• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

Question Jen Sung makes questionable decision? [RUMOR] NVidia tries to disable GPU mining?

Page 18 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Shmee

Memory and Storage, Graphics Cards
Super Moderator
Sep 13, 2008
5,006
831
126
I don't think people should be told what workload they can and cannot do with there card(s), especially if they had to pay extra for one. The best thing IMO, would be if sellers had a limit a person could buy, to give different people a chance to get them. Some stores already do this.
 

Dribble

Golden Member
Aug 9, 2005
1,899
426
136
What a joke. This can be currently fully workarounded with virtualization since you can spoof the PCI Vendor ID and Device ID.
How? As far as I know no one has broken the signed bios/drivers. The current 3060 only works because Nvidia released a engineering driver without it disabled, not because anyone overcame the security. Release a new card where that driver won't work and you are back to them successfully messing up mining.

In addition they have a good carrot/stick thing going on. The stick is they are hinting at releasing new cards with new security silently, and hinting that they might change what the crypto algorithms driver blocks. Hence you now take a risk ordering a bunch of these to mine and finding out they are rubbish. The carrot is "here our are mining specific cards that you know will mine well" - see recent press releases about mining companies buying lots of these, that's not accidental, they are designed to make the miners think about what they should buy.

If I were a mining company I'd just not bother with 3060's. I'd have to be very careful which ones I bought as I know some of them won't work well, even the old ones that do work require me to change everything to windows as the engineering driver is for windows (all the mining companies use linux). It's just not worth it, I'd be better off buying something else instead which is exactly what Nvidia want to happen.
 
Last edited:
  • Like
Reactions: ryan20fun

zir_blazer

Golden Member
Jun 6, 2013
1,017
236
116
How? As far as I know no one has broken the signed bios/drivers. The current 3060 only works because Nvidia released a engineering driver without it disabled, not because anyone overcame the security. Release a new card where that driver won't work and you are back to them successfully messing up mining.
If the only thing they did was to change the Device ID, it can be fully workarounded, since you just use the old Driver with the new card after spoofing that. Or the Drivers can be hacked so they bind to the new card, too. I think that it is rather self explanatory.


In addition they have a good carrot/stick thing going on. The stick is they are hinting at releasing new cards with new security silently, and hinting that they might change what the crypto algorithms driver blocks. Hence you now take a risk ordering a bunch of these to mine and finding out they are rubbish. The carrot is "here our are mining specific cards that you know will mine well" - see recent press releases about mining companies buying lots of these, that's not accidental, they are designed to make the miners think about what they should buy.

If I were a mining company I'd just not bother with 3060's. I'd have to be very careful which ones I bought as I know some of them won't work well, even the old ones that do work require me to change everything to windows as the engineering driver is for windows (all the mining companies use linux). It's just not worth it, I'd be better off buying something else instead which is exactly what Nvidia want to happen.
Pure garbage to get press releases doing virtue signaling to pander to gamers to look like they are doing something useful.

Mining cards just raise the ASP of the same silicon and guarantees that they aren't going to compete in the second hand used market after their useful mining life ends. These cards are as good as bricks, so I suppose that the ones that purchase them know that they can pay for themselves in a short time frame then give some more profit, otherwise no one would purchase them.
 

Mopetar

Diamond Member
Jan 31, 2011
5,756
2,520
136
I don't think people should be told what workload they can and cannot do with there card(s), especially if they had to pay extra for one. The best thing IMO, would be if sellers had a limit a person could buy, to give different people a chance to get them. Some stores already do this.
It doesn't really matter if you impose those limits. Even regular people who realize that they can flip a GPU for a substantial profit will do so when there's this much of a gap between MSRP and market price. The only way a limited purchase system actual gets cards into the hands of customers is if the store charges actual market price because then there's no incentive for anyone to flip cards and gamers willing to pay that actual price can purchase a card without having to buy from a scalper.
 

guidryp

Golden Member
Apr 3, 2006
1,170
1,249
136
If the only thing they did was to change the Device ID, it can be fully workarounded, since you just use the old Driver with the new card after spoofing that. Or the Drivers can be hacked so they bind to the new card, too. I think that it is rather self explanatory.
Signed drivers are extremely difficult to hack. That's the point of driver signing.
 
  • Like
Reactions: ryan20fun

VirtualLarry

No Lifer
Aug 25, 2001
50,897
6,315
126
Signed drivers are extremely difficult to hack. That's the point of driver signing.
I believe, that on Windows (10), only the driver binaries themselves are "signed", the installation .INF file that binds the PCI DeviceIDs to the actually binaries, is not signed, and thus the associations can be modified. At least, how I understand it.

Possibly, I'm wrong, and the ".cat" files, include the .INF in part of the manifest with the signature info too. But I didn't think so.
 

guidryp

Golden Member
Apr 3, 2006
1,170
1,249
136
I believe, that on Windows (10), only the driver binaries themselves are "signed", the installation .INF file that binds the PCI DeviceIDs to the actually binaries, is not signed, and thus the associations can be modified. At least, how I understand it.

Possibly, I'm wrong, and the ".cat" files, include the .INF in part of the manifest with the signature info too. But I didn't think so.
I guess we will find out when they release a new mining blocked card and people attempt to bypass it again.
 

Mopetar

Diamond Member
Jan 31, 2011
5,756
2,520
136
Signed drivers are extremely difficult to hack. That's the point of driver signing.
I wonder if a bunch of people who've developed an infrastructure surrounding solving cryptographic problems would have problems with signed drivers? Sure a single person would probably have a hell of a time, but a whole distributed network of people. . . Now where could we ever find one of those? Maybe if we made some kind of digital coin that could be given out for solving this problem it would incentivize people into doing the work!
 

Dribble

Golden Member
Aug 9, 2005
1,899
426
136
If the only thing they did was to change the Device ID, it can be fully workarounded, since you just use the old Driver with the new card after spoofing that. Or the Drivers can be hacked so they bind to the new card, too. I think that it is rather self explanatory.
So Nvidia went to all that effort and then put out all those press releases for something that is trivial to work around. I suspect it's not quite that simple. Nothing is unhackable, but equally you can't just assume everything is trivial to hack.
I wonder if a bunch of people who've developed an infrastructure surrounding solving cryptographic problems would have problems with signed drivers? Sure a single person would probably have a hell of a time, but a whole distributed network of people. . . Now where could we ever find one of those? Maybe if we made some kind of digital coin that could be given out for solving this problem it would incentivize people into doing the work!
They are probably using one of the standard crypro libraries that's already used to send real money banking information, company secrets, your credit card details on every purchase, etc. If you could break it then you'll be using it for a lot more than saving a few $$$ on your crypto gpu's.
 
  • Like
Reactions: guidryp

coercitiv

Diamond Member
Jan 24, 2014
4,308
5,514
136
They are probably using one of the standard crypro libraries that's already used to send real money banking information, company secrets, your credit card details on every purchase, etc. If you could break it then you'll be using it for a lot more than saving a few $$$ on your crypto gpu's.
People in the security world have this joke: your threat model isn't my threat model. Keep in mind the entire Nvidia cryptographic protection was rendered useless by one human error (releasing a signer driver with the limiter partially turned off).

With this context in mind, here's a recent joke made by a hacker during a live demo of breaking into a police union website:
Don't crack hashes. Backdoor login pages. It's better for the environment.
 

Mopetar

Diamond Member
Jan 31, 2011
5,756
2,520
136
Miners really don't overwork the cards, particularly since most are mining Etherium which hits a memory bandwidth bottleneck long before anything else. Since electricity costs cut into profits, most will even undervolt a stock card to cut down the power draw.

Really the only way they might be considered overworked is that they run around the clock, but that really isn't any different than hardware in a data center. Anyone who's just a small time miner using their own GPU is just finding ways to use their card while not gaming on it.
 

GodisanAtheist

Diamond Member
Nov 16, 2006
3,002
1,481
136
Miners really don't overwork the cards, particularly since most are mining Etherium which hits a memory bandwidth bottleneck long before anything else. Since electricity costs cut into profits, most will even undervolt a stock card to cut down the power draw.

Really the only way they might be considered overworked is that they run around the clock, but that really isn't any different than hardware in a data center. Anyone who's just a small time miner using their own GPU is just finding ways to use their card while not gaming on it.
- I personally would have no issue buying parts second hand from a miner, assuming the appropriate second hand discount is applied.

I have honestly had much worse luck with new products than used ones. No one knows if the new product is going to work out of the box, but the used product has been confirmed to be working by the prior owner. Many parts of an aftermarket GPU are so absurdly overmanufactured that they tend to either fail right away due to manufacturing defects or never fail during the useful life of the card at all.

I keep carrying forward parts of my system from prior builds to keep costs down, like some sort of Thessus' ship scenario. Some of the oldest parts of my PC are over 10 years old and many of those parts are $20 for a pack of 3 type case fans, AIO pumps, SSDs etc.
 

Mopetar

Diamond Member
Jan 31, 2011
5,756
2,520
136
- I personally would have no issue buying parts second hand from a miner, assuming the appropriate second hand discount is applied.

I have honestly had much worse luck with new products than used ones. No one knows if the new product is going to work out of the box, but the used product has been confirmed to be working by the prior owner. Many parts of an aftermarket GPU are so absurdly overmanufactured that they tend to either fail right away due to manufacturing defects or never fail during the useful life of the card at all.

I keep carrying forward parts of my system from prior builds to keep costs down, like some sort of Thessus' ship scenario. Some of the oldest parts of my PC are over 10 years old and many of those parts are $20 for a pack of 3 type case fans, AIO pumps, SSDs etc.
Most electronics have the standard bathtub curve when it comes to product failure, so buying slightly used is less risk assuming it wasn't someone who tortured their technology.

I think a lot of the bad rap comes from BitCoin miners back in the day who just crammed as many cards into a system as possible without any concerns for thermals. They probably killed a few cards that would have normally been fine, but people serious about mining have refined their operations a lot and have far better setups.
 

VirtualLarry

No Lifer
Aug 25, 2001
50,897
6,315
126
I think a lot of the bad rap comes from BitCoin miners back in the day who just crammed as many cards into a system as possible without any concerns for thermals. They probably killed a few cards that would have normally been fine, but people serious about mining have refined their operations a lot and have far better setups.
That was me circa 2017 mining, I was a total noob them, and doing things on the cheap, using cheap ASRock B350M Pro4 or whatever AM4 boards, to mine the daylights out of a pair of RX 470/570 cards, using double-slot spacing (right up against each other, thermal throttling, never heard of a "mining frame").
 

MrTeal

Diamond Member
Dec 7, 2003
3,103
775
136
Most electronics have the standard bathtub curve when it comes to product failure, so buying slightly used is less risk assuming it wasn't someone who tortured their technology.

I think a lot of the bad rap comes from BitCoin miners back in the day who just crammed as many cards into a system as possible without any concerns for thermals. They probably killed a few cards that would have normally been fine, but people serious about mining have refined their operations a lot and have far better setups.
Bitcoin also scaled in performance exactly with clock speed, right up until the edge of stability. There was a big incentive there to crank up the core clocks and push them as far as they'd go.
 

GodisanAtheist

Diamond Member
Nov 16, 2006
3,002
1,481
136
Bitcoin also scaled in performance exactly with clock speed, right up until the edge of stability. There was a big incentive there to crank up the core clocks and push them as far as they'd go.
-Even then, silicon degradation is fairly overstated and even a chip pushed to the edge of stability might have it's time to failure point pushed downward, we're usually looking at a reduction of 10 years to 8 years (using fictional numbers here).

Yes, that is a large and significant reduction in total lifespan of the core and associated components... But the remaining lifespan is still far longer than anyone would expect to reasonably keep using a GPU for any sort of relevant gaming.
 

amenx

Platinum Member
Dec 17, 2004
2,806
563
136
It doesn't really matter if you impose those limits. Even regular people who realize that they can flip a GPU for a substantial profit will do so when there's this much of a gap between MSRP and market price. The only way a limited purchase system actual gets cards into the hands of customers is if the store charges actual market price because then there's no incentive for anyone to flip cards and gamers willing to pay that actual price can purchase a card without having to buy from a scalper.
The problem lies at the source, China, Taiwan where most of these manufacturers are and where they are approached by mining farms to buy in bulk most their production at higher cost than MSRP. Much easier, more profitable, no RMAs or customer service to contend with, just easy bulk sales and they are done. These Chinese miners, farms have closer relations with their home based manufacturers than retailers in the west hoping to get whatever token quantities they can from them.

We've actually seen instances where miners have obtained quantities of cards that havent even been released (ie, 3060, 3060ti) or reviewed at the time they got them. That shows how close they are to the manufacturers.
 
Last edited:

MrTeal

Diamond Member
Dec 7, 2003
3,103
775
136
-Even then, silicon degradation is fairly overstated and even a chip pushed to the edge of stability might have it's time to failure point pushed downward, we're usually looking at a reduction of 10 years to 8 years (using fictional numbers here).

Yes, that is a large and significant reduction in total lifespan of the core and associated components... But the remaining lifespan is still far longer than anyone would expect to reasonably keep using a GPU for any sort of relevant gaming.
Depends on what you were using. Even with the blower running 100%, the VRM on a 5970 would sit at 125C 24/7 mining Bitcoin. The 5870 didn't get quite the hot, but they still were extremely toasty. A lot of those mining cards did not live happy lives and suffered painful deaths.
 

PingSpike

Lifer
Feb 25, 2004
21,483
387
126
I have honestly had much worse luck with new products than used ones. No one knows if the new product is going to work out of the box, but the used product has been confirmed to be working by the prior owner. Many parts of an aftermarket GPU are so absurdly overmanufactured that they tend to either fail right away due to manufacturing defects or never fail during the useful life of the card at all.
I too have noticed this bizarre change in some respects. I needed an m.2 to pcie converter awhile back, the reviews for all the ones available mentioned a lot of OOB failures. So I bought used ones from a miner. I figured if they hadn't burned that guys house down they were probably OK.
 
  • Like
Reactions: GodisanAtheist

GodisanAtheist

Diamond Member
Nov 16, 2006
3,002
1,481
136
I too have noticed this bizarre change in some respects. I needed an m.2 to pcie converter awhile back, the reviews for all the ones available mentioned a lot of OOB failures. So I bought used ones from a miner. I figured if they hadn't burned that guys house down they were probably OK.
- I mean, you're basically leveraging survivorship bias in your favor with used parts, and saving some money at the same time...
 
  • Like
Reactions: Leeea

ASK THE COMMUNITY