Page 1 of 1

RAM identification and speeds

Posted: Thu Aug 31, 2006 9:44 am
by His Royal Majesty King V
All,

Sorry I haven't been here much, but between a really funky job market, and now having a Royal Majesty Prince V, time's been a bit lacking!

Well, onto the questions:

I've got two SDRAM chips I can't quite identify. They're definitely older units.

I've been using Everest Home Edition v2.20, and if you go to Motherboard -> SPD, it'll list the RAM modules installed, and give you quite a bit of information, most importantly to my question Memory Timings.

I'm using a P2-266 as my test-mule, and have thus gleaned quite a bit of info about some old DIMMs that I have (including a 4-chip, single-sided 32MB piece that's brutally fast at 100MHz . . runs 2-2-2-5, which I didn't think such small modules were available at).

Anyway, there are two DIMMS I have that Everest can't identify. One of them is lacking that little SPD chip, though strangely, the other one has one, but Everest still simply won't list it at all.

So, what I'm wondering is, does anyone know, or is there any free software out there, that will tell me what memory timings these two DIMMs will run trouble-free at a given FSB?

The one without the SPD chip has NEC chips, 32MB total, 16 chips (8 on each side), with the following markings:

Code: Select all

NEC JAPAN
D4516821AG5
-A10-7JF
  9805AY007
I believe it's a PC100 unit, but that's all I've been able to determine, and I'm not even 100% sure on that.
There are also 2 stickers on it, each of which fits onto a single one of the chips, one of which reads:
EU32SDRAM808S
and the other:
72.24341.00N


The second DIMM I think is a PC66 unit, though I've come across references stating that some PC66 modules will perform better than some PC100's. This is the one that has an SPD chip on it, but Everest acts as if it doesn't. It's also a 16-chip unit (8 chips on each side), and 32MB. Anyway, the chips on this one read:

Code: Select all

 KOREA C206
 9819 FFF0
HM5216805TT10H
There are also some stickers, one of which fits on one memory chip:
290069
010K
S-DR LD-32-HT
Void if Removed

and the second of which spans 4 of the chips and reads:
RM-DM-SDRAM-32M-S-DRAM LD-32HT
(a UPC bar code, with the following numbers under it: 0 75028 00000 4)
090198-UME-460390

Also printed on the corner directly on the module is:
POWMEM
ASSY IN USA


I've done quite a bit of searching on Google, but it's been mostly fruitless, other than that the former has NEC chips (duh) and is PC100, the latter has Hitachi chips (most likely), and is probably PC66.

Anyone have access to datasheets, or know where I can get software that will definitely identify the SPD-equipped one? I tried searching Hitachi's and NEC's websites to no avail. My goal is really to find out what the ideal timings are at a given FSB.

Posted: Thu Aug 31, 2006 10:09 am
by Uranium235
You could try Sisoft Sandra, sometimes that works when Everest won't.

There's also DocMemory SPD Reader 1.4
http://www.majorgeeks.com/DocMemory_SPD ... d4383.html

:)

Posted: Thu Aug 31, 2006 10:12 am
by cake
Does this tool work?
http://www.heise.de/ct/ftp/ctspd.shtml

Was a while since I used it myself..

And congratulations on Prince V!! :D

Posted: Thu Aug 31, 2006 3:38 pm
by His Royal Majesty King V
U235: The DocMemory one doesn't seem to work, just hangs. The information at the website does seem to imply that it only works on Intel and SiS chipsets with DDR. My stuff's Intel, but ancient SDR equipment.

EDIT: Sandra works, but doesn't give a whole lot of info. I had to use an older version that would work with Win98se. Just gives CL values.

Cake: The ctspd program worked after a manner. It gives me a LOT of information, though some of it seems to contradict what Everest says. It did however also give me information about that one Hitachi-chipped module, but I don't know how to directly correspond it to what Everest says.

For example, on the 32MB module that Everest *does* recognize, I get the following info:

Code: Select all

Memory Timings:
   @100MHz:   3.0-3-3-5 (CL-RCD-RP-RAS)
   @ 76MHz:   2.0-2-2-4 (CL-RCD-RP-RAS)
However, for that same module, ctspd says:

Code: Select all

                              SPD-EEPROM      66MHz      100MHz
CAS Latency                      2/3            2           3
_RAS to CAS delay tRCD          26 ns           2           3
RAS precharge tRP               26 ns           2           3
RAS cycle time tRC                              8           8
RAS active time tRAS            50 ns           5           5
Ok, so I get which labels correspond to each other between the two, but why do they disagree, specifically with RAS/tRAS?

Here's what the 64MB module says:
Everest:

Code: Select all

Memory Timings:
   @100MHz:   3.0-3-3-5 (CL-RCD-RP-RAS)
   @111MHz:   2.0-3-3-6 (CL-RCD-RP-RAS)
Huh? CL is faster at a higher FSB?!

EDIT: Sandra confirms it as well: CL3 at up to 100MHz, CL2 at up to 111MHz... weird.

ctspd:

Code: Select all

                              SPD-EEPROM      66MHz      100MHz
CAS Latency                      2/3            2
_RAS to CAS delay tRCD          26 ns           2
RAS precharge tRP               26 ns           2
RAS cycle time tRC                              8
RAS active time tRAS            50 ns           5
Apparently this thing's SPD specifically specs it at 66MHz, though it appears to be equal in speed (if not better, if Everest is to be believed) to the previous one.


And here's the one that was an unknown for Everest, so I've only given the ctspd values:

Code: Select all

                              SPD-EEPROM      66MHz      100MHz
CAS Latency                     1/2/3           1
_RAS to CAS delay tRCD          30 ns           2
RAS precharge tRP               30 ns           2
RAS cycle time tRC                              8
RAS active time tRAS            50 ns           5
Again, ctspd seems to think it's "officially" 66MHz only (as shown by the box labeled "PC100 compatible"), but what's with that funky extra-fast CAS Latency, particularly given that RP and RC are a hair slower?

I think I'm more confused than when I started out.

Anyone have any thoughts or insights?

Posted: Thu Aug 31, 2006 10:40 pm
by Jim
Unless you are dead set on using them, I can snail mail you a matched pair of single sided, (8 chips), Infineon 128s I have sitting around that I will never use. I think they are PC100, though they might be 133.

Posted: Sat Sep 02, 2006 7:39 am
by tazwegion
It's been a little while since I tested any RAM modules, but Memtest86+ provides information r/t motherboard chipset, processor & memory etc. worth a look in perhaps?

Additionally... I have a matched pair of AOpen 64Mb SDram sticks (PC100) which ran flawlessly @ 133Mhz in an old Soltek Skt. A I used to have ;)

Posted: Tue Sep 05, 2006 5:11 pm
by Super_Relay
CPUz will also read the SPD data from modules.

I use it at work all the time to figure out what ram i need to upgrade a machine without having to take the case off to look

Posted: Wed Sep 06, 2006 5:57 am
by KachiWachi
Just some random thoughts -

SPD modules can be wrong. If the manufacturer programs incorrect data into it, the system would never know it.

No SPD/damaged SPD, no readout in programs that read the SPD. :(

NEC JAPAN
D4516821AG5-A10-7JF
9805AY007
EU32SDRAM808S

D45 = NEC SDRAM
16 =16
8 = x8
2 = 2 bank
1 = LVTTL
A = Revision
G5 = Package (44 pin TSOP II)
-A10 = -10 = 10nS = 100MHz
-7JF
9805AY007 = Date Code (Week 5, 1998)
EU32SDRAM808S = 32MB SDRAM

Datasheet -> http://www.elpida.com/eolpdfs/E0122N10_EOL.pdf

KOREA C206
9819 FFF0
HM5216805TT10H
290069 010K S-DR LD-32-HT
RM-DM-SDRAM-32M-S-DRAM LD-32HT

HM5 = Hitachi
2
16 = 16
8 = x8
05TT
10H = 10 = 10 ns = 100MHz
9819 = Date Code (Week 19, 1998)
Other = 32MB SDRAM

Datasheet -> http://www.elpida.com/eolpdfs/E0168H10_EOL.pdf

Posted: Wed Sep 06, 2006 5:56 pm
by DonPedro
@kachiwachi

of course the spd-data might be incomplete or false, but the program ctspd (cake gave the link to it some posts above) detects faulty information and displays whether the data is complete/incomplete or if the checksum of the data gives an error.

Posted: Thu Sep 07, 2006 4:55 am
by KachiWachi
Must be a damn sophisticated program then (for 400 KB)...if it actually runs memory tests and can adjust timing parameters on the fly (to find false/incomplete/faulty information). :wink:

I'm sure all it does is dump the hex code from the SPD, then decodes the parameters...just as Everest (or any other program) would.

In any case...the posted datasheets will tell all.

Posted: Thu Sep 07, 2006 12:39 pm
by DonPedro
no, it does not run memory tests nor does it allow for on the fly changing of parameters.

what it does is it reads out the spd-parameters, checks for completeness and checks whether the data is logically "valid".

why not just downlaod it and give it a try?