First platform to use dual channel memory?

[lehtv]

I'm having a surprisingly hard time googling an answer to this. Someone help?

What was the first consumer PC platform to use dual channel memory? Was it used at the same time as the first DDR SDRAM?

Thanks!

 

[Golgatha]

I'm having a surprisingly hard time googling an answer to this. Someone help?

What was the first consumer PC platform to use dual channel memory? Was it used at the same time as the first DDR SDRAM?

Thanks!

Intel E7205. I had that hotness back in my college days.

Edit: Well, there was the RDRAM stuff, but it was super expensive and almost no one used it. But, according to this article, it was dual channel as well and did predate dual channel DDR.

https://www.pctechguide.com/chipsets/what-is-the-intel-e7205-chipset-what-features-does-it-have

The Intel 850 chipset would be the first one to support dual channel memory and it was introduced in Nov 2000.

 

[UsandThem]

I think it was the 845 chipset, not long after the whole Rambus debacle.

https://en.m.wikipedia.org/wiki/List_of_Intel_chipsets

Edit: I read the question wrong, person above had right answer.

 

[lupi]

RDRAM, wow what a lulz memory.

 

[dave_the_nerd]

Does memory interleaving "count" as dual channel? Because that was around in the early '90s.

Also, I think the nForce deserves an honorable mention in the "consumer" category, since it supported existing Socket A Athlon CPUs (from 1999), supported dual channel DDR, and wasn't a server/workstation part. Came out in 2001. Back when upgrading a motherboard sometimes made more sense than upgrading a CPU.

 

[dave_the_nerd]

The Intel 850 chipset would be the first one to support dual channel memory and it was introduced in Nov 2000.

If you're counting the 850 (Dual channel RDRAM) the 840 (same) would have been earlier. (1999).

 

[VirtualLarry]

I was going to say 845G or 865G. But I guess the RDRAM chipsets predate those.

 

[razel]

Does memory interleaving "count" as dual channel? Because that was around in the early '90s.

That's what I was thinking too. I remember having RAM sticks be in pairs only in the 486/Cyrix days.

 

[Golgatha]

That's what I was thinking too. I remember having RAM sticks be in pairs only in the 486/Cyrix days.

That didn't have anything to do with parallelism. You had to match the width of the system buses and each chip was 8-bits wide, so you needed 32 bits (4 chips) for 386/486 and some early Pentium systems.

http://www.tomshardware.com/reviews/PARALLEL-PROCESSING,1705-2.html

So, you increased bandwidth by widening the bus on those older systems, but they were still single channel systems. In dual channel systems, you distribute the data over both channels, the bus is wider (128 bit effective), and in the case of DDR2/DDR3/DDR4 you're sending more bits per clock cycle.

 

[0ldman79]

Necro-post, but for archival purposes for anyone else searching this, the Intel 840 chipset for the PIII predates the 850 and is actually the first dual-channel memory controller.

They tried RDRAM on the PIII first in November of 1999. It was a high bandwidth, high latency memory connected to a low bandwidth, low latency interface. Needless to say it did not perform well.

Edit, got my numbers crossed up. I originally said 820, it was the 840. 820 was single channel.

 

[DrMrLordX]

RDRAM dual-channel setups didn't work the same as DDR dual-channel though. If you plugged in extra RIMMs, it would increase system latency. So there was little benefit to adding extra RIMMs.

If that can still be thought of as dual-channel then so be it. The first dual-channel system you would actually want would be E7205 (Granite Bay).

 

[bononos]

RDRAM dual-channel setups didn't work the same as DDR dual-channel though. If you plugged in extra RIMMs, it would increase system latency. So there was little benefit to adding extra RIMMs.
..............

No wonder rambus was a pile of garbage from start to end.

 

[0ldman79]

They were still classified as dual channel. I got it wrong initially, the 820 was single channel, the 840 was dual channel.

They were a poor design, but they probably did help with the evolution of the memory systems, they had some forward thinking features as well as a whole lot of "don't do this if you want to succeed" business practices.

Granite bay was introduced as a replacement for the RDRAM chipsets due to the expense, poor performance and bad publicity that was brought in with it.

I remembered there was a single channel RDRAM, that was the 820. The 840 was dual channel. Both were on the PIII.

 

[SPBHM]

RDRAM on Pentium 4 was dual channel, but if I remember right it ONLY worked in dual channel, a single stick would not work, and it was 2x 32bits.


in any case, the first desktop dual channel DDR (2x 64bits channel, like it still is today) was in 2001 with the Nvidia Nforce for Socket A.
Intel only had dual channel DDR in 2003, they were heavy into RDRAM at the time, and their first DDR chipset 845 was single channel.

 

[0ldman79]

The 840 on the PIII was dual channel in November of 99.

It was funny, the later model PIIIs, the laptops were the only way to get good performance. The RDRAM killed the PIII with it's latency and the 815 SDRAM chipsets were limited to 512MB of SDRAM when they were finally released, initially only available on the mobile platforms. The 810 was on desktop, but it had integrated graphics and I don't believe AGP was even an option. We had a ton of those at the school I worked at. Awesome for students systems, not so good for anything else.

The 815 wasn't available on the desktop for a while.

 

[Insert_Nickname]

The RDRAM killed the PIII with it's latency

Indeed. But the P3 FSB design wasn't really up to running RDRAM either from what I remember. Too narrow. The P4's FSB was purpose designed for RDRAM.

and the 815 SDRAM chipsets were limited to 512MB of SDRAM when they were finally released, initially only available on the mobile platforms.

With RAM prices back then, that would only be an issue for the last Tualatin P3's. 512MB RAM was a huge amount of memory at the turn of the millennium.

 

[0ldman79]

I paid $78 for 512MB for my PIII laptop shortly after I got it. The laptop would have remained useful for a lot longer as it was pretty much the same performance as the Pentium M, but my Pentium M laptop could handle 4GB of RAM, my PIII maxed out at 512.

You're dead on about the PIII bus though, the PIII was designed around mid range bandwidth and low latency. The bus was 64bit 133MHz if memory serves. The memory was 64bit 133MHz. RDRAM ran at 600 or 800MHz, the latency wasn't that great and it had to translate from 133MHz to 800MHz, there's a buffer, sync issues, etc..

They seriously attached a memory standard that would cripple the PIII to the last generation of the PIII. You either had to go with Via or Sis to have memory capacity and SDRAM. Pretty sure Via had a short lived DDR chipset for the PIII. Sis just had poor performance, but you could have more RAM. That whole situation helped AMD a great deal.

 

[Golgatha]

I paid $78 for 512MB for my PIII laptop shortly after I got it. The laptop would have remained useful for a lot longer as it was pretty much the same performance as the Pentium M, but my Pentium M laptop could handle 4GB of RAM, my PIII maxed out at 512.

You're dead on about the PIII bus though, the PIII was designed around mid range bandwidth and low latency. The bus was 64bit 133MHz if memory serves. The memory was 64bit 133MHz. RDRAM ran at 600 or 800MHz, the latency wasn't that great and it had to translate from 133MHz to 800MHz, there's a buffer, sync issues, etc..

They seriously attached a memory standard that would cripple the PIII to the last generation of the PIII. You either had to go with Via or Sis to have memory capacity and SDRAM. Pretty sure Via had a short lived DDR chipset for the PIII. Sis just had poor performance, but you could have more RAM. That whole situation helped AMD a great deal.

This mirrors my purchasing decisions for computer hardware throughout late college and grad school pretty accurately.

I had an ASUS Via 133 chipset motherboard due to the cost of RDRAM and there being no performance advantage, and even a disadvantage in some benchmarks, over SDRAM. I overclocked a P3 Coppermine 700A to 933Mhz with 1GB of SDRAM on that platform, and it lasted me for quite some time. Also, at release, the Pentium 4 was a performance dog at the time when paired with SDRAM.

A couple of years later, I upgraded to another ASUS board with the E7205 (RDRAM was stupid expensive at the time) dual channel DDR chipset, and overclocked a Pentium 4 1.8A to 2.4Ghz on that platform. In between then and there, I tried out an Athlon XP 1600+ system with an nVidia nForce chipset, but it was unstable as all hell from what I recall. So, the performance of AMD was enticing at the time, and I did try their platform, but ultimately went back to Intel for stability.

 

[Arkaign]

The really high end RDRAM setups with P4 were the only ones that ever really showed good results, but the price absolutely wasn't worth it. Later on I got a dirt cheap PC in a larger auction that had 4x256GB PC1066 RDRAM and it was really quick. Think it was on a revised 850 chipset, worked with P4B Northwoods s478, and it definitely was faster at 3ghz+ than my old P4C 3.2 w/DDR333, and faster than my AXP 2500@3200+.

Would be interesting to make a retro RDRAM PC now, wonder if it's gotten too expensive already though.

 

[0ldman79]

The instability on the KT133, KT266 and Nforce chipsets was primarily due to the hard drive controller drivers.

If you disabled DMA access the Via chipsets were far more stable and both platforms were more stable when using the default Windows driver over the official driver.

I would install chipset drivers for all functionality then roll back the IDE controller to MS default. Disabling DMA on Via helped but the MS driver worked without having to do that, later versions didn't have performance loss vs the official drivers.

I ran a KT133, KT266 and Nforce chipset on the same box (lightning is a @#$@#) around that time.

I've got some RDRAM, probably only 512MB or so, but no platform to plug it in to.

 

[Insert_Nickname]

I upgraded to another ASUS board with the E7205 (RDRAM was stupid expensive at the time) dual channel DDR chipset

I have fond memories of that one. It had the performance of the 850E, but at much lower cost.

The P4 really came into it's own with the 875/865 chipsets, which were more-or-less consumer versions of the E7205.

In between then and there, I tried out an Athlon XP 1600+ system with an nVidia nForce chipset, but it was unstable as all hell from what I recall. So, the performance of AMD was enticing at the time, and I did try their platform, but ultimately went back to Intel for stability.

Yeah, the usual suspect was the HDD controller. Drivers were unstable, but you could use the standard MS driver for better results. Don't get me started on VIA or SiS chipsets.

Edit; Oh, ninja'ed by 0ldman79...

 

[Golgatha]

I have fond memories of that one. It had the performance of the 850E, but at much lower cost.

The P4 really came into it's own with the 875/865 chipsets, which were more-or-less consumer versions of the E7205.



Yeah, the usual suspect was the HDD controller. Drivers were unstable, but you could use the standard MS driver for better results. Don't get me started on VIA or SiS chipsets.

Edit; Oh, ninja'ed by 0ldman79...

Yeah, I think my next build was a dual core Pentium D 805 which OCed to 4.0Ghz on an 875 board.

My Via 133 setup was rock solid stable at 933Mhz, but I do recall using a Promise ATA 66 or 100 (can't remember) RAID 0 card for my main drives, which had it's own set of drivers.

 

[0ldman79]

Edit; Oh, ninja'ed by 0ldman79...

lol

You got beaten to the punch by an old guy!

 

[DrMrLordX]

lol

You got beaten to the punch by an old guy!

It's always the ninjas that get you, no matter what their age.

 

[0ldman79]

Hahaha!

I'll take it.

lol