Soul_Because of this statement "NVIDIA appears to have introduced a new flagship model with the passively-cooled 170HX"

Let's review..

UskompufNVIDIA announced the first four Crypto Mining Processor (CMP) cards earlier this year with performance ranging from 26 MH/s to 86 MH/s. These cards were all based on existing Turing/Ampere silicon and featured board partner-designed cooling systems. NVIDIA appears to have introduced a new flagship model with the passively-cooled 170HX that is based on the NVIDIA A100 accelerator which features a GA100 GPU.

This new model is the first mining card to be designed by NVIDIA and features 4480 CUDA cores paired with 8 GB of HBM2E memory which are both considerably less than what is found in other GA100 based products. NVIDIA has also purposively limited the PCIe interface to Gen 1 x4 to ensure the card cannot be used for tasks outside of cryptocurrency mining. The 170HX has a TDP of 250 W and runs at a base clock of 1140 MHz with a locked-down BIOS that does not allow memory overclocking resulting in a hash rate of 164 MH/s when using the Etash algorithm.

Taken in context, this seems to be the best and most capable mining card they offer. So in no way is the statement you quoted misleading on any level. Then when you take into account the fact they are citing this source and this one, Uskompuf is only offering some insight based on that data.

You still have not shown where TPU is at fault.

ColddeckedEtherium needs like 5 gb. By the time it needs more than 8gb it should already be POS. I don't think there's any coin that takes more than 8gb to mine.

Right. After the comments from Solaris and Trog, I did a bit of quick reading. The idea I had was that amount of VRAM was eqaully important as the speed of the RAM & GPU. That was my bad..

lexluthermiesterLet's review..


Taken in context, this seems to be the best and most capable mining card they offer. So in no way is the statement you quoted misleading on any level. Then when you take into account the fact they are citing this source and this one, Uskompuf is only offering some insight based on that data.

You still have not shown where TPU is at fault.


Right. After the comments from Solaris and Trog, I did a bit of quick reading. The idea I had was that amount of VRAM was eqaully important as the speed of the RAM & GPU. That was my bad..

You are trying way too hard to grasp at straws right now. Mr. Defense attorney, just adding fluff around a statement does not bring context. It just brings in redundant information. Fact of of the matter is simple, it is not passively cooled and TPU incorrectly mentioned it as such.

So take it whichever way you would like. If you just wish me to tell you that "you win", I will happily do so to get you the sleep you so dearly need.

Soul_Fact of of the matter is simple, it is not passively cooled and TPU incorrectly mentioned it as such.

And YOU were telling ME to read the article? See the posted pictures? See the two photos that CLEARLY show the inside of the vents on the card? Notice anything in particular? Like the fact that we can see straight through it with nothing obstructing the view? There's no fan in there. Nor is there one mounted on that chassis of the card.

Want to think about that for a few moments?

lexluthermiesterAnd YOU were telling ME to read the article? See the posted pictures? See the two photos that CLEARLY show the inside of the vents on the card? Notice anything in particular? Like the fact that we can see straight through it with nothing obstructing the view? There's no fan in there. Nor is there one mounted on that chassis of the card.

Want to think about that for a few moments?

LMAO!!! Just because it doesnt show it mounted in the chassis, doesnt mean that it is not meant to be chassis cooled.

So look at those pictures again. And tell me how you would cool 250 watt TDP, with size that it has and with a heatsink which is closed from 4 directions.

It is closed from 4 directions because it needs to streamline east-west airflow. I guess I just read the article a bit closer than you, and looked at the pictures a bit closer than you. And here you are standing and trying to tell me to "think for a few moments"? Laughable.

Soul_You are right, I might be nit picking, but technicalities you know. Passively cooled means that it is able to stay within thermal limits without need of a fan. Chassis cooled means that it doesn't need fan on chip, but needs chassis airflow to stay within thermal limits. Two different things.

Lol i get you passively cooled sounds better than dependant on case fan cool. But still little nitpicky.

ColddeckedLol i get you but you but yea its a bit nitpicky. But yea passively cooled sounds better than dependant on case fan cool.

That is true..

Soul_So look at those pictures again. And tell me how you would cool 250 watt TDP, with size that it has and with a heatsink which is closed from 4 directions.

Not familiar with server rack cooling are you? This card is designed for that kind of cooling scheme in mind. So this is in fact YOUR failure to understand a detail, not TPU/ Uskompuf. Or you're trolling. Either way, the problem here is you.

ColddeckedLol i get you but you but yea its a bit nitpicky. But yea passively cooled sounds better than dependant on case fan cool.

The card in question has no mechanism to cool itself and relys on an external source of cooling. That is the definition of passive cooling.

ColddeckedBut still little nitpicky.

And semantics..

lexluthermiesterNot familiar with server rack cooling are you? This card is designed for that kind of cooling scheme in mind. So this is in fact YOUR failure to understand a detail, not TPU/ Uskompuf. Or you're trolling. Either way, the problem here is you.

Please enlighten me how rack cooling works. LOL!!!

Rack cooling's responsibility is not to extract heat from components inside the chassis. Rack cooling's responsibility is to extract the heat from hot aisle and pump cold air to cool aisle. It is chassis cooling's responsibility to suck in the air from cold aisle, cool internal chassis components and pump hot air into hot aisle.

You will just keep on digging yourself deeper. I have designed datacentres for 15 years of my life. I am an engineer by education.

Soul_You will just keep on digging yourself deeper.

Am I? Are you sure?...

Soul_I have designed datacentres for 15 years of my life. I am an engineer by education.

Ah, so you admit you're trolling. Do you see how easy it was to get you to admit that? Well done there. Well done indeed.

lexluthermiesterAm I? Are you sure?...

Ah, so you admit you're trolling. Do you see how easy it was to get you to admit that? Well done there. Well done indeed.

LOL!! Yes, every one can see that you are looking good by every comment that you make here. Good work!! I am just proud to be in conversation with such an enlightened individual.

Now that is trolling...

Ethash algo for ETC maybe?

This card seems tailor-made for ethereum mining. It is quite similiar to the vega cards in many ways. The core count really doesn't matter (underclocking is the norm), vram bandwidth and tight timings is paramount (hence the hbm) while 8 gb of capacity would be enough for 5-6 years should the mining model remain pow. Plenty of time roi wise.

Now enters the shift to pos, and everything goes tumbling down. Just breaking even on this thing is quite a gamble...

111frodonshould the mining model remain pow

Which it's not. Proof of Stake is already being adopted and deployed...

111frodonNow enters the shift to pos, and everything goes tumbling down. Just breaking even on this thing is quite a gamble...

Right. So the question is, will NVidia release this card to retail once the public figures out they would not make much money with such a card...

lexluthermiesterUm, I'll admit my knowledge of mining is less than expansive, but last time I checked where GPU coin mining is concerned, the more VRAM the better and 8GB is on the lean side. Is that not so?

@trog100 , @R-T-B
You're miners, want to chime in? Anyone else is of course invited to join them.

VRAM good. But only to a point. After about 6GBs, extreme diminishing returns at least for ethereum.

R-T-BVRAM good. But only to a point. After about 6GBs, extreme diminishing returns at least for ethereum.

I did not know that before today. Was under the impression that modern mining was RAM intensive, needing both speed and capacity.

lexluthermiesterI did not know that before today. Was under the impression that modern mining was RAM intensive, needing both speed and capacity.

It's sort of like gaming. Not having enough friggin sucks. Having more than enough is just wasted.

trog100i think memory speed is more important than the amount of it.. it seems 8 gig is enough.. i expect eth to go up significantly in price as well.. it will be interesting to see how this effects mining rewards.. mine seem to be going up but they are a bit all over the place at the moment..

trog

Memory size obviously matters. If you take ETH for example, 4GB is'nt enough anymore. 8GB would set you for quite some time still if the thing does'nt turn into Proof of stake.

Jismif the thing does'nt turn into Proof of stake.

That is in process. The conversion has begun.

R-T-BIt's sort of like gaming. Not having enough friggin sucks. Having more than enough is just wasted.

Fair enough. I've learned more about mining today than I ever wanted too. HAHAHA!

lexluthermiesterThat is in process. The conversion has begun.


Fair enough. I've learned more about mining today than I ever wanted too. HAHAHA!

its more about when mining stops than when staking starts.. that date is still in doubt..

staking is a bit like putting money in the bank and gaining interest from it.. it already goes on..

my mining rewards are going up along with the price of eth.. i expect the price of eth to at least double or treble before the switch over to proof of stake happens..

if this is the case eth mining still has some mileage left in it before the switch over to proof of stake happens some time in 2022..

trog

ps.. i have seen some examples of the gains from staking and compounding crypto.. over ten years or so they are impressive.. crypto is the future i really dont see how it can fail to be.. the days of conventional banking are coming to an end.. like it or not lex.. its gonna happen.. :)

trog100if this is the case eth mining still has some mileage left in it before the switch over to proof of stake happens some time in 2022..

Yeah we were talking about this in the other thread, where PoS work has started getting rejected and those portions of Eth are lost forever. So we both know the conversion has started.

trog100the days of conventional banking are coming to an end.. like it or not lex.. its gonna happen..

That is a pipe dream. Will never happen.

Looks like a 3060 with HBM2.

It would have been nice for my ETH farm six months ago but it's specifically targeting ETH with memory bandwidth above anything else, and the problem with that is that ETH is planning (and making reasonable progress towards) becoming a Proof-of-Stake cryptocurrency in December, at which point this card will be e-waste.

It's no good for Ravencoin or ERG because it lacks the compute power those coins need. Maybe there are other coins that have requirements for high bandwidth and low compute, but they're neither popular nor currently profitable, so why would anyone risk it?

So, as an ETH-specific card, it's likely to be launched at almost exactly the point in time that mining ETH is supposed to stop, permanently. Now there's a strong chance that ETH will be mineable well into 2022 due to delays in the Proof-of-Stake transition, but as a business model, buying ETH-specific cards like this now is almost guaranteed failure unless the price of these is low (unlikely with HBM2) and ETH's PoS transition is very delayed indeed. Even then, you'd still be better off getting a more balanced card that can more profitably mine other coins once the inevitable end to ETH mining eventually happens.

I should also point out that my 2019-vintage RX5700 farm gets exactly 164MH/s from three cards, using around 330W on the GPUs alone, so whilst this is a decent efficiency at 250W for exactly the same hashrate as old RX5700 cards, it's not really good enough to justify buying single-purpose hardware with no future and no resale value.

Dyatlov AIf bitcoin price goes down, it cannot be sold.

Yes, cause it goes to bottom and never recovers and just dies.

To be fair, a lot of ppl sold their farms after the bull run to 18k. Unsure how they felt when BTC went to 60k.

Nvidia: The way it's meant to be mined.

Chrispy_Looks like a 3060 with HBM2.

It would have been nice for my ETH farm six months ago but it's specifically targeting ETH with memory bandwidth above anything else, and the problem with that is that ETH is planning (and making reasonable progress towards) becoming a Proof-of-Stake cryptocurrency in December, at which point this card will be e-waste.

[ ... ]

For how many times has the transition been postponed now? The Umpteenth?
Spoiler alert: It'll never happen, the developers own far too many ETHs still that they do value. Abandoning mining will kill ETH stone dead, zeroing out the developers wallets. They're not that dumb.

SelayaFor how many times has the transition been postponed now? The Umpteenth?
Spoiler alert: It'll never happen, the developers own far too many ETHs still that they do value. Abandoning mining will kill ETH stone dead, zeroing out the developers wallets. They're not that dumb.

As a miner, I really hope you're right. I'll be happy to keep making $100+ a day beyond December if mining can continue.

Owning ETH favours PoS though, so the developers are the ones trying to make ETH a greener cryptocurrency. Moving away from mining also increases scarcity and stops inflation, so there's a double incentive for the ETH developers/holders to hasten the transition from PoW to PoS since scarcity increases the relative value of any held ETH.

Postponing the PoS transition is what all miners want, me included, but the London fork on 5th Aug is something like step 3 on the 5-step journey to a full PoS transition and its been very well received by the market and shows that the PoS transition roadmap can continue as planned. A lot of us miners were hoping that EIP-1559 would be a bit of a flop and provide more arguments to delay the PoS transition, but it's had the opposite effect and boosted ETH's value by 100% in under 30 days.