Cards being damaged by the 6950->6970 bios flash

[biostud]

Basically you just have to ask yourself whether an 6950 o/c + 6970 Flash is worth the risk compared to a 6950 o/c. Personally I wouldn't do it, since I seriously doubt that I would notice the 5-10% speed difference, while I would definitely notice a dead card.

 

[-Slacker-]

14 unsuccessful unlock attempts or buggy cards after unlock (total) out of 470 attempts sounds like a pretty good success rate to me.

If you're not willing to put your chips all in with a straight flush because someone else might have a royal flush, then you shouldn't be playing poker.

I'm talking of course about the techpowerup review.

Basically you just have to ask yourself whether an 6950 o/c + 6970 Flash is worth the risk compared to a 6950 o/c. Personally I wouldn't do it, since I seriously doubt that I would notice the 5-10% speed difference, while I would definitely notice a dead card.

Think of it this way.

A hd 6970 is $340 after rebate.

The hd 6950 has an equal or greater chance than 95% to unlock successfully.

95% out of $340 is $323

But a hd 6950 is $257 AR.

Seems like a pretty deal to me.

 

[HendrixFan]

Nope, I think you got it wrong, there's a full fledged 6970 BIOS unlock method or there is a shader unlock method. If you just use the "shader unlock method," you just enable the shaders that are turned off, the voltages and clocks remain untouched.

Exactly. This is the route I chose.

I ran it for a couple days with the shaders unlocked with no problems, both in desktop and gaming. Then I loaded up MSI Afterburner to overclock the card up to the 6970 speeds, without touching the voltages.

As is the case with any aftermarket mods, you do need to know what you are doing and you do need to progress with care. It was well known as the shader unlock mod was hitting the internet that the RAM on the 6950 was not rated to run at 6970 speeds. Anyone who was trying to get to 6970 speeds should have proceeded with caution.

With most overclocking, the people who fry parts probably were lazy or too aggressive about the process.

 

[Elfear]

No issues with my unlocked and oced 6950s.

 

[ronnn]

Not sure I get this thread. We all know modifying a computer part carries risk. The real question is how much risk and is the reward worth it.

So far the people here seem to have had good success and I haven't read anything about cards burning out. But I don't read much.

So it seems on the surface modifying a 6950 is safer (at least in the short term) than overclocking the 570. But if you want to live with minimum risk, buy a card that has developed a good track record and stay at stock.

 

[Wreckage]

Not sure I get this thread. We all know modifying a computer part carries risk. The real question is how much risk and is the reward worth it.

The reason for this thread is that a discussion regarding these failures was dragging another thread far offtopic.

Plus, many people have been acting like it has a 100% success rate with zero chance of failure or damage. However it would appear that at least 20 or more cards on Techpowerup alone have sustained permanent damage.

So I guess there's at least 2 reasons for this thread.

Also the heart of the issue that was being discussed in the other thread, was whether the different power requirements between the 6950 and 6970 could be causing some of the failures. It's worth noting that not all failures cause permanent damage to the card.

 

[zerogear]

I have had my 6950 flashed to 6970 for about 3 weeks and haven't had any problems since. There are people here that are just trying to run their personal agenda. Yes there are people that are going to have problems just like with running any other electronic out of spec. But why is a NVIDIA employee bringing it up? This is the reason I keep buying AMD.

Same here. I think some peoples cards were also probably just faulty to begin with.

 

[BFG10K]

Nope, I think you got it wrong, there's a full fledged 6970 BIOS unlock method or there is a shader unlock method. If you just use the "shader unlock method," you just enable the shaders that are turned off, the voltages and clocks remain untouched.

Obviously I’m talking about the situation where the clocks/voltage is raised too. Seriously, you know exactly what I mean so stop with the semantic games.

 

[tincart]

Reading through the TPU thread, I'm not seeing any evidence of permanent damage. Some people are reporting problems after flashing to a 6970 BIOS, but no one is giving info on what, if anything, they did to try and fix the problem once it developed.

 

[Wreckage]

Reading through the TPU thread, I'm not seeing any evidence of permanent damage. Some people are reporting problems after flashing to a 6970 BIOS, but no one is giving info on what, if anything, they did to try and fix the problem once it developed.

The very first post

In attempt to get to the root cause of why modded 6950 cards are starting to develop permanent damage, I've created this poll. If you have permanent damage to your 6950, even after going back to the stock bios, please take the poll and choose which option most closely describes the status of your card when it FIRST STARTED TO DEVELOP PERMANENT DAMAGE

Looks like they are up to 23 cards now.

 

[happy medium]

Obviously I’m talking about the situation where the clocks/voltage is raised too. Seriously, you know exactly what I mean so stop with the semantic games.

I agree, I said this back at post # 13 (I think) but it was ignored basically.

 

[Skurge]

I agree, I said this back at post # 13 (I think) but it was ignored basically.

Raise the voltage on any card and you risk damage, especially with some people telling members to overclock their cards and cpus willy nilly without warning them about the risk. You know who you are.

So far, no-one has shown that anything other than the ram timings, ram clocks and ram voltages being messing with is whats causing the problem. Messing with the timings on ANY card would cause permanent damage pretty quickly, I dare you to go ahead and try it. I also doubt the GPU itself would be damaged. Running disabled parts on a lower binned gpu would not cause damage or the GPUs would be damaged right from the factory when they test the GPU for it to be a Cayman XT or Cayman PRO.

 

[badb0y]

Obviously I’m talking about the situation where the clocks/voltage is raised too. Seriously, you know exactly what I mean so stop with the semantic games.

I hope your lube is ready. The person you quoted, HendrixFan, is basically saying that the VRM won't be affected if you use a BIOS shader unlock method which basically means that you takes YOUR card's BIOS, unlocks the shaders on it, and then flash it on to the card. The voltages, clocks, ram voltage, timing etc. are untouched.

You continue to bring up the 6970 BIOS in your reply to him when he isn't even talking about the 6970's BIOS. In conclusion, with all due respect, you need to stop arguing semantics and you need to learn about what you are talking about before you post.

This entire thread is arguing semantics, 97% success rate is phenomenal for turning your 260$ card to a 350$ card.

/thread.

 

[Skurge]

I'd like to know if ANY of those damage cards were simply shader unlocked 6950 and not 6950 flashed to 6970.

This is from a poster at TPU

I see it that some 6950s just simply have broken SPs.
And those few people who broke their card by flashing in a 6970 bios and had defects after going back to 6950 bios flashed with a bios other than a backup from their own card resulting in disabling a wrong SP partition, that is, defective SPs still being enabled and a working partition getting disabled.

Its funny, when the 465s could be unlocked to 470s (mind you that was a very limited selection of them) I don't remember all this hoopla on the forum. Figures.

 

[wahdangun]

I'd like to know if ANY of those damage cards were simply shader unlocked 6950 and not 6950 flashed to 6970.

This is from a poster at TPU



Its funny, when the 465s could be unlocked to 470s (mind you that was a very limited selection of them) I don't remember all this hoopla on the forum. Figures.

because there are too many extreme nvdia fanboy in here, its irritating tough, thats why i rarely come here

 

[evolucion8]

I'd like to know if ANY of those damage cards were simply shader unlocked 6950 and not 6950 flashed to 6970.

This is from a poster at TPU



Its funny, when the 465s could be unlocked to 470s (mind you that was a very limited selection of them) I don't remember all this hoopla on the forum. Figures.

But could be possible to flash the card's back with its original stock BIOS? It should disable the defective cluster as it was at factory, couldn't?

because there are too many extreme nvdia fanboy in here, its irritating tough, thats why i rarely come here

You are right, but the best thing is that when you prove them wrong, they stay shut for a while, and the other ones that can't stop spreading marketing propaganda, they run out of ideas and start to lie and troll so badly that everybody on the thread ignores them. :biggrin:

 

[OVerLoRDI]

This is a case of lack of control in the data. We have scattered data with way too many variables.

1. What kind of mod? shaders unlocked? 6970 clocks?
2. How much OC? how much voltage?
3. What kind of PSU?
4. The same ram chips used on the 6950 as 6970?

Obviously anecdote evidence of one case is nothing, but I will offer mine. I have 2 gigabyte 6950s unlocked to 6970s @ 6970 speeds, powered by a 850 watt Corsair PSU. I haven't had any issues, and my cards do a huge amount of day after day number crunching for distributed computing, so if this failure is common I wouldn't be surprised if my cards were some of the first to have issues.

My point is: too many variables and not enough data.