Merge ddr2 FB-DIMM category with DDR2

[yosarianilives]

As far as I can tell FB-DIMMs are just a fancy ddr2 dimm essentially. The base memory technology is still ddr2. It's basically the same as having a seperate category for sodimms or ecc sticks. These categories should be merged for a more cohesive DB. I think they only reason they ever got seperated was someone with a skulltrail system wanted to feel special and different than people with "pleb" non buffered ddr2 oc server boards.

[Mr.Scott]

I disagree.

If they were interchangeable it would be a different story.

 

Quote

 

Fully Buffered DIMMs (FB-DIMM) have an AMB (Advanced Memory Buffer) unlike other buffers this has a Serial Interface to the memory controller, this allows greater data width without having to increase the number of pins on the memory controller. This method is only used in DDR2 modules.

FB-DIMMS have 240 pins and are the same length as DDR DIMMS but differ by having notches in different positions so standard DDR2 modules will not fit into a FB-DIMM system.  FB-DIMMs require a FB-DIMM compatible system.

 

 

Edited July 31, 2019 by Mr.Scott

[mickulty]

I was gonna hold off on posting this to give him chance to delete the thread but since people have seen it now;

FFS, just because you want to sub 6-core socket 604 Xeons in team cup.  "for a more cohesive DB" my arse.  You've never even OC'd an FB-DIMM and you're saying they're the same?

Did you just get sick of trying to find hardware loopholes, so now you want to mess with the DB to make your own, while accusing others of "wanting to feel special"?

You know HWBOT staff are volunteers, right?  Don't waste their time with stupid crap like this.

BTW as far as the underlying logic, if anything it's weirder that DDR3 and DDR4 RDIMMs aren't listed separately since the CAD bus is done totally differently on them.

[ground]

31 minutes ago, mickulty said:

BTW as far as the underlying logic, if anything it's weirder that DDR3 and DDR4 RDIMMs aren't listed separately since the CAD bus is done totally differently on them.

I mean for those there are at least IMCs that can use udimms as well as rdimms without any issue, sometimes even on the same boards (x99 especially). Pretty sure that doesn't apply to any platforms with FBdimms or do I remember it wrong?

[mickulty]

2 minutes ago, ground1556 said:

I mean for those there are at least IMCs that can use udimms as well as rdimms without any issue, sometimes even on the same boards (x99 especially). Pretty sure that doesn't apply to any platforms with FBdimms or do I remember it wrong?

That would be correct, they have a ridiculously fast serial bus to the AMB chip that's something like 12x the ('true') memory clock.  This isn't shared, rather each DIMM passes on commands to the next.

[Noxinite]

Err, what even is this thread.

[ground]

Indiana Yos on the hunt for the loophole

[Bilko]

47 minutes ago, ground1556 said:

Indiana Yos on the hunt for the loophole

Hey I like those movies, don't you ruin them for me now....

[yosarianilives]

I may've been reminded of it by looking at team up chips (although it's really not needed to win that stage at all) but I've thought it weird that two different ddr2s are separated for what at most amounts to a difference of ddr2 oc on p45 vs x48 but because the controller is on dimm instead of on mobo everyone acts like it's completely different. 

[yosarianilives]

And tbh it'd be silly for team cup as it would enable only about 6 chips that might have an advantage over s775 c2q in one bench, which if you know about the bench on this platform you know the memory layout precludes it actually being competitive 

[Mr.Scott]

11 minutes ago, yosarianilives said:

but I've thought it weird that two different ddr2s are separated for what at most amounts to a difference of ddr2 oc on p45 vs x48 but because the controller is on dimm instead of on mobo everyone acts like it's completely different. 

Slots not being the same make it completely different. No rocket science involved here.

[yosarianilives]

Just now, Mr.Scott said:

Slots not being the same make it completely different. No rocket science involved here.

So we should have a separate category for sodimms, microdimm, minidimm? Also let's not forget soldered straight onto mobo or the billion and one different kinds of simms. Yeah that's basically a non argument considering that no other mem is divided by socket. 

[ZFeSS]

2 hours ago, yosarianilives said:

So we should have a separate category for sodimms, microdimm, minidimm?

If you can post FB-DIMMs in any other board we'll discuss it.

[yosarianilives]

13 minutes ago, ZFeSS said:

If you can post FB-DIMMs in any other board we'll discuss it.

Post ecc dimm on non ecc memory controller then we can talk about how adding a buffer chip to a dimm makes it a new type of memory 

[ZFeSS]

RTFM, please, before noob posting.

[mickulty]

8 hours ago, yosarianilives said:

the memory layout precludes it actually being competitive 

What's wrong with the memory, isn't it just DDR2? /s

[max1024]

The idea of combining FB-dimm with ordinary DDR2 is meaningless imho. Firstly, completely different preassignment and hardware component. For all frequency and speed characteristics, the FB-dimm will lose, no examples are needed here. This is the same as combining FPM and EDO based on the conclusion, that they are structurally the same and the number of contacts is identical.

I suggest that energy be directed in a different direction. Somehow, in search of one user, I stumbled upon a situation when I found more than a hundred registered overclockers from my own micro-country, with zero points and no results. It seems to me that there will be hundreds of thousands of such users who have been hanging for many years, and this is the place in the tables, and a waste of time in calculating their zero results. I propose to remove all those who registered more than a year ago and did not give a single result, I think this will benefit the HWBOT database.

Edited August 1, 2019 by max1024

[yosarianilives]

That's like saying that we should give am2 it's own ddr2 ranking because it will always lose to s775. The max attainable speed and frequency has never entered into how categories are set. As for FPM and EDO they are seperate memory standards and shall be, not really a relevant comparison when comparing ddr2 ram with or without an extra on dimm controller.

[Leeghoofd]

[max1024]

22 hours ago, yosarianilives said:

As for FPM and EDO they are seperate memory standards and shall be, not really a relevant comparison when comparing ddr2 ram with or without an extra on dimm controller.

And I consider FB-dimm to be a separate standard that does not overlap with DDR2. All previous years and the editors of HWBOT, this truth did not cause questions. How does it make sense to raise this issue if it has already been resolved before us? There are a lot of other issues that need to be resolved, at least, to keep me up, which will be of real benefit to all.

[skulstation]

As a fb-dimm user skulltrail and tyan dual 771 board.
Don't mess whit fb-dimm , let it have its own category

PS: was not intend to post in here first

[yosarianilives]

Well after reading again through this thread I've realized I might've been "a little" off the mark. Definitely very different than ddr2 on a logical side, which imho is what matters the most as logical standards can work on all kinds of physical standards.

 

So now forgive me if I completely understand how this memory works but could you achieve higher clocks if you replaced the buffer with an fpga that can attain higher clocks? Seems like something we may one day see if it's possible.