Friday, August 4th 2017

AMD RX Vega Mining Performance Reportedly Doubled With Driver Updates

Disclaimer: take this post with a bucket of salt. However, the information here, if true, could heavily impact AMD's RX Vega cards' stock at launch and in the subsequent days, so, we're sharing this so our readers can decide on whether they want to pull the trigger for a Vega card at launch, as soon as possible, or risk what would seem like the equivalent of a mining Black Friday crowd gobbling up AMD's RX Vega models' stock. Remember that AMD has already justified delays for increased stock so as to limit the impact of miners on the available supply.

The information has been put out by two different sources already. The first source we encountered (and which has been covered by some media outlets solo) has been one post from one of OC UK's staff, Gibbo, who in a forum post, said "Seems the hash rate on VEGA is 70-100 per card, which is insanely good. Trying to devise some kind of plan so gamers can get them at MSRP without the miners wiping all the stock out within 5 minutes of product going live."
The fact that a staffer from OC UK is actually looking for ways to prevent miners from wiping the stock speaks more to the credence of the information than the "70-100 hash rate" claim. Apparently, this information was conveyed to Gibbo from an AMD AIB partner, who remains unknown at time of writing. This information has already been sort of confirmed by a second source, coming out of Videocardz's Why Cry. In a post, the editor reported how he already had come in possession of similar information through his sources, who put the hash rate of RX Vega close to double that of the Frontier Edition's original hash rate, which was ~30 MH/s in Ethereum mining. This means the hash rate could be ~60 mark, which is still close to OC UK's Gibbo's reference to a "70-100" hash rate. This increase in hash rate was apparently indirectly enabled by updated driver features for the RX Vega cards. Apparently, these were included in the drivers to improve features and gaming performance - which also indirectly resulted in increased mining capabilities.

We had already covered in our Vega Architecture Technical Overview article that AMD's Vega NGCU computation capabilities were bolstered through added support for over 40 new ISA instructions, which result in increased IPC over Polaris - and which were also mentioned by VSG at the time as being "very relevant for GPU mining."
Adding to this story is the fact that recent optimizations from a Reddit user to a Monero mining program and an underclock to 1.3 GHz have brought the Frontier Edition's mining hashrates to around 1.16 kH/s - 34% faster than a GeForce GTX 1080 Ti (around 0.76 kH/s according to Nicehash), and 43% faster than a single Radeon RX 580 8 GB. This means that the $999 Frontier Edition currently stands at double the mining performance of the GTX 1080 on Monero - and the gaming RX Vega, with its $499 price-tag, should follow suit. And these are optimizations achieved by a single user, for a cryptocurrency that is admittedly not as popular as Ethereum or others. But increasing levels of performance in some mining algorithms really does leave the door open for exploration of improved speeds on others, and you can rest assured that miners will, at the very least, attempt to achieve these optimizations in other cryptocurrencies.
All in all, if true, these reports lend credence to AMD's take on the RX Vega delays for stock build-up. And the situation seems to be less straightforward than one might hope when it comes to disabling these instructions, or only enabling them at a latter date, after gamers had already had some time to purchase their desired cards. Because these driver-level updates were apparently done with the intent to bolster gaming performance, I believe it's safe to assume AMD can't easily neuter the mining improvements without putting the increased gaming performance at risk as well. Let's see how this pans out, but consider yourselves at least warned - the RX Vega may see much reduced stock and increased pricing throughout if this scenario pans out. in the meantime, those Radeon Packs with both their shortcomings and opportunities are looking like an increasingly interesting way to get ahold of one of AMD's latest...
Sources: OCUK's Gibbo, Videocardz, NAG, Dirtbagdh @ Reddit
Add your own comment

216 Comments on AMD RX Vega Mining Performance Reportedly Doubled With Driver Updates

#151
Vya Domus
FordGT90ConceptEh, I don't buy the argument that running a card at 100% load for extended periods of time really impacts life span. It just exposes design flaws faster. A card with crappy hardware might last just over a year under normal gaming load but when stressed, they could crap out in a month. A card with good hardware shouldn't care if it was stressed for a year or five years.

Where's the evidence that this card that survived a month was ever used for mining?
This has occurred multiple times , where they incredibly unlucky ? Maybe , more than likely not. Mining does kill cards more often than gaming.

It's so easy to put a broken card in the oven and often make it work just like new , you don't seriously imply this isn't a thing.
Posted on Reply
#152
cdawall
where the hell are my stars
Vya DomusThis has occurred multiple times , where they incredibly unlucky ? Maybe , more than likely not. Mining does kill cards more often than gaming.

It's so easy to put a broken card in the oven and often make it work just like new , you don't seriously imply this isn't a thing.
Depends on what kind of gaming we are talking about. Running games 16hours a day everyday at full tdp wouldn't be anymore stressful than mining at 60-70% tdp
Posted on Reply
#153
R-T-B
RejZoRThere is no manufacturing capacity that can satisfy this mining idiocy. And no one is going to build a new plant just to satisfy these demands when it can die over night. Just too much of a gamble increasing manufacturing capacities 10 fold and then it all dies off over night. It's why we have such clusterf**k. If there was guaranteed capacity needed, the'd done it already.
That's complete BS. Eventually, hardware acquisition cost exceeds proft margin, even for miners, as the amount of possible profit remains fixed.

The fact is the manufacturers are enjoying the high prices as much as anyone, and that's why they aren't ramping up production more than anything.
RejZoRWe are blaming crypto miner degenerates for the major part...
So it's "degenerate" to make money off a product (AND run a payment network) you were planing on gaming on?

If you ask me honestly, if anyone is degenerate, it's the gamers. But I don't get judgemental and sling insults, so no one is.
bugI think they can. Afaik, mining is mostly about FP66
Not at all really, no.
Prince ValiantCouldn't they cripple performance with x+ GPUs detected in a system unless you buy a card specifically for it? Or do cards not communicate at all with mining?
Cards do not use the bus for anything but submitting hashes with mining, so no, they don't communicate.
Captain_TomThat is when you switch to a different coin lol.
Irrelevant. Everyone cashes out in bitcoin.
notbWhat you're doing here is pretty unfair and IMO even offending to part of this community.

Most people that get into an investment before the bubble make a profit - this is fairly basic stuff.
But most people that start after the "greed" moment (from the graph in @efikkan 's post) don't make a profit.
So if this turns out to be a bubble and profitability stays on current level (or more likely: drops even further), people that started in June will lose money.

Your role in this is pretty simple. You're the one who made the hugely popular mining thread - you've presented mining as a great investment with no risk ("free money"), you've consulted people on setting up their rigs etc. It was most likely the first time many people on TPU really decided to learn something about mining (me included) and some of them invested into it as a result.

After merely 2 months you're simply telling those people that they started to late. Nice going.
More mining doesn't help him though, it hurts him, which pretty much shoots a hole in any benefit argument you can make against him.

Also, people have been trying to apply that graph to bitcoin since bitcoin was $20.00. It doesn't work.
Posted on Reply
#154
Vya Domus
R-T-Band that's why they aren't ramping up production more than anything.
I am still highly doubtful of that , like I previously said they already have deals in place in terms of how many wafers they can get. Ramping up production could be simply too costly for them or impossible , remember AMD/Nvidia aren't making more per card , they are just selling more of them than usual.
Posted on Reply
#155
justanother gamer
running a R 7 1800x with a R9 270x (pulled from Old system which is still running a I930 :) just waiting for AMD to put the vega 64 liq cooled on the market. Many cant afford the cards even the low end but the 700 is still a good deal, but hey thats their delema. ive never data mined but may look into it, just wondering do data miners have their own secret hand shake :))
Posted on Reply
#156
cdawall
where the hell are my stars
Vya DomusI am still highly doubtful of that , like I previously said they already have deals in place in terms of how many wafers they can get. Ramping up production could be simply too costly for them or impossible , remember AMD/Nvidia aren't making more per card , they are just selling more of them than usual.
Actually this is completely wrong. AMD/Nvidia is making more per wafer since they can now sell "damaged" cards to miners as mining cards. So per card cost is down increasing per card profit.
Posted on Reply
#157
notb
R-T-BThat's complete BS. Eventually, hardware acquisition cost exceeds proft margin, even for miners, as the amount of possible profit remains fixed.
(...)
More mining doesn't help him though, it hurts him, which pretty much shoots a hole in any benefit argument you can make against him.
These arguments are incorrect. Possible profit (in a period of time) is not fixed. Supply of crypto is infinite and demand is still growing.
The big unknown is whether central banks and financial institutions choose to utilize open currencies (BTC, ETH or something) or make fully controlled new ones.
Posted on Reply
#158
Vya Domus
cdawallActually this is completely wrong. AMD/Nvidia is making more per wafer since they can now sell "damaged" cards to miners as mining cards. So per card cost is down increasing per card profit.
What do you mean damaged ? They seem to be fully enabled chips from what I see meaning they come from the same supply of wafers that the consumer variants come from.
Posted on Reply
#159
AsRock
TPU addict
RejZoRIt's almost as if AMD doesn't really care about gamers. Everyone should refuse buying second hand cards. That's f**k miners up as they won't get any financial returns from the hardware. And refuse to buy new ones if prices get stupid. That'll force AMD to decide who they want to support. Gamers or dumb a** miners. Cancerous mining nonsense...
Could not agree more but.
erockerMaybe AMD needs to consider a separate SKU and/or driver for mining cards.
Yeah but then people be bitching that it was not done with the other cards as seen as there was no hardware changes required.

AMD is in a Catch 22, dammed if they do and dammed if they don't.

As seen as the AMD cards are just as exspenive as the high end nVidia cards i may as well go for one of those now even more so with Arma being my main game still.

AMD should higher prices so these company's are grabbing all the profits for them self's, when really it should be AMD getting the money.
Posted on Reply
#160
R-T-B
notbThese arguments are incorrect. Possible profit (in a period of time) is not fixed. Supply of crypto is infinite and demand is still growing.
No, because everyone exchanges via BTC. BTC has a fixed, decreasing release rate and market cap is not infinitely growing. It would reach a point of equilibrium.
Posted on Reply
#161
hat
Enthusiast
@Captain_Tom
I paid almost $900 for two 1070s, and have so far earned roughly $80 by mining
Posted on Reply
#162
cdawall
where the hell are my stars
Vya DomusWhat do you mean damaged ? They seem to be fully enabled chips from what I see meaning they come from the same supply of wafers that the consumer variants come from.
570D that a good chunk of the AMD stuff is based off of is a further disabled Polaris die, the 1060/1080's have outputs removed on the vast majority the clocks on some are also dropped well under reference for the OE dies, no TDP listings etc. These will be bad bin chips, 30 day warranty means basically DOA only warranty. You don't think they took the highest end chips and resold them to miners at lower cost do you?
Posted on Reply
#163
Vya Domus
cdawall570D that a good chunk of the AMD stuff is based off of is a further disabled Polaris die, the 1060/1080's have outputs removed on the vast majority the clocks on some are also dropped well under reference for the OE dies, no TDP listings etc. These will be bad bin chips, 30 day warranty means basically DOA only warranty. You don't think they took the highest end chips and resold them to miners at lower cost do you?
And most chips that are still functional but don't meet the voltage/clocks requirements are far in between. These can be returned in the 30 days window along side the DOA stuff too. Yes they can resell some chips that they couldn't before but it's still not enough to make them clap their hands in excitement nor does it reduces the per wafer cost greatly on their end. The money they pay to fabs and the supply is still the same and it's still a limiting factor.

Point is , if they really wanted to profit from this situation in a meaningful way they would need completely different deals with the fabs and much higher MSRPs from the get go.
Posted on Reply
#164
notb
R-T-BNo, because everyone exchanges via BTC. BTC has a fixed, decreasing release rate and market cap is not infinitely growing. It would reach a point of equilibrium.
"via" - not "to"

Also, you are aware of the fact than when e.g. nicehash delays your payment (even for weeks), they're not keeping those BTC on their books, right? :-)
Posted on Reply
#165
R-T-B
notb"via" - not "to"
No, via AND to. You switch to BTC and cash out. You STILL exchange to BTC prior to sale. Everyone does.
Also, you are aware of the fact than when e.g. nicehash delays your payment (even for weeks), they're not keeping those BTC on their books, right? :)
I have no idea, I have never used nicehash. The last pool I mined on was btcguild. It certainly kept books.
Posted on Reply
#166
notb
R-T-BNo, via AND to. You switch to BTC and cash out. You STILL exchange to BTC prior to sale. Everyone does.
OK, so you're not using a mining pool now, but you know how they work (at least from the user/practical point of view).
So lets say you're mining ETH or ZEC or something, but you "switch to BTC" (whatever that means) and cash out as soon as possible - that's the general workflow.

So here are 2 questions for you:
1) How many wallets and in which cryptos do you need to do that?
2) Assuming you're mining at day 0, how will crypto transfers look between your wallets?

I believe that analyzing this scenario will tell us a lot about what's actually happening. :)

And BTW: I've just started mining! :-D
Posted on Reply
#167
R-T-B
notbOK, so you're not using a mining pool now, but you know how they work (at least from the user/practical point of view).
So lets say you're mining ETH or ZEC or something, but you "switch to BTC" (whatever that means) and cash out as soon as possible - that's the general workflow.

So here are 2 questions for you:
1) How many wallets and in which cryptos do you need to do that?
2) Assuming you're mining at day 0, how will crypto transfers look between your wallets?

I believe that analyzing this scenario will tell us a lot about what's actually happening. :)

And BTW: I've just started mining! :-D
There are many ways to answer that question, but generally you need a wallet for every currency type involved. Usually users today trust the exchange to manage the books there but I wouldn't, personally. Maybe I'm old fashioned, but either way I promise you books are kept and every exchange to btc noted.

Good luck mining. I'd have ran like hell if I were you. Investing is far easier and even it is stressful as heck.
Posted on Reply
#168
cdawall
where the hell are my stars
Vya DomusAnd most chips that are still functional but don't meet the voltage/clocks requirements are far in between. These can be returned in the 30 days window along side the DOA stuff too. Yes they can resell some chips that they couldn't before but it's still not enough to make them clap their hands in excitement nor does it reduces the per wafer cost greatly on their end. The money they pay to fabs and the supply is still the same and it's still a limiting factor.

Point is , if they really wanted to profit from this situation in a meaningful way they would need completely different deals with the fabs and much higher MSRPs from the get go.
More cards is more cards, less cost is more profit. You also don't know yields so how can you say if this made a debt or not? Do you work for tsmc/glofo?
Posted on Reply
#169
Prince Valiant
the54thvoidThis should just depress poeple:

www.tweaktown.com/articles/8229/ethereum-mining-1gh-40-gpus-5000-per-month/index.html


To be fair, his mining set up required shit loads of devotion and loads of cash but still, to me it's just another blight on our world of quick cash grabs at the expense of others.

Stop mining and do some folding. You might get some cash back but it wont cure your cancer.
The box stacking in that picture is horrendous.
cdawallActually this is completely wrong. AMD/Nvidia is making more per wafer since they can now sell "damaged" cards to miners as mining cards. So per card cost is down increasing per card profit.
Any proof of this happening or are you speculating?
Posted on Reply
#170
R-T-B
FordGT90ConceptI was hoping to find out details on the power consumption but doesn't look he did it. His bill was probably $1000/mo. If this keeps up, I wouldn't be surprised if power companies add surcharges for people that surge in power consumption.
Not a new idea by any means. Mines been doing this for literally DECADES before computers existed, let alone miners.
Posted on Reply
#171
cdawall
where the hell are my stars
Prince ValiantAny proof of this happening or are you speculating?
RX570D/RX470D feature 1792 shaders vs 2048.

P106/P104 based nvidia products lacking a video output means artifacting doesn't matter as long as they don't crash the driver.
Posted on Reply
#172
Vya Domus
cdawallDo you work for tsmc/glofo?
Then your guess is as good as mine. I'm just basing my assumptions on how it typically is.
Posted on Reply
#173
cdawall
where the hell are my stars
Vya DomusThen your guess is as good as mine. I'm just basing my assumptions on how it typically is.
How what typically is, we have never in the history of gpu production had such a huge run on equipment.
Posted on Reply
#174
Vya Domus
That's the thing , this situation is occurring in a time when the manufacturing of these devices is still driven by the "old rules".

Your saying AMD/Nvidia managed to adapt in no time and make huge gains in terms of profitability in an industry where deals are made years before production even begins. Although they are certainly not losing cash , they are not in a perfect situation from which they can profit at full capacity either.
Posted on Reply
#175
cdawall
where the hell are my stars
Vya DomusThat's the thing , this situation is occurring in a time when the manufacturing of these devices is still driven by the "old rules".

Your saying AMD/Nvidia managed to adapt in no time and make huge gains in profitability in an industry where deals are made years before production even begins. Although they are certainly not losing cash , they are not in a perfect situation from which they can profit at full capacity either.
I never said huge gains in any post. I said they gained additional funds by selling lesser cards for a profit.

Number of cards sold is up quite a bit however. Which is very interesting considering it was down q1 of this year. The only reports really showing this are AMD's q2 results.
Posted on Reply
Add your own comment
Dec 23rd, 2024 04:26 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts