Widespread Voltage Issues on DLT3C

I am on my fourth DLT3C 1700+ chip now, and I've found some interesting results. First, and I don't remember who posted the link, but the DLT3C TBredB chips allow for more amperage through the chip, allowing the voltage to run lower to achieve the same level of stability at a given speed. This is a good thing until you start hitting an upper voltage wall that is much lower than other chips. Out of the four chips, here are the voltage points at which the chip will refuse to post at ANY speed:

#1: 1.65V
#2: 1.775V
#3: 1.85V
#4: 1.75V

I have also noticed other threads in which people have stated that their DLT3C chips are refusing to go over 1.8V or other similar numbers, and there are other threads in which people are stating that they are using 2.05V or higher. The wide disparity in how these chips are handling overvolting from stock is disturbing to say the least. I'm not complaining for speed reasons: even the lowest voltage chip listed above still got to 2.3 Ghz, and Chip #2 got up to 2.55 Ghz at 1.75V, so I'm definitely not complaining about performance. What worries me, however, is the longevity of these chips at higher voltages. Since some are proving that they can't handle overvolting even in small amounts, it makes me wonder if there could be a problem in the design of the cores themselves that cause problems to crop up at the higher amperage ratings that AMD is using, and if these problems were detected and that THIS is the reason that the chips were devalued from higher speeds down to 1700+ chips.

This may be something that everybody should watch, because if these chips are not able to handle the total W load through their internals, it may cause some of the chips that are run at higher voltage to suffer something very similar to SNDS in short order. While I haven't lost a chip yet, I am monitoring the ones I have installed very closely for any gradual buildup of instability. I would caution people against using higher voltages until we know more about this issue.

The problem is that I don't think we will know more about the ramifications of this problem until processors start eating it ... some of you don't care and are looking for an excuse to upgrade again. If that's the case, feel free to be guinea pigs, but for those that want their chips to last a bit longer, it may be wise to back off of them a bit.

While all chips aren't identical in their tolerances, this large of a disparity has got me worried. Yeah, maybe I'm overreacting, but I always feel that it's better safe than sorry.

I'm wondering if maybe burning in at the max posting voltage would let you up the voltage another step or two... or it might just hasten your chips demise.... who amoung you is brave eniough to find out. (The Gummy BearSSSSSSS! They're in my HEAD!!!!!! They're chewing on my BRAIIIIIIIIN!!!!!!!!!!!!!!!!)

Nice chip that hit 2.55 at 1.775 by the way. (licks lips) Mmmm Processing...

Could this be the begining of SADS (Sudden Athlon Death Syndrome)?

That's my worry. On the 1.75V chip I did a thorough burn-in as an experiment and left it burning for 72 hours. At the end it still wouldn't handle any more voltage and I got only 1 measly Mhz in FSB out of it. I haven't noticed any added instability yet, but then again it hasn't been pushed too much either. That particular chip lips just over 2300 Mhz at 1.725V.

Here is another case of voltage related reboots:

http://forum.oc-forums.com/vb/showthread.php?s=&postid=1701799#post1701799

One of the chips has now died. It lasted less than three weeks running 24/7 at 1.725V. Max temp recorded on this chip was 43C. I didn't really even push it much ... it was max stable at 2475 and I was running it at 2350. I am seriously thinking there is a potential problem with these chips.

The chip was a JIUHB 0308XPMW.

Note, this chip is going to be part of an RMA experiment, just like the last bad 1700+ I got was. I have the replacement coming in today. If it won't hit 1.8V, I will crap myself.

I never even heard of SADS. how common is it?

airspirit, edit your posts more. you got like 4 in a row. I did the same thing, i had 5in a row once. just try to keep it under 3or2 in a row. it dosent bother me that much, but i know others that are bothered by it.

modenaf1 said:

I never even heard of SADS. how common is it?

airspirit, edit your posts more. you got like 4 in a row. I did the same thing, i had 5in a row once. just try to keep it under 3or2 in a row. it dosent bother me that much, but i know others that are bothered by it.

Click to expand...

You have to have a good power supply for these chips as they take a lot of amperage to run. That's one of the keys for the 1.5V chip. Mine? I've never taken it over 1.7 Volts. I'm at 2.31gHz so far and I'm happy. 200mHz (.2gHz) either way is not really going to be noticeable in the real world, so why bother frying the chip?

I'm running them on a 480W Antec TruePower and the rest are on 500W generic PSUs. I think they are getting enough juice. That's not the issue: the issue is the chip itself. Curiously enough, I just got another one from Excaliber yesterday that will run up to 1.9V on the same equipment. This chip is an absolute wonder, giving me stable speeds of 2650 Mhz watercooled at 1.775V, but I'm keeping it at 2500 Mhz so I can keep the voltage lower. I'm paranoid about cooking more of these things.

sanford1 said:

You have to have a good power supply for these chips as they take a lot of amperage to run. That's one of the keys for the 1.5V chip. Mine? I've never taken it over 1.7 Volts. I'm at 2.31gHz so far and I'm happy. 200mHz (.2gHz) either way is not really going to be noticeable in the real world, so why bother frying the chip?

Click to expand...

im thinking of getting one. i have a generic 300w psu. will my chip suffer SADS? will it even work?

This is my attempt to explain why the 1700+ is so sensitive to higher Vcore and temperature.

For my 1700+, I can run it at:
2.2 GHz stable using 1.5V, 34 C load, 3000 rpm fan
2.4 GHz stable using 1.65V, 38 C load, 3000 rpm fan
2.5 GHz stable using 1.85V, 48 C load, 4800 rpm fan

For reasons of CPU health, environment reason of heat and noise (air cooling), it is obvious to run the CPU at 2.4 GHz at 1.65V, 38 C load, 3000 rpm fan, or even at 2.2 GHz 1.5V, ...

The 0.2V Vcore to get the CPU from 2.4 GHz to 2.5 GHz, are mostly wasted as heat instead of useful power for computation !!!! I.e. hitting a wall.

Human cannot distinguish between 2.4 GHz and 2.5-2.6 GHz, except for those benchmark programs for picture taking, ...

hitechjb1 said:

Relationship of clock, die temperature and Vcore

As far as Vcore, clock and die temperatue relationship, a chip (CPU) can be modeled as a capacitor C and a resistor R in parallel driven by Vcore. C models the useful active power to substain the computation by charging and discharging 100 millions of internal capacitors (from coupling between transistors, wires and silicon subtrate). R models the wasted leakage power through the internal current paths through the dozens millions of transistors.

If the die temp is kept low enough, in theory, todays XP and P4 can be clocked as high as 3 GHz, 4 GHz. The power (the C component) going into the chip to run the clock at a frequency f and Vcore V is given by

P_active = C V^2 f

And this can go on to 3-4 GHz if the die is kept below certain temp. Most of the power are used to power the clock faster as Vcore is increased.

But in reality, for any cooling used, air, water, vapor, liquid nitrogen, ..., the die temperature will eventually increase as Vcore increases due to leakage current which heats up the chip. Though at a different rate depends on what cooling is used. The leakage current is small at low temp, and increases with temp increases and also at a faster rate as temp increases. The power that heats up the chip (the R component) is given by

P_leak = V^2 / R

From my experiment with the TB B 1700+ DLT3C, when die temp reaches around 40C, the chip leakage current begins to increase at a faster pace, and heats up the chip more. Once this starts, any Vcore increase will split between powering faster clock, but also in heating up the chip. The exact Vcore when this occurs varies from chip to chip (100-200 mV difference), it depends on the "gene" how a particular CPU was born.

P = P_active + P_leak = CV^2 f + V^2 / R

After passing that temp threshold, the portion P_leak going into heating the chip (the R component) will become larger and larger, as Vcore is increased. In other word, the useful P_active to power the chip faster (the C component) will eventually become very very small. And the chip is just being heat up and cannot be clocked faster any more.


If better cooling than air such as vapor, liquid nitrogen are used, the chip may reach that temperature threshold (40C) at a much higher Vcore. That is, more power can be used to power the P_active portion to a much higher clock frequency before the CPU reaches that temperature threshold of high "run away" leakage. But still, the leakage current will still increase, but at a much slower pace due to lower die temperature, until which the die temp reaches, say 40C again, but at a much higher Vcore for a much higher clock frequency (hopefully 3-4 GHz). And at that point, overclocking will still begin to hit a wall like air cooling then.

Click to expand...

hitechjb1 said:

Human cannot distinguish between 2.4 GHz and 2.5-2.6 GHz, except for those benchmark programs for picture taking, ...

Click to expand...

This isnt entirely true. To the eye, you cant really tell a difference. And who really cares what benchmarks programs say. But try encoding video and you will definitely know the difference of a few hundred MB. Example, My wifes pc is nearly identical to mine except for the processors. Im running a 1700+ T-bred oc'ed to 1.94 in mine while she has a pally 1800+ running stock in hers. Converting the same DiVX to MPEG2 using the same programs and her pc takes a average of 30 minutes longer because of a approx. 400mhz difference.

And I know exactly what you mean about the voltage limits on some of the 1700+'s. Mine is a older one( 0251 I think ) and it will not run over 1.8v no matter what, even at stock speeds.

Pyros said:

This isnt entirely true. To the eye, you cant really tell a difference. And who really cares what benchmarks programs say. But try encoding video and you will definitely know the difference of a few hundred MB. Example, My wifes pc is nearly identical to mine except for the processors. Im running a 1700+ T-bred oc'ed to 1.94 in mine while she has a pally 1800+ running stock in hers. Converting the same DiVX to MPEG2 using the same programs and her pc takes a average of 30 minutes longer because of a approx. 400mhz difference.

And I know exactly what you mean about the voltage limits on some of the 1700+'s. Mine is a older one( 0251 I think ) and it will not run over 1.8v no matter what, even at stock speeds.

Click to expand...

As you said, your Tbred is running at 1.94 GHz, and her Pally 1800+ is running at stock which is 1.533 MHz. There is a difference of 400 MHz, which probably is distinguishable by human looking at the overall DIVX speed, which is like interpreting benchies.

The main point is every MHz counts. It is not that 100-200 MHz is not important, especially it is almost free (that is what oc is about).

But if needed to raise Vcore 0.2V for 100 MHz, that is not justify for the cost (heat and noise and CPU health) for continuous daily running.

Pyros said:

This isnt entirely true. To the eye, you cant really tell a difference. And who really cares what benchmarks programs say. But try encoding video and you will definitely know the difference of a few hundred MB. Example, My wifes pc is nearly identical to mine except for the processors. Im running a 1700+ T-bred oc'ed to 1.94 in mine while she has a pally 1800+ running stock in hers. Converting the same DiVX to MPEG2 using the same programs and her pc takes a average of 30 minutes longer because of a approx. 400mhz difference.

And I know exactly what you mean about the voltage limits on some of the 1700+'s. Mine is a older one( 0251 I think ) and it will not run over 1.8v no matter what, even at stock speeds.

Click to expand...

Thanks for the informative post... I have a 1700+ coming in the mail... now I'll be extra careful with the Vcore.

Im running at 2.4 (12x200) at 1.7v. Above 1.8v I believe the dielectric inside the chip its self starts to break down and after a while, you dont have a nice fast chip, you have an expencive wire :/ not good...

Mine's at a 1.65V setting according to the bios. But MBM reads it as 1.67 volts, yet the bios hardware monitor reads 1.64. I tend to believe the bios but without measuring it myself with one of my 4 wire voltmeters I don't know.

Hey airspirit, you wanna sell Proc #2 to me ?

I'm going to be doing a prometia test on this new 1700+ sometime in the next week or two. I've had it up to over 2800 Mhz WATERCOOLED, and I think my RAM just can't hack it at that point. I was only at 1.825V to do that speed. I'm using OCZ DDR400 (teh sux), and I had to slow the timings dramatically to even maintain those speeds (12.5x226->2825 Mhz). My computer booted into windows, and it appears that the minor instability I'm seeing is coming from the RAM. At 2850 (12.5x228), I have massive instability. I'm going to hopefully be plugging this into a Prommy system with some DDR433 and a new NF7, so I'll have to see what it can do at -40F, hehe. The only thing that is keeping from trying higher are those RAM issues and the fact that I WILL NOT run this chip over 1.825 for any reason. In the end, I'm going to keep it under 1.775 and take whatever speeds I can get.

I'm not going to sell this chip. No way. No how. Not gonna do it. Wouldn't be prudent. Oh, and did I mention that I haven't really even burnt this thing in yet?