d1stortion wrote:

The first chipset with SDRAM support was 430VX in early 1996, but P6 CPUs didn't get it until 440LX in 1997, where it was necessary to take full advantage of AGP. Why did they put it so early on the Socket 7 platform?

Because Intel has the power to encourage the change to new standards while small companies can not afford to put SDRAM support and sockets on a board, just because it is new on the market. The real breakthrough for SDRAM came with the L2 caches of the P6 CPUs and FSB100.

Is there a noticeable difference between EDO and SDRAM on that platform, or was it just a "future proof" type of thing?

According to Vetz benches the speed is about the same, with advantages for SDRAM for burst access and EDO for random. The 430VX and TX were only the introduction platforms for SDRAM, as both cache only 64 MB. Maybe Intel hoped that SDRAM would become faster cheaper as the main choice for consumer systems.

Which memory type were 430VX/TX buyers advised to go with, assuming the mobo supported both?

In the time of VX EDO mainly due to the non availability of SDRAM. At the time of TX I would say it was even. As there was no real speed difference the price was the key argument.

System builders also toyed with removing L2 cache and letting EDO and SDRAM make up the difference compared to FPM RAM + L2. Cheaper.

Also, early SDRAM chipsets like 430VX require an older type of SDRAM chips. "2 clk" SDRAM.

I'm pretty sure SDRAM first became prevalent when the Pentium II started gaining popularity. I've encountered K6-2 systems with SDRAM, but never Pentium MMXs.

The IBM PC300GL series used for the Pentium I machines a boards based on 430VX and SDRAM.

Yeah I've looked at his results now, crazy to think that in Doom EDO is a bit faster than SDRAM. Also here:

http://www.anandtech.com/show/72/6 wrote:

The SDRAM performance of the TX chipset isn't "bone shattering" due to a timing problem with the chipset (which has been corrected in the Pentium II successor to the TX chipset, the LX chipset), however the stability of SDRAM and the TX chipset at higher bus speeds is much greater than that of EDO and the HX Chipset.

So those early implementations were quirky... I guess if you had sufficiently fast EDO from a previous system it would have been a no-brainer just reusing that instead of going SDRAM.

It is also funny from today's perspective how they were complaining about TX not having AGP... 🤣 considering how the chipset couldn't make full use of SDRAM anyway, and AGP not bringing any real benefits for several years after that. But Intel marketed it as the next best thing in graphics so from a 1997 perspective it's certainly understandable.

Late 1997, 1998. It's pretty much how the guys said, Intel has the power to push new technologies and AGP and 100MHz FSB made SDRAM's case 😀

I'd think more 1996....

Also don't confuse SDRAM with standard 5v FPM/EDO DIMMs. There were a handful of PC boards that took them, otherwise they were almost exclusively used on pre-G3 PCI Powermacs.

a K6-III with SDRAM seems like a good idea because of its on chip L2 cache

While all my Super Socket 7 boards have SDRAM, when I was little I had a Pentium 133 and that had 72 pin RAM. It was only with the Pentium II 300 that I first saw SDRAM 😀

Some Socket 7 boards have both RAM options available.

Also, I have a 16MB SDRAM memory and it gets detected as EDO memory. Also benchmarks slower. How does this work?

Early '98 is when I started seeing SDRAM everywhere.
I don't have an EDO S7 board, but I do have an EDO Slot 1 board as well as a standard BX board. SDRAM is hugely faster than EDO when it is used with a P2 CPU. My P2-300 barely outperforms a 250MMX on the FX/EDO board, but it cruises along on the BX/SDRAM board.

For a K6-2 system, how much of a difference does using SDRAM make vs using EDO?

mr_bigmouth_502 wrote:

For a K6-2 system, how much of a difference does using SDRAM make vs using EDO?

In DOS I eyeball it at around 20-40%% difference? But that's just a rough guess. I haven't looked into this in greater detail.

I'm going to go with 1997. SDRAM wasn't really worth it in socket7 systems except that you didn't need to install in pairs.

mr_bigmouth_502 wrote:

For a K6-2 system, how much of a difference does using SDRAM make vs using EDO?

According to my results on a VIA MVP3 board there is a 1-2% performance increase. Note that this was on a Pentium 100. If this difference increases with a fast K6-2 I don't know.

Mau1wurf1977 wrote:

mr_bigmouth_502 wrote:

For a K6-2 system, how much of a difference does using SDRAM make vs using EDO?

In DOS I eyeball it at around 20-40%% difference? But that's just a rough guess. I haven't looked into this in greater detail.

vetz wrote:

mr_bigmouth_502 wrote:

For a K6-2 system, how much of a difference does using SDRAM make vs using EDO?

According to my results on a VIA MVP3 board there is a 1-2% performance increase. Note that this was on a Pentium 100. If this difference increases with a fast K6-2 I don't know.

The reason why I ask is because I have a K6-2 system in storage, an old Packard Bell to be exact, and the Win98SE install on it seems to be rather snappy, despite only having EDO ram. I also remember coming across a similar, albeit later system a couple years back at the same recycling center, but I turned it down because it had an AGP slot, and I've heard that SS7 boards with AGP slots are usually duds. I did snatch the CPU from it though, I believe it's either a 350 or 500MHz model, though I can't remember specifically.

Wtf? Considering that AGP slots are one of the defining features of the platform and I can't even think of any SS7 boards without AGP... perhaps you mixed it up with the quirky AGP implementation on those chipsets, but I don't think that is enough to call all of those boards garbage.

mr_bigmouth_502 wrote:

The reason why I ask is because I have a K6-2 system in storage, an old Packard Bell to be exact, and the Win98SE install on it seems to be rather snappy, despite only having EDO ram. I also remember coming across a similar, albeit later system a couple years back at the same recycling center, but I turned it down because it had an AGP slot, and I've heard that SS7 boards with AGP slots are usually duds. I did snatch the CPU from it though, I believe it's either a 350 or 500MHz model, though I can't remember specifically.

Hmm I might do some tests on this.

Also I asked this above in the thread. I got a stick of 16MB SDRAM, but it shows up on the POST screen as EDO and also performs lower. It's the only stick that does this.

Also I would need 2x 72 pin sticks for testing right? Currently I only got one working one from the 486, but more coming. But I'll definitely do some memory benchmarks and benchmark memory sensitive games.

d1stortion wrote:

Wtf? Considering that AGP slots are one of the defining features of the platform and I can't even think of any SS7 boards without AGP... perhaps you mixed it up with the quirky AGP implementation on those chipsets, but I don't think that is enough to call all of those boards garbage.

So it's just the AGP implementation that's not any good?