Nvidia Unhackable RTX 3060 anti-miners driver has just been hacked (this time for real)

[Moonzy]

8 minutes ago, RejZoR said:

I hate ppl who multi-quote every fucking sentence.

I'd appreciate if you'd at least quote or mention me when you're replying, because I seldom follow threads, I manually check them.

 

9 minutes ago, RejZoR said:

Learn to fuicking reply in paragraphs, jesus.

1) why the strong language though

2) I figured it gives context to what I'm replying to, that's why I quote things I'm specifically replying to.

 

I wish this doesn't generate any malice outside of this thread, but that's probably a lost hope for any discussion that gets this heated. (Personally I don't think ill of anyone unless they're being disrespectful or rude while we're discussing, which I don't think occured here)

 

Thread lock for bickering please, end this suffering

[leadeater]

1 hour ago, LAwLz said:

My guess is that peoples' hatred for Nvidia to fail at anything and everything they do is stronger than peoples' hatred for miners.

Well over the last year that balance scale has started tipping, and quickly. I'm not so sure which is hated more right now.

[Kilrah]

If you look at the big picture Nvidia have made great products, the only problem is they can't make enough of them but everybody else is in the same boat anyway.

 

Communication could have been done better, but it's not like it changes anything anyway.

[LAwLz]

49 minutes ago, Kilrah said:

Still WAY less than mining

I think that's an assumption you're making without having looked into it.

 

 

Mining is estimated to use 128.77 TWh annually. 

 

 

PC gaming alone is estimated to use 75 billion kWh annually (aka 75 TWh).

Console gaming is estimated to be slightly bigger than PC gaming. A console uses less power when playing (around 175 watts) but typically uses a more power hungry display (TV vs computer monitor) so if PC gaming is 75 TWh then console gaming might be 40 TWh? Less per user but slightly more users.

So we could estimate that gaming uses 115 TWh of power every year, not including data centers which is too big of an unknown to count.

 

So if we count "gaming" as a country then it would be at least the 31th most power hungry country in the world (only counting PC and console gaming, not servers or mobile).

Mining would be the 28th most power hungry country in the world.

 

 

 

If the argument is that "it uses lots of power, so therefore it's bad" then gaming is bad too, to almost the same degree as mining (gaming is estimated to use about 90% of the power mining is estimated to use).

 

 

Also, the whole "it uses a lot of power" has only been true for a little while. Up until January 2021 it was estimated below the 100 TWh mark which meant gaming has been using far more power than mining has up until very recently.

 

I think it's a bit hypocritical that people have been participating doing something that has used 100 TWh annually, but then as soon as something else also starts using that much power it suddenly becomes an environmental issue.

 

 

I don't buy the whole "mining needs to be stopped because it uses a lot of power" argument, at least not from people who play video games. I think people who say that are hypocrits.

[MageTank]

6 hours ago, RejZoR said:

Coz it's a made up thing based literally on nothing, that can crash any time and the whole time it's consuming jiggawatts of power worldwide while creating massive shortages of GPU's meant for gaming coz miners drive them out of factories in Asia on pallets. You could hardly made up a dumber thing than cryptomining.

[RejZoR]

2 hours ago, MageTank said:

LITERALLY, when people literally don't get it.

[SerpentXSF]

Yea, and it spread like wild fire.... so the DEV driver is everywhere now. Someone is comparing 470.05 against the public LOCKED drivers now, so yea only a matter of time..... same is gonna happen to 3080 Ti

 

Another miner

 

and...Another miner

[MageTank]

1 hour ago, RejZoR said:

LITERALLY, when people literally don't get it.

Oh, I am confident people understand that you dislike mining/crypto. There is nothing to misconstrue there. You may not have any objective reasoning behind it (if you do, you've yet to provide any), but the blatant misuse of the word "literally" is unforgivable. This generation is using it akin to that of the word  "like", and it's starting to lose its meaning.

 

Try figuratively. It might not have the same ring to it, but it would be far more accurate given the metaphorical context. It also helps people take your arguments more seriously when everything isn't literally literal in literalism, you feel me? Not literally speaking, of course... we need to maintain that social distance.

[IkeaGnome]

I may as well attack myself here. If miners bad, why aren't miners bad?

https://www.theworldcounts.com/challenges/planet-earth/mining/energy-use-in-the-mining-industry/story

[Fnige]

10 hours ago, Spotty said:

There's been enough threads lately that have gone down the rabbit hole of "is cryptocurrency ethical?". Let's give it a rest, alright?

They said no

[Moonzy]

3060 test result: (performance is of the 3060, other card performance is normal unless stated otherwise)

 

System:

Ryzen 5 3600

B550 Tomahawk MAG ATX

 

PCIE 2.0

Spoiler

- 3060 x16: 25MH/s

 

PCIE 3.0

Spoiler

- 3060 x16: 48MH/s

- 3060 x8: 48 MH/s

- 3060 x4: 25 MH/s

- 3060 x1 (riser): 25MH/s

 

- 3060 x16, 3060 ti x4: 48MH/s

     - Display move to 3060ti: 25MH/s 

 

Speculated requirements:

- pcie 3.0 or above, x8 connection or above

- having a display plugged into the 3060

 

Works alongside other GPU, as long as above conditions are met.

 

Given the above conditions, I tried plugging in a HDMI to VGA converter without actually plugging in a display, I get the full hash

I'll move it to one of my rigs and see how it fares

 

Thoughts:

Good news to gamers who wish to mine while not using their PC, miners will have to get creative if they wanna use it

 

But the existence of this driver tells us that the limit is on the driver side(?) Can't be too sure though

[Craftyawesome]

7 minutes ago, Moonzy said:

3060 test result: (performance is of the 3060, other card performance is normal unless stated otherwise)

 

System:

Ryzen 5 3600

B550 Tomahawk MAG ATX

 

PCIE 2.0

  Hide contents

- 3060 x16: 25MH/s

 

PCIE 3.0

  Hide contents

- 3060 x16: 48MH/s

- 3060 x8: 48 MH/s

- 3060 x4: 25 MH/s

- 3060 x1 (riser): 25MH/s

 

- 3060 x16, 3060 ti x4: 48MH/s

     - Display move to 3060ti: 25MH/s 

 

Speculated requirements:

- pcie 3.0 or above, x8 connection or above

- having a display plugged into the 3060

 

Works alongside other GPU, as long as above conditions are met.

 

Given the above conditions, I tried plugging in a HDMI to VGA converter without actually plugging in a display, I get the full hash

 

Thoughts:

Good news to gamers who wish to mine while not using their PC, miners will have to get creative if they wanna use it

 

But the existence of this driver tells us that the limit is on the driver side(?) Can't be too sure though

Do you think Nvidia is/was considering releasing this intentionally because this primarily benefits people who mine part time and not large miners who might buy CMP? How else would requirements like this get there? Protection against leaks? They said it was an accident, though https://wccftech.com/nvidia-confirms-470-05-driver-enabled-mining-on-geforce-rtx-3060/

 

It was supposed to be more than the driver.

 

So presumably the card validates the driver in some way? 

 

[Moonzy]

1 minute ago, Craftyawesome said:

Do you think Nvidia is/was considering releasing this intentionally because this primarily benefits people who mine part time and not large miners who might buy CMP?

Yes, I think they're responding to the feedback that the limit would hurt gamers more than large scale miners

 

This is the most logical move

 

1 minute ago, Craftyawesome said:

They said it was an accident, though https://wccftech.com/nvidia-confirms-470-05-driver-enabled-mining-on-geforce-rtx-3060/

From what I read online, the driver was pulled down because there's another issue with it, not because of the "unlocked" eth mining capability

 

I sure hope Nvidia don't backtrack this, it benefits gamers who just spent few hundreds on a GPU, it helps lessen their pain slightly if they choose to mine with it.

[leadeater]

57 minutes ago, Moonzy said:

Works alongside other GPU, as long as above conditions are met.

Not that miners would do it but that means (potentially) all you need is a PCIe backplane with a PCIe switch SoC in it that gives each slot x8 connectivity and just shares a larger uplink to a PCIe HBA in the system. That along with VGA converters and you could have multi GPU mining again, unless there is another condition you were not able to check for and that is multiple GPUs (of 3060's), if hash rate drops for 2nd or 3rd even with all these other conditions met.

[Craftyawesome]

32 minutes ago, Moonzy said:

From what I read online, the driver was pulled down because there's another issue with it, not because of the "unlocked" eth mining capability

Source? They maybe also said that, but they definitely said "A developer driver inadvertently included code used for internal development which removes the hash rate limiter on RTX 3060 in some configurations”

 

13 minutes ago, leadeater said:

Not that miners would do it but that means (potentially) all you need is a PCIe backplane with a PCIe switch SoC in it that gives each slot x8 connectivity and just shares a larger uplink to a PCIe HBA in the system.

Is this cost effective over buying 3060 TIs? (Assuming you are paying similar scalper premiums on both of them)

[Master Disaster]

52 minutes ago, Craftyawesome said:

Do you think Nvidia is/was considering releasing this intentionally because this primarily benefits people who mine part time and not large miners who might buy CMP? How else would requirements like this get there? Protection against leaks? They said it was an accident, though https://wccftech.com/nvidia-confirms-470-05-driver-enabled-mining-on-geforce-rtx-3060/

 

It was supposed to be more than the driver.

 

So presumably the card validates the driver in some way? 

 

I call BS on this. If what BDR said was accurate then the broken driver simply would not work. The handshaking process would simply fail when trying to authenticate the driver.

 

The point of the 3 step authentication would be to have a redundant check that all 3 are in place yet Nvidia somehow bypassed it with only 1 of the steps. That doesn't compute.

[Rauten]

1 hour ago, Moonzy said:

PCIE 2.0

  Hide contents

- 3060 x16: 25MH/s

 

PCIE 3.0

  Hide contents

- 3060 x8: 48 MH/s

 

Ok, so I know little to absolutely nothing about mining but, this just doesn't make sense to me.

Isn't PCIE 2.0x16 supposed to be the same speed (within margin of error) as PCIE 3.0x8?

Why the near 2x boost in mining performance?

[Moonzy]

18 hours ago, Rauten said:

Isn't PCIE 2.0x16 supposed to be the same speed (within margin of error) as PCIE 3.0x8?

Why the near 2x boost in mining performance?

Basically Nvidia is limiting your eth hash rate if you're operating below pcie 3.0 x8

 

Which is probably not what 99% of the gamers would use, the 1% are the ones stuck on 2nd gen intel with pcie 2.0

 

Miners use pcie risers that are only x1, and most of the time running pcie gen 2 or less because of stability issues over the risers

Mining only requires pcie 1.1 x1 to function, so it's nothing to do with bandwidth

 

18 hours ago, Craftyawesome said:

Source?

Dude, trust me!

 

I saw it while I was researching this, it was mentioned in a forum posts elsewhere but I didn't remember seeing a source, so take it with a big grain of salt

 

I'm interested to see if Nvidia truly implemented a way for gamers to mine, that'd be nice of them

[Craftyawesome]

7 minutes ago, Master Disaster said:

I call BS on this. If what BDR said was accurate then the broken driver simply would not work. The handshaking process would simply fail when trying to authenticate the driver.

 

The point of the 3 step authentication would be to have a redundant check that all 3 are in place yet Nvidia somehow bypassed it with only 1 of the steps. That doesn't compute.

That's how it should work, but maybe the card just doesn't authenticate enough. If it only checks for an official unmodified driver, this is one.

 

Also IIRC, someone managed to use a 3060 with the 3070 drivers before launch, and had the speed limited. So either the code was already in the drivers or there is something in the card itself.

[leadeater]

1 hour ago, Craftyawesome said:

Is this cost effective over buying 3060 TIs? (Assuming you are paying similar scalper premiums on both of them)

Nooope, not by a long shot lol

[Master Disaster]

49 minutes ago, Craftyawesome said:

That's how it should work, but maybe the card just doesn't authenticate enough. If it only checks for an official unmodified driver, this is one.

 

Also IIRC, someone managed to use a 3060 with the 3070 drivers before launch, and had the speed limited. So either the code was already in the drivers or there is something in the card itself.

Correct but it still doesn't address the main point which is...

 

If there's a driver check, firmware check and silicon check all in place to prevent this then how where Nvidia able to bypass it using only a driver on cards that have been sold at retail.

 

I could understand if they also had a different VBIOS for developer use which, when combined with the dev driver allows the check to by bypassed however that's not the case. Nvidia bypassed a three step redundant check by altering only one step on retail cards being sold to the public.

 

Either

A) They baked this feature into all the cards and are testing it internally for later release or

B) They're lying about the 3 step checking process

[LAwLz]

1 hour ago, Master Disaster said:

Correct but it still doesn't address the main point which is...

 

If there's a driver check, firmware check and silicon check all in place to prevent this then how where Nvidia able to bypass it using only a driver on cards that have been sold at retail.

 

I could understand if they also had a different VBIOS for developer use which, when combined with the dev driver allows the check to by bypassed however that's not the case. Nvidia bypassed a three step redundant check by altering only one step on retail cards being sold to the public.

 

Either

A) They baked this feature into all the cards and are testing it internally for later release or

B) They're lying about the 3 step checking process

I think you're misunderstanding what BDR said.

 

Correct me if I am wrong, but you are interpreting his post as him saying "the driver has a hash rate limiter, the silicon has a hash rate limiter, and the firmware has a hash rate limiter, and all three are separate systems which will throttle the hashrate".

Because clearly you just had to change the driver to get the full hash rate, you are calling BDR a liar. Correct?

 

 

What BDR actually said (or at least meant) was that it is the driver that controls the throttling, but the 3060 includes special firmware and hardware which makes sure you can't just install an unauthorized driver for the 3060 with doesn't include the hash rate limiter.

 

What BDR said:

Quote

We have a lock that requires three keys.

 

 

What you interpreted it as:

Quote

We have three separate locks.

 

 

 

Since the driver was signed by Nvidia, it passes the secure handshake controlled when checked by the 3060 silicon and firmware.

From what I have read, the secure handshake hasn't been cracked yet, nor has the "anti-miner driver been hacked" either. Nvidia just released a driver that had looser restrictions for how the 3060 operates during mining workloads. Nvidia are still in control over how the card works when mining. Until someone managed to bypass the secure handshake which allows for third party drivers to be installed, nothing has been cracked.

 

 

Think of it like this.

• Blizzard releases Diablo 4 and on day 1 they also launch a bunch of really expensive DLC with heavy DRM. One of the DLC items is a really powerful sword.
• A lot of customers complain because they want the sword, but because the DRM is so good, nobody manages to crack it. The only way to get the sword is to buy it.
• After a lot of complaining, Blizzard releases a beta-update where the DLC sword becomes a rare drop from a boss. They want to experiment if it's a good idea to let free to play players have a way of getting the sword as well, without having to buy the DLC.
• News sites run headlines about how Diablo 4's DRM has been cracked because now you can suddenly get the sword without buying the DLC.

 

That's what is happening here, except replace Blizzard with Nvidia and DLC with mining.

Nothing has been cracked. It's still Nvidia that's in control over how the card works. 

[Distinctly Average]

21 hours ago, LAwLz said:

I think that's an assumption you're making without having looked into it.

 

 

Mining is estimated to use 128.77 TWh annually. 

 

 

PC gaming alone is estimated to use 75 billion kWh annually (aka 75 TWh).

Console gaming is estimated to be slightly bigger than PC gaming. A console uses less power when playing (around 175 watts) but typically uses a more power hungry display (TV vs computer monitor) so if PC gaming is 75 TWh then console gaming might be 40 TWh? Less per user but slightly more users.

So we could estimate that gaming uses 115 TWh of power every year, not including data centers which is too big of an unknown to count.

 

So if we count "gaming" as a country then it would be at least the 31th most power hungry country in the world (only counting PC and console gaming, not servers or mobile).

Mining would be the 28th most power hungry country in the world.

 

 

 

If the argument is that "it uses lots of power, so therefore it's bad" then gaming is bad too, to almost the same degree as mining (gaming is estimated to use about 90% of the power mining is estimated to use).

 

 

Also, the whole "it uses a lot of power" has only been true for a little while. Up until January 2021 it was estimated below the 100 TWh mark which meant gaming has been using far more power than mining has up until very recently.

 

I think it's a bit hypocritical that people have been participating doing something that has used 100 TWh annually, but then as soon as something else also starts using that much power it suddenly becomes an environmental issue.

 

 

I don't buy the whole "mining needs to be stopped because it uses a lot of power" argument, at least not from people who play video games. I think people who say that are hypocrits.

Of course it is a bit hypocritical, but why not? The same could be said for a vegan like me who is driving a car with parts that may have come from animals, Thing is, we have to start somewhere or nothing will ever change. We cannot endlessly consume for nothing but wealth. I am sure the fossil fuel and electricity generation industry are investing heavily in crypto at the moment. Hopefully etherium are right and the industry will change, as the exponential rise cannot continue forever.

 

While on the subject of gaming, we seem to be obsessed with low power CPUs while wanting more power from GPUs like the 30xx series. That will have to reach a peak soon too and nVidia/AMD will need to find ways of

making their products more energy efficient. Hopefully sooner rather than later.

[Kisai]

3 hours ago, Moonzy said:

Basically Nvidia is limiting your eth hash rate if you're operating below pcie 3.0 x8

 

Which is probably what 99% of the gamers would use, the 1% are the ones stuck on 2nd gen intel

 

Miners use pcie risers that are only x1, and most of the time running pcie gen 2 or less because of stability issues over the risers

Mining only requires pcie 1.1 x1 to function, so it's nothing to do with bandwidth

 

Funny enough, I speculated on this months ago.

 

https://linustechtips.com/topic/1290435-nvidia-and-amd-could-easily-stop-this-mining-gpu-issue/page/2/?tab=comments#comment-14367837

Quote

The only way to go “Game machine or workstation” only would require that all 16 lanes work at PCIe 4.0,” which means anyone who isn’t for some reason (eg a PCIe 3.0 system or a budget chipset) would then complain about it as well. eGPU’s only have 4 lanes in thunderbolt, that would also leave them out as well.

 

Miners tend to just use their cards in a headless PCIe 1.1 1-lane mode to maximize the amount of GPU’s they can put on one system board. That’s the only weak point there is here to blocking mining, and that again, only means they buy more Systems, it does nothing for GPU availability.

 

The thing is, at least from your test, is that it looks like this is exactly what is happening here, if the card doesn't run with a monitor attached at x8 or x16, it switches to a crippled mode.

 

It also shows a major oversight if that was the intent as well, the card needs to query the monitor DDC and if it doesn't respond, assume the monitor is turned off and switch to "power saving" mode. 

 

This could be a sign that nVidia might be making deliberately be making CMP and data center A100 parts "perform better" over the geforce parts used in servers. Think about things like stadia and competing cloud gaming platforms.

 

I'm not sure I like this. Any artificial limit/lockout on hardware that isn't designed to protect the device from melting/catching fire, is usually a bad faith action taken by the manufacturer to secure more profits by forcing customers to overpay for a feature that isn't even bin'd out.

 

 

 

[Moonzy]

3 hours ago, Distinctly Average said:

nVidia/AMD will need to find ways of making their products more energy efficient. Hopefully sooner rather than later.

leaked photos of next gen GPU, courtesy of @Den-Fi

Spoiler