*bump*
Duonger is doing the best he can to help me out on this one. Infineon is only interested in selling big quantities, so big that I can never do all the testing myself.
This is all I can say now ;)
best regards,
alexio
Printable View
*bump*
Duonger is doing the best he can to help me out on this one. Infineon is only interested in selling big quantities, so big that I can never do all the testing myself.
This is all I can say now ;)
best regards,
alexio
Kudos for even trying to tackle this - hope something comes out of this for you, even if its not what you intensionaly set out for.
Keep looking at ALL options and from all angles.
If you want make a real certification, so you refund ... if the module can't make the frequency.
For me it's not possible to sell a module with certification of more than 260 MHz 2-2-2-5 1T @ 3.6V. It's not a 256 Mb of old BH-5 in Single Side.
It's 512 Mb Dual Side module with UTT BH-5 0.175µm.
Maybe you can sell few special module @ 270 Mhz with 3.7V, but required a ACTIVE COOLING and mobo as DFI nF4 ... and you will have 1 module of 200 who can make it.
Also, don't confuse UTT CH-5 0.13µm chip (Cheap), and UTT BH-5 0.175 µm chip (Expensive).
You want sell 1Gb DDR 400 2-2-2-5 @ 3.0V for 100$ ... but with 100$ ou can't buy 32 chips.
Don't speak to fast, please analyse all before :)
We speak about 512 Mb module and STABLE (CERTIFIED), fatory 1000 modules of better TCCD CHIP (437 or 440), on a BP 808 PCB.Quote:
Originally Posted by j_p_h2002
Take the 100 better modules, and maybe you can certified 310 MHz 2.5-4-3-5, not more. Also, we speak about UTT BH-5, not TCCD ;)
I know, on the DDR400 it's impossible to make a profit. I expect that 90% of the chips can run DDR480 and that would be a nice speed to start testing at (with some yields for overclocking). Anything that fails will be sold as DDR400 2-2-2-5, but ofcourse tested again to pick the really bad or broken chips out.Quote:
Originally Posted by computersmsa
I know about prices now. I can't say much about testing atm. I can say that if I continue with this pricing will possibly be higher than stated before, that was ofcourse an indication. Final pricing will depent on the yields of the chips.
Big quantities will probably be sold by mushkin but I cannot say anything about that now. Duonger is looking into the possibility of testing the chips the way I proposed.
best regards,
alexio
just a second, you mean 90% of ic's completely untested up to that point?Quote:
Originally Posted by alexio
You expect 90 % of UTT BH-5 can 290 MHz 2-2-2-5 stable ?
I'm sorry, finish dream => 0.0000000 %
I speak about 512 Mb modules DS ! Maybe 256 Mb SS you can have 5% at 280 MHz 2-2-2-5.
Edit : Sorry, you tell 240 MHz, -> Yes that's right :)
exactly my point.Quote:
Originally Posted by computersmsa
afaik, at least 20% of the ic's will end up being just garbage.
No, I can't say more though about testing at this point. Testing each chip at their rated speed, timings and voltage first I guess. Any sales from Infineon and other components for the PCB will be via Mushkin. No deals yet between Mushkin and me. No deals at all. I just heard something about pricing.Quote:
Originally Posted by high5
I've said all I can say now. Not because I don't wan't to share information with you. It's just that I don't have much solid information now. I can't give information about pricing of the chips.
best regards,
alexio
Not all people get the point that on a 512MB DS module are 16 chip's, so if 1 chip can't do 2-2-2-5 the whole module can't work at that timings, the other chips can maybe do 280MHz @ 2-2-2-5 or better, but because 1 chip can't work @ 2-2-2-5 timings the module is a bad overclocker.
And how said it is 90% of the UTT that can do 290MHz 2-2-2-5, ofcourse there are a lot of chips that are ****** up, but that doesn't mean a lot of chip's are good.
If you can sell 1Gb Kit certified DDR560 2-2-2-5 @ 3.7V @ 300$ (Not DDR600), in one mouth you have sell 1000 Kits and you can buy one Ferrari 575 Maranello :D
DDR480 is 240mhz my friend, guess you should know :stick:Quote:
Originally Posted by computersmsa
I can't tell you more about testing now. I expect a reasonable percentage to be able to run DDR560.
regards,
alexio
Yes, I have edited my post before :D
Sorry :D You have right :)
"I expect a reasonable percentage to be able to run DDR560"
If I have understand, you want test chip per chip and put 16 powerfull chip per modules, Right ? You will have 16x luck in more that' right ... but you have a hardware for test chip per chip ?
For me, if you factory BH-5 UTT on BP808 (512Mb DS), you have 0.1% chance to can sell 280 MHz 2-2-2-5 certified module.
Test chip per chip, maybe 2% not more ... :(
But test chip per chip is a very good idea (VERY EXPENSIVE) !
Also, dont forget that certifier 280 MHz 2-2-2-5 @ 3.7V => You must have a module right for 282-283 MHz 2-2-2-5 @ 3.7V (not 3.78V) and not a incredible FAN, light FAN.
So, you have good idea but it's very hard (I don't speak about DDR600, just DDR560).
errrm, without ALL the chips on the pcb you will have no idea if the chips can run at the same speed with all the chips on the pcb. if you are testing 8/16 chips at a time, how would you know which was the bad chip??
basically, what this is turning into is, mushkin will buy some memory chips from a manufacturer cheap cos they are untested, then, they will test them and release speed binned chips? OMG, thats amazing, how did no one do this before?
Also, isn't just a touch unlikely, that on all the bh-5 512 sticks that the people on this forum have had over the last 3 years(we must be talking thousands of sticks), each and every single stick had one bad chip that caused it to not hit 300Mhz. quite frankly, bollocks. they can't do that speed, simply as that, there is no way in hell you can release a product. 2 times in 3 years we've seen that speed, unstable, with HUGE voltage and on 256 single sticks. those were the only two sticks, in 3 years, with bh-5 that had only good chips on the pcb? ok, lets just for a second assume each and every one of those sticks over 3 years that have been used, does have 1 faulty chip at least, then simply put you will not find enough chips to make sticks.
prediction, mushkin will see that some people are willing to throw even more voltage at to get slightly higher than their redline speeds. so they will now get some bh-5, rate it to 3.8v, make it stupidly expensive to cover all the rma's, and sell some newer redline thats garenteed to 270Mhz.
i've got 4 sticks of twinny bh-5 that all do 260Mhz, 250 stable, 260 a little unstable, but not used more than 3.2v yet. my previous sticks of ch-5 utt did about the same little more voltage. my old bh-5 would all do 260Mhz, my 256 sticks would do a little more.
i for one wouldn't mind seeing a little more reality on these forums. maybe i'm jsut getting old.
Quote:
Originally Posted by drunkenmaster
The only problem with your logic:
Mushkin sells tons and tons of black level II TCCD rated 2-2-2-5. As i'm sure they will sell tons of redline that is pc4000.
Just because thats all anyone sells doesn't mean the chips cannot do it. they only sell taht speed, because that is whats ecinomically FEASABLE to make a good profit.
He is talking about testing each IC individually I believe, and the ones that do well, are to be combined onto one PCB and tested. For a ram manufacturer like mushkin, doing that for every stick would be nuts if they wanted 1000-2000 sticks like that.
Now, maybe 10-20 sticks may be more managable by a single person. :stick:
Computersmsa does this already *somewhat*, as i saw in his madshrimps thread. not each IC, but the sticks.
the reason we dont see ddr600 2-2-2-5 on OCZ and mushkin stuff, is because, if 1 module does 265mhz, and one does 305mhz, the 265mhz will still pass as the pc4000, making that 305chip unused.
its all very very dependent on where you place the performance bar, and which stages of testing its at a certain point. :fact:
Couldn't you have some automated machine that has a robot arm, which picks up a chip from a belt, put in into one of a set of testing slots, while testing runs it picks up the one that just finished testings (FILO buffer), and moves it to a set of bins/belts based on clockspeed. Then each bin is used to make the RAM sticks. You should be able to do 1000+ chips per day like this. I think that's enough production capacity.
Quantities are huge, just think about atleast 100k chips. Now guess what happens if even 1% is 300+mhz capable ;)
DDR600 might be closer than you think, atleast DDR560 is.
Lets say that 1% is capable of doing 300+mhz. Now a 256mb stick has 8 chips.
0.01*0.01*0.01*0.01*0.01*0.01*0.01*0.01 = ;)
To reward your boldness, I hope that 5% of the chips are capable of 300Mhz.Quote:
Originally Posted by alexio
But what makes you believe that the order of magnitude of 300Mhz capable chips is that and not one hundredth or one thousandth. I hope your sure of your beliefs, since your success of failure will depend on numbers like this one.
can the chips themselves be tested by themselves??? or do you have to mount them to the PCB and hope all of them work at a good speed??
if so I will pay for some 300-320 stable chips :D
if you can test the individually, taht will make this a much easier task. heck, bet you could sell some nice 256mb ddr600 sticks then.Quote:
Originally Posted by iboomalot
Quote:
Originally Posted by alexio
firstly, you're completely guessing that 1% can do 600Mhz stable. also as i said, there have been, i said thousands, but look how many members we have, and how many sticks each of us have had. though the chances are low, not one single stick of 512 mb has ever hit 600Mhz. anyway, assuming there are just a few chips in the 100k that do 600Mhz, do you know how long it will take you to find each and every one of those chips? finding enough for one stick of 512mb could take you a month. it will never happen. the bh-5 die, design, spec and transistors are simply not capable of those speeds. remember also, not only has ONLY 256 mb sticks done ddr600, there is a LOT less stress on the pcb alone, also the power lines, the traces on the pcb and mobo, than there is for a 512 stick, let alone two. and we have never seen a 256mb stick do it at 3.8v. we've seen 280 fairly often(still not that often) and a couple higher speeds. but to claim that there is even a remote chance of this is just, well, so far out there, that its funny.
ddr560 sticks should be easy. its really this simple, there are 10's of thousands of sticks out there with bh-5 on, we know EXACTLY what bh-5 can do. we know some bh-5 can do 280mhz. maybe 290, but thats rare. those are the sticks with only good chips on. by pure chance there are a couple that will have the best chips on and we've seen them already.
as i also said before how are you going to stably test each chip on its own. one chip running on a pcb is not the same as running 8 chips or 16 chips at full load. everything changes. it might well be possible that 1 chip on its own can do 320Mhz, but with 8 chips in they will never break 260Mhz. testing individually, if even possible, is very unlikely to prove possible (maybe not at all or simply, 100k chips, even by mushkin let alone by one guy, will take so long that there probably won't actually be ddr boards being used by enthusiasts by the time you make the first stick.
maybe i'm being harsh, but honestly, i'm trying to stop you wasting too much time. try and think realistically, how long it would take to test a single chip, ie memtest stable, then test each chip that seems to run better on a pcb with more of the binned chips. then every time a stick of 8 chips won't do ddr600 you test for ddr560, then ddr450, then ddr400.
I think you completely missed my post entirely.Quote:
Originally Posted by drunkenmaster
Mushkin and OCZ arent "handpicking" these sticks everyone is buying. if a certain IC does 265mhz, and one does 350mhz, they both pass as pc4000, the highest speed grade, and can end up on the same stick. :fact:
Using a DIFFERENT, EXPENSIVE, TIME CONSUMING method, such as hand testing each IC, DDR600 should be easily obtained- but it sint going to be cheap.
Quote:
Originally Posted by drunkenmaster
Your saying that since bh5 that is tested to do 400mhz and passes is all slapped on a stick and it is extremely rare that even a 256mb stick can do 300mhz that it won't happen? There ARE chips that will do 300mhz, I bet they are 'fairly' common, as in you probably have at least one on every 2 512mb sticks, but they cannot be taken advantage of because you also have the chips that are holding you back.
Think about it, out of the 1000's of bh5 sticks we have had on this forum, and I'm only using 256mb for simplicity. I bet less than 10% have seen higher than 3.5v, so I'm sure the number of 300mhz sticks would be slightly higher, not to mention not everyone here knows as much as some other people who could take the same stick and get 20mhz higher by just tweaking the alpha timings and such. Now there are 8 chips per stick, and all of those chips were random, no high speed binning done. So I think if you take say 8,000 chips, that are tested for pc3200, or some 'even' pc3500, and just randomly put any of those chips together in sets of 8 (for a 256mb stick), and you actually have a couple that will do 300mhz, I think those are pretty good odds when you think about it.
Excellent point, and seconded.Quote:
Originally Posted by kryptobs2000
You guys are dreaming big time :( As a matter of fact, I don't understand why would this thread be started in a first place without any research done before. Alexio doesn't even know If he's able to get Winbond/Infineon to talk to him and sell him the ICs, less testing them. Then, claims of DDR600 6-2-2-2.0-1T rock stable... We have not seen DDR560 "rock stable" (neither on 256MB no 512Mb sticks). DDR560 can be achieved, lightly benchable... Rock stable DDR560 will be with some magic sticks, 1 in 1000 maybe. Speed binning ICs seperately @ DDR600 3.8v, one by one is just unreal.
I would rather have alexio come here with a "very cherry picked" sample of DDR600 "rock stable" sticks that speculating about it and letting others dream his dream that will most likely not come true :(