PDA

View Full Version : the bh-5/bh-6 legend



saaya
10-15-2004, 12:15 PM
What is BH5?

bh5 is known as THE overclockers memory around the world. it ran at 200mhz+ speeds with 2-2-2 timings WITHOUT a voltage bump even though it was rated to 166-200mhz 2.5-3-3.

chronically it all started with the winbond ah-6 wich already oced to 180mhz with 2-2-2 in the early DDR days with only pc2100 and pc2400 (100/133mhz) memory around.
later winbond released the bh-6, an improved ah-6 made in a smaller manufacturing process. those bh-6 chips are identical to bh5, bh5 are made in a yet smaller production process though, wich made them cheaper to manufacture and reduced the vdimm needed to reach certain speeds.

both, bh5 and bh6 oc to amazing speeds of 300mhz and more with very high vdimm voltage (4v+). the newest and last winbond ic is the ch-5, its a "crippled" bh5. the design was simplified to reduce the production costs yet again, unfortunatly it reduced the performence of those chips a bit, even though the newest ch5 chips seem to be able to run at 2-2-2 as well, their performence is still lower than bh5 or bh6 running at 2-2-2 timings.

i actually ran a few benchmarks and couldnt see any improvement at all between 2-3-2 and 2-2-2 with my stick of ch5. anyways, not all of those chips oc to amazing speeds around 300mhz. most max out at 240mhz (with around 3.2v) and some even max out a lot earlier (i had a stick that wouldnt even run 190mhz! with 3.2v).

winbond originally planned to introduce bh-4 memory, an improved ch-5 memory that could run 2-2-2 and reach much higher speeds as it was also built in the smaller 130nm process, unfortunately winbonds managers decided that the memory ic business was not lucrative enough with its massive fluctuations and samsung and hynix throwing their ic's on the market for next to nothing. furthermore none of the memory manufacturers realized the potential of the bh series and its marvelous capability to run with 2-2-2 timings at very high speeds. so only a very few bh-4 samples were ever made.



here is a pic of one:
http://xtremesystems.org/forums/attachment.php?attachmentid=17760


there are some bh-5 chips that have a very odd stepping code that doesnt match the usual bh-5 stepping code and i suspect those to be the few bh-4 chips from the first production line. there was no interest for bh-4 at that time so winbond could have relabeled their small bh-4 series to bh-5 to be able to sell them. it was the first series of bh-4, so that explains why they dont overclock extraordinarily compared to the normal bh-5 memory.

almost at the same time as overclockers and tweakers discovered bh-5 as beeing the best overclocking and tweaking memory by far, winbond had unfortunately already decided to leave the DRAM market. the request for bh-5, wich was used on cheap value ram by that time (!) was growing incredibly high. the bh-5 prices were rising by 20$ per stick for some days.

all manufacturers tried to get some of the very limited stock of bh-5 that was still available. mushkin and ocz released a last "limited edition" of memory based on bh5. as the demand for bh-5 memory kept rising and rising they kept looking for bh-5 stocks and found some bh-6 supply instead wich they used for the final bh series memory, another very limited "special edition" from ocz and muskin was released.

even today, almost 1 year after winbond dicontinued the bh-5 series, there is still no memory that matches its performence and overclockability. ocz has had a mayor success with its EB series wich runs at 2.5-2-2 or 3-2-2 yet still performs as well as bh-5 and bh-6 with 2-2-2 timings at the same speed, in some situations its even faster! this is possible due to some latency tweaking and improved pcb design from ocz.

however, the EB series overclocked slightly worse than bh-5 and bh-6 and needed 3-2-2 timings to reach high overclocks of 280-300mhz (on some good sticks) wich made it a little slow than bh-5 and bh-6 wich also reached 280-300 (on some good sticks).

in the meantime samsung has released a new memory ic, called tccd. this allround talent can run with 2-2-2 timings up to 230mhz, 2.5-3-3 timings up to 300mhz (on good sticks) and up to 340mhz with 3-4-4-8 timings (on good sticks). so it beats the bh-5/bh-6 series with brute force, a huge clockspeed, but still lacks the low latencies at high clockspeeds like the bh-5/bh-6 series that reaches 250-300mhz with 2-2-2 timings on average (!) on athlon64 boards with very high vdimm.

thats the history of the winbond bh-5/bh-6 legend... :D

winbond (and all ic manufacturers) are always looking to reduce the production costs and improve the yields to make more money, so they keep changing details in the manufacturing process and even the ic design. those changes mean that some steppings oc better than others (wich is the reason people ask for cpu steppings as well) wich was finally the reason why i came up with the idear to collect the steppings to give people an idear what chips probably oc better than others.

i did some more research on the manufacturing process of the diferent chips, and found this:


AH-6 180nm
BH-6 166nm
CH-6 166nm
AH-5 (doesnt exist?)
BH-5 150nm
CH-5 130nm
BH-4 130nm

this explains why bh-5 and bh-6 perform almost the same, they are made in almost the same manufacturing process.

i wrote this down before i and everybody else forgets all this infos about bh5 and bh6 and for those who are new to overclocking and dont know what all this bh5 and bh6 worshipping is all about :lol: furthermore bh5 and bh6 deserves to remembered for a long time :D

Arnar
10-15-2004, 04:56 PM
real nice article, very interesting reading ;)

ent
10-16-2004, 12:23 AM
Really good one, well done saaya.

One question:
Was Corsair XMS 3000C2 rev. 1.1 (rated 2-3-3-6) BH-6?? (sorry but I have to know it before tomorrow :-))

Elisha
10-16-2004, 05:20 AM
well done.

saaya
10-16-2004, 05:35 AM
thx everybody :)

thougt id write down those infos before i and everybody forgets about it :D
bh-5 and bh-6 deserve to be remembered for a long time :D

sorry ent, cant help you with that. there was a thread called "where to find BH5 and BH6 on the internet" i think, its full with infos about what steppings and sticks have bh5 and bh6 :)

ent
10-16-2004, 06:50 AM
thx everybody :)

thougt id write down those infos before i and everybody forgets about it :D
bh-5 and bh-6 deserve to be remembered for a long time :D

sorry ent, cant help you with that. there was a thread called "where to find BH5 and BH6 on the internet" i think, its full with infos about what steppings and sticks have bh5 and bh6 :)
You mean this thread: http://www.xtremesystems.org/forums/showthread.php?t=27888

I read it, but i haven't found info about 3000 C2 v 1.1. In this list (http://pub.lorenz.bei.t-online.de/winbondlist.htm) there is no 3000 C2 v 1.1 also :( 3200C2 and 3500C2 were BH, so maybe also 3000...

eva2000
10-16-2004, 07:02 AM
XMS3200C2v1.1 = BH-6
XMS3500C2v1.1 = BH-5

ent
10-16-2004, 11:04 PM
XMS3200C2v1.1 = BH-6
XMS3500C2v1.1 = BH-5
:D That's nice :D but how about 3000C2 :confused:

GazC
10-17-2004, 01:02 PM
Saaya,

I find it interesting that you say the CH6 is based on the same manufacturing process as BH6. I was always under the impression that CH6 was "down binned" CH5. I would have thought that considering CH5 and 6 are both C revision, they would have been fabricated by the same process. I also thought the same was true of BH5 and 6.

centaur
10-19-2004, 06:26 AM
:D That's nice :D but how about 3000C2 :confused:

XMS-PC3000C2 Rev1.1 (2-3-3-6-1T) Samsung TCB3/Micron (6ns rev B) and few Winbond BH-6

found here: http://mar.h0sted.org/ramliste250804.htm

btw: i got 2 mushkin pc3200 512mb sticks today. one is a bh-5 and the other a bh-6 :D

ent
10-20-2004, 03:20 AM
XMS-PC3000C2 Rev1.1 (2-3-3-6-1T) Samsung TCB3/Micron (6ns rev B) and few Winbond BH-6

found here: http://mar.h0sted.org/ramliste250804.htm

btw: i got 2 mushkin pc3200 512mb sticks today. one is a bh-5 and the other a bh-6 :D
THX Centaur :D I didn't buy them cause I haven't known if they are BH-6 for sure :D

GazC
10-21-2004, 02:43 PM
Saaya,

I find it interesting that you say the CH6 is based on the same manufacturing process as BH6. I was always under the impression that CH6 was "down binned" CH5. I would have thought that considering CH5 and 6 are both C revision, they would have been fabricated by the same process. I also thought the same was true of BH5 and 6.
...

WiCKeD
10-24-2004, 11:02 AM
Excellent article, saaya! For the record I have a DH-6 chip on the bottom of my Seagate 15k.3 hard drive... Going by the size, it looks like it must be a 180nm chip, at least.

Noktar
10-24-2004, 10:00 PM
Very nice article.
Lot's of facts I havn't known.

Btw. the Corsair Rev. 1.1, can it be only BH5 or BH6 too ?

Coz friend of mine has a Rev. 1.1 Corsair, but he won't remove the HS.

I think it's BH5 but u know better I think !

Learn
10-25-2004, 12:15 PM
nice read saaya :thumbsup:

Noktar, go here => http://www.houseofhelp.com/v2/showthread.php?t=20550

blinky
10-28-2004, 12:15 AM
THX Centaur :D I didn't buy them cause I haven't known if they are BH-6 for sure :Dif this helps you at all, XMS pc2700LL is all BH6 im pretty sure

WeStSiDePLaYa
10-28-2004, 01:13 PM
http://pub.lorenz.bei.t-online.de/winbondlist.htm

lalPOOO
10-28-2004, 05:01 PM
BH-4 Never made it past prototype stage I think..?

STEvil
10-28-2004, 05:51 PM
There are odd BH-6 chips floating about on the bottom of Maxtor 7200rpm 40gb hdd's too............. ;)

mrnuke
10-28-2004, 07:58 PM
what is the default voltage for bh-5? is it 2.5v at 2,2,2,5?

WeStSiDePLaYa
10-28-2004, 09:10 PM
There are odd BH-6 chips floating about on the bottom of Maxtor 7200rpm 40gb hdd's too............. ;)


interesting thats the same drive i have.

nope just checked, i have crappy hynix!

http://www.comexgroup.com/storage/_memory/img_adata/adata500_chip.jpg

STEvil
10-28-2004, 09:39 PM
yeah, I think the chips change from drive to drive... tictac has a seagate drive or somethign with a bh-6 chip on it too.

Revv23
11-25-2004, 05:15 PM
does winbond know how much the demand for their chips has raised since they left the market?

if they would ahve fired up the machines again over the summer they would have been sitting on a gold mine...


another cool thing is how chepa BH-5 was, lets not forget, i got my first stick of 512mb Bh-6 for $95, and later when BH-5 was all but gone i picked up neweggs last 1Gb pack of KHX3000 (relabled 2700 that took forever to sell, but had either bh5/bh-6 sticks on it) for just $220, and thats when demand for these modules were high!

yay i just picked up another 512 of mushy lvII black (2x256) looks like mushkin saved even more stock hehe :)

http://www.xtremesystems.org/forums/showthread.php?p=610400#post610400

HermS
11-27-2004, 01:25 PM
Great article this, definitely a 'must-read' :cool:

Oddly enough I noticed that the Maxtor 80Gb SATA drive I bought yesterday has a DH-6 chip on it which was a littel surprising! :eek:

Revv23
12-03-2004, 03:36 PM
ok so i got my mushkin and i was hoping for single sided diims as im Overclocking my NFII board, and it comes and theres 4 IC's on each side :(


was this common with mushkin?

Revv23
02-09-2005, 05:04 PM
not sure if anyone has said this, but your info on the manufacturing process is incorrect sayaa


No need to argue, just read up on their respective datasheets from Winbond.

From the Winbond 942516CH datasheet (for all speed grades from CH-75 to CH-5), the manufacturing process is described as 0.13um. It also says the 942516CH is sorted into 4 speed grades (-75,-7,-6,-5).

I would be very surprised if Winbond has lines that produce ram using 0.18um, 0.166um, 0.15um & 0.13um in an overlapped timeframe. For starters, it means massive investment wasted, as 0.166 & 0.15 are not necessary as Winbond is already migrating to 0.13 at the time when the BH series was in production. The incremental benefits offered by 0.166 & 0.15 over 0.18 does not justify the investment needed. Secondly, why would Winbond waste production cost in producing chips in a larger process when a smaller & more cost effective process is available (not to mention the cost in inventory controls)? Thirdly, what happens to those CH-5 that cannot function at 200MHz but can do so at 166MHz?? Do they junk them altogether?? They surely can't be called CH-6 if it's from a different process!!!

The words AH, BH, CH tells u which generaation it belongs, & the -75, -7, -6, -5 tells u it's speed. If u look at the ICs themselves, it's printed "BH-5" or "CH-5", BH5 or CH5 is just an abbreviation.

http://www.winbond.com.tw/c-winbondhtm/partner/b_2_g_1_a.htm#06

Datasheet for CH series available at bottom of the page.

owies
03-12-2005, 07:17 AM
Have also DH6 cahe on my 2 120GB Maxtors:
http://members.lycos.co.uk/owiespl/Hdd/Maxtor/cache/dh6%20uk%b3ad.jpg

And BH6 on 40GB again Maxtor:
http://members.lycos.co.uk/owiespl/Hdd/Maxtor/cache/Maxtor%20cache%20on%20BH6%20+%20loopa%20paruje.JPG
Anybody knows how to oc it ? On 40Gigs i find a pin who i gues is a clock input, but have no idea how to give them more freg without spec. chip. :/
pic of 35pin of BH6 cahe (http://members.lycos.co.uk/owiespl/Hdd/Maxtor/cache/rezystor%20zoom.JPG)

Revv23
03-12-2005, 09:05 AM
lol are you kidding?

why would you want to overclock your HDD cache?

just imagine what an unstable IC would cause, and im sure it wont improve performance.

HARDCORECLOCKER
03-12-2005, 02:06 PM
:D Very nice review Sascha - made me just feel a bit sad remembering all the nice time with BH-5.

But I still use it and who'd put a cent on it that also on NF3 / NF4 it's an interesting opportunity to TCCD.

It take my breath away the day the message came on the ticker that Winbond stopped the production.

So all I can say is I wish everyone remebers this legendary RAM forever and I hope it will also be used in future systems.

:toast:

owies
03-18-2005, 08:22 AM
no, realy, it`s simple, I just can`t stay with non oced BH, :) And it may be a lot fun witch oc hdd cache. I know, performance, it`s most depend on hdd engine but i`m just curious ;)
Btw, of coz, art is great :]

owies
03-24-2005, 09:43 AM
Ok, it`s time to ask about stupid thing:
I`m just wondering what means that marks: W942508BH-5
can`t find nothing usefull on web, only some trash...

largon
09-11-2005, 01:03 PM
Thread revival!! :)

This thread is linked from saaya's Sticky-thread "bh5 steppping database (http://www.xtremesystems.org/forums/showthread.php?t=35598)".
I thought it would be good to update it a bit:

Winbond DDR-SDRAM ICs:

W942508AH = 175nm * datasheet (http://www.winbond.com/s-winbondhtm/partner/PDFresult.asp?Pname=743)
W942508AH-6 (http://www.techpowerup.com/memdb/img/chip10.jpg) = 175nm *
W942508AH-7 = 175nm *

W942508BH = 175nm ^ datasheet (http://www.winbond.com/s-winbondhtm/partner/PDFresult.asp?Pname=841)
W942508BH-4 (http://www.comexgroup.com/storage/_memory/img_adata/adata500_chip.jpg) = 175nm * (discontinued @ sample stage)
W942508BH-5 (http://www.techpowerup.com/memdb/img/chip1.jpg) = 175nm *
W942508BH-6 (http://www.techpowerup.com/memdb/img/chip4.jpg) = 175nm * datasheet (http://www.winbond.com/s-winbondhtm/partner/PDFresult.asp?Pname=795) (preliminary)

W942508CH = 130nm ^
W942508CH-5 (http://www.techpowerup.com/memdb/img/chip2.jpg) = 130nm * datasheet (http://www.winbond.com/s-winbondhtm/partner/PDFresult.asp?Pname=892)
W942508CH-6 (http://www.techpowerup.com/memdb/img/chip5.jpg) = 130nm *

W942508DH = ???nm (in development (http://xtremesystems.org/forums/showthread.php?p=1030024#post1030024), datasheet n/a)

* = DISCONTINUED
^ = UnTesTed-version available
note: AH- and BH-dies are of the same process size (see datasheets). This
leads to a conclusion: prefixes such as B in BH-5 or C in CH-5 do
not code process sizes, but are likely to discriminate chip generations.
Just like E and F (http://www.samsung.com/Products/Semiconductor/common/product_list.aspx?family_cd=DRM030103) are generation prefixes for Samsung K4H560838_-TCCD.

Winbond DRAM-IC markings: (incomplete)

WAABBCCDE-F

W - Winbond product

AA - IC Description
94 = DDR-SDRAM
98 = SDRAM

BB - DRAM density
25 = 256Mbit (32 megawords)
12 = 128Mbit (16 megawords)
64 = 64Mbit (8 megawords)
16 = 16Mbit (2 megawords)

CC - DRAM configuration
04 = 4 bits/word
08 = 8 bits/word
16/G6 = 16 bits/word
32/G2 = 32 bits/word

D - IC generation (A/B/C/D/E) does not refer to process size

E - Package type
H = TSOP II 50/54/66/86
B = VFBGA 54/60/90

-F - Speed grade in nanoseconds (-4/-5/-6/-7/-7.5)

largon
09-25-2005, 03:07 AM
It seem the 1st post has some critical factual errors:



What is BH5?
--
later winbond released the bh-6, an improved ah-6 made in a smaller manufacturing process. those bh-6 chips are identical to bh5, bh5 are made in a yet smaller production process though--W942508BH and W942508AH are of same process size.

both, bh5 and bh6 oc to amazing speeds of 300mhz and more with very high vdimm voltage (4v+).-- Screenshots of 300MHz 2-2-2-x do exist.
But more than 300MHz? No.

winbond originally planned to introduce bh-4 memory, an improved ch-5 memory that could run 2-2-2 and reach much higher speeds as it was also built in the smaller 130nm process --BH-4 is only a higher bin BH-5. Nothing new actually.

i did some more research on the manufacturing process of the diferent chips, and found this:


AH-6 180nm
BH-6 166nm
CH-6 166nm
AH-5 (doesnt exist?)
BH-5 150nm
CH-5 130nm
BH-4 130nm

this explains why bh-5 and bh-6 perform almost the same, they are made in almost the same manufacturing process.It seems only 1 (one) of those number is a fact...

Saaya, may I ask what was your source? :)

ZX7891
09-25-2005, 05:41 AM
UTT manufacturing process is UTT-CH and BH built on?

eliminate
09-25-2005, 05:58 AM
I have the OCZ Gold Series RAM which is supposedly BH-5.

Would my RAM be able to reach 300 at 2-2-2 or do I have to relax my timings a little with these. The Mushkin RedLine RAM uses the exact same chips as the OCZ Gold.

largon
09-25-2005, 06:59 AM
ZX7891,
There's 2 kinds of UTT:
1. CH-UTT = untested Winbond CH-die
2. BH-UTT = untested Winbond BH-die


eliminate,
Mushkin Redline is CH-UTT. Gold VX on the other hand is BH-UTT, like you said.

Here's a good way to identify Winbond chips:
If your sticks can't do tRFC 12 stable, they aren't BH-die.