 fps

In summary, FPM was 0.8% faster than EDO in Quake at 320x200, and 0.45% faster in Quake at 640x480. FPM was 0.26% faster than EDO in DOOM.

These results are perplexing and was the driving force for starting this thread. At the same memory timings, does EDO have any benefit over FPM, or does FPM have the edge on EDO given this criterion?

I do not think the BIOS changes timing between the EDO and FPM on these boards. I did run a simple memtest a while ago and found FPM with Parity off was a bit faster than Parity on but not by much. Lack of advanced BIOS memory timing settings is most likely the reason. added: the minor differences you see are probably more to how the BIOS+chipset handles the two types without those advanced timings.

You think some of these hidden settings might work with EDO instead of FPM? They didn't work with FPM.

WOW ! Yes the first two plus RAS precharge could useful but need to find my bios mem manual to be sure. Odd changes could also prevent a bootup but worth a try.

Unfortunately, the first two are already at their lowest values.

feipoa wrote on 2020-03-22, 00:57:

EDO x-2-2-2 and FPM x-3-3-3

Is this what the BIOSes' AUTO feature sets it to? If FPM and EDO use the same BIOS timings, is there any difference in benchmark results?

This is on manual with timing set to these values.

This is the best timing HX can do with EDO or FPM. Having faster burst read cycle is the feature of EDO. In theory I could slow the EDO down to x-3-3-3 which would obviously remove any performance benefit over FPM.

The write cycle is x-2-2-2 for both EDO and FPM.

feipoa wrote on 2020-03-22, 05:22:

You think some of these hidden settings might work with EDO instead of FPM? They didn't work with FPM.

yes the first setting - (x222) is value for EDO and the one after slash (x333) is for FPM.

The options for DRAM Burst Timing are:
x222 / x333
x333 / x444
x444 / x444

The system automatically defaults to x222 / x333 when 60 ns EDO or FPM is used, yet benchmarks results for FPM are a smidge better.

On your 430hx system, does 50 or 60 ns FPM run reliably on x222?

So looks like FPM can work just as fast as EDO, if timings are tweaked.

I don't think you can drive FPM RAMs at x-2-2-2 at 66 MHz no matter if 60ns or 50ns.

Unlike EDO the FPM cannot remove CAS during data transfer and thus it needs extra cycle and at best three cycles for burst transfer. The initial cycle is the same as in EDO.

As I mentioned above the "x222 / x333" setting in BIOS means that EDO will be clocked 5-2-2-2 and FPM 5-3-3-3. So 4 sequential transfers will take 11 cycles instead of 14. Resulting in 27% best case improvement which translates in the 5% real-world improvement observed in Quake.

feipoa wrote on 2020-03-22, 01:49:

In summary, FPM was 0.8% faster than EDO in Quake at 320x200, and 0.45% faster in Quake at 640x480. FPM was 0.26% faster than EDO in DOOM.

i recently found that using the same timings with different sticks can cause differences that are right in that ballpark, even when each pair is e.g. 60ns and EDO. size is another factor, just adding extra memory could repeatably reduce pcpbench scores by 0.1fps.

OEM systems with dual CPUs are often a bit exotic. They are likely to be using more than 4 rows of RAM which require buffering. This means that both lead-off and burst rates are increased by one to account for buffer delay. This effectively eliminates aby benefits of EDO. Or they could be penalised by ECC/Parity checks (irrespectively to what you set in BIOS).

But normal situation in normal system, the EDO is faster than FPM which should be clearly obvious in RAM sequential read speed tests as well as in Quake (low res). There is nothing about EDO per se that could reduce performance compared to FPM if everything else is equal.

The RAM modules do not contain buffers, just a parity chip. And it must use FPM or EDO with that parity chip otherwise the screen stays blank at power-up. I suppose I could test some different FPM modules like auron suggested to see if there's any difference.

The buffering doesn't have to be on module. It can also happen on the motherboard - look for 74FCT3245A devices or similar close to SIMMs slots.

How many memory slots that can take double density modules are on your board? If more than 4 then there must be some buffering involved.

The 430HX chipset has actually a duplicate copy of MA signals which optionally allows for direct connection to SIMMs (without buffer) and maintain x-222 EDO burst performance on 6-8 SIMM socket motheboards when less than 4 SIMMs are used. However, not every motherboard used this. My GA-586HX does. The HX chipset was a weirdo and often deployed in servers where having more RAM was more important than performance.

AFAIK there are also HX boards with DIMM slots (EDO only) and they are always buffered resulting in x-333 fastest timing.