GPU market up—Intel and Nvidia graphics winners in Q4, AMD down

[jh30uk]

You are showing your age now.


Is that Johnny Rotten?







Naa I've only used it twice

 

[Tonester0011]

Even the best capacitors, such as Japanese 10K black capacitors (used in high end motherboards), have a lifetime of 10000 hours when operating at full capacity, that's just over a year.

That's to meet it's required specification and also running at maximum temperature, just because you've ran it past 10k hours doesn't mean they'll die at 10001

Running components 24/7 will limit the life, that's for sure but not by a significant amount unless they've been running with terrible airflow (granted the chances of terrible airflow is increased in the mining scene especially with DIY miners who aint a clue)



(Worth noting i still wouldn't touch an ex miner unless the price was well below market )

 

[humbug]

That's probably correct, I don't understand how people can claim that running a graphics card at full load 24/7 will not cause quicker degradation of the GPU. Even the best capacitors, such as Japanese 10K black capacitors (used in high end motherboards), have a lifetime of 10000 hours when operating at full capacity, that's just over a year.

The components used in high-end graphics cards will be about the same quality as this - with some components probably having a lower quality, so when they finally give up the ghost you're looking at a GPU with a lifetime of less than a year when mining. Of course there will be exceptions as always with the silicon lottery, but that will be the general "trend".

So, no, I wouldn't be happy to receive a graphics card from someone whom used it for 24/7 mining for a long time as I know it would have significant component degradation.

A lot wrong with that post which draws you to this conclusion.

so when they finally give up the ghost you're looking at a GPU with a lifetime of less than a year when mining.

People the world over are mining on GPU's today that they have been mining on for 3, 4, 5 years.

Its simplistic logic to dictate because a Capacitor has a life span on one year the component will last for one year.

One thing you did say which if you had realised the significance of you would not have made your above statement.

The life span on these components are measured in requirements of maximum stress usage.

Those components are never under any circumstance anywhere near that level of stress.

Modern GPU's are not 1980's Amstrad all in one HIFI systems.

The work required for a component to power another is shared between multiple componants when one or two are adiquate to do the job on their own.

This is done on modern Motherboard and GPU PCB's SO THAT no one component is EVER under any significant stress, as matter of fact designed to be under as lesser stress as possible.
The reason is what all these manufactures want to sell you about their products, reliability and above all for enthusiasts; stability. A component under stress is not stable.

My Motherboards CPU power Phase has 8 chokes. 8 set array capacitors and 8 set Mosfets, it only needs 2 to do its job.

 

[PhysicsMan94]

A lot wrong with that post which draws you to this conclusion.

People the world over are mining on GPU's today that they have been mining on for 3, 4, 5 years.

Its simplistic logic to dictate because a Capacitor has a life span on one year the component will last for one year.

One thing you did say which if you had realised the significance of you would not have made your above statement.

The life span on these components are measured in requirements of maximum stress usage.

Those components are never under any circumstance anywhere near that level of stress.

Modern GPU's are not 1980's Amstrad all in one HIFI systems.

The work required for a component to power another is shared between multiple componants when one or two are adiquate to do the job on their own.

This is done on modern Motherboard and GPU PCB's SO THAT no one component is NEVER under any significant stress, as matter of fact designed to be under as lesser stress as possible.
The reason is what all these manufactures whant to sell you about their products, reliability and above all for enthusiasts; stability. A component under stress is not stable.

Mt Motherboards CPU power Phase has 8 chokes. 8 set array capacitors and 8 set Mosfets, it only needs 2 to do its job.

Lots of hearsay there, fact of the matter is that mining crypto currencies is incredibly stressful on a GPU (heck, it's designed to be given that these algorithms try to squeeze every last bit of performance from the core) and hence the degradation of the components will happen much faster than on a GPU which is not set to mining 24/7.

As a general rule, I do not trust people - people are stupid - thus I would not be happy buying a GPU from someone which has been mining 24/7 for a considerable length of time; I can't know whether they've had adequate air flow reaching their graphics cards or not and I'd rather not risk my own money to find out.

Anyway, how did we even end up on this topic? I think we shouldn't let this thread turn into another pointless debate which will inevitably end up nowhere.

 

[triss]

Right kiss and make up and lets agree to disagree and move on guy's(girls) not guessing what sex everyone is
slightly back on track i do agree having intel's gpu's in the stats is a bit out of whack, is there any stats on dedicated gpu's

 

[humbug]

I'm not kissing him ^^^^^

Lots of hearsay there, fact of the matter is that mining crypto currencies is incredibly stressful on a GPU (heck, it's designed to be given that these algorithms try to squeeze every last bit of performance from the core) and hence the degradation of the components will happen much faster than on a GPU which is not set to mining 24/7.

As a general rule, I do not trust people - people are stupid - thus I would not be happy buying a GPU from someone which has been mining 24/7 for a considerable length of time; I can't know whether they've had adequate air flow reaching their graphics cards or not and I'd rather not risk my own money to find out.

Anyway, how did we even end up on this topic? I think we shouldn't let this thread turn into another pointless debate which will inevitably end up nowhere.

We came onto this topic because people started spouting facts about mining that could not be more wrong.

Degradation of a component is caused by heat during use, not use on its own. thats why the life span of the components you cited are measure with the caveat of heat, usually 95 or 105c depending on the quality of said component.

If people run those components at 105c then they may only last a year.

Mining does not cause those components to run at 105c, people do. components running at less than 105c will last longer, much longer, this is why there are GPU's which have been mining for 5 years, their owners know (and Understand) heat is what degrades them.

 

[melmac]

A good explanation there Mr Physics

Except it's wrong.

That 10K hours is measured at maximum ripple current and highest temp inside the cap itself. For every 10 degrees drop in temp you get 4-7 times the lifespan. Those caps he mentions are rated for 105 degrees.

So at 95 degrees, your capacitor will last 40K hours. But I doubt that they will be running at anywhere near that temp, even if the GPU is running at 95 Degrees, unless there is a problem with the capacitor itself.

 

[melmac]

That's probably correct, I don't understand how people can claim that running a graphics card at full load 24/7 will not cause quicker degradation of the GPU. Even the best capacitors, such as Japanese 10K black capacitors (used in high end motherboards), have a lifetime of 10000 hours when operating at full capacity, that's just over a year.

The components used in high-end graphics cards will be about the same quality as this - with some components probably having a lower quality, so when they finally give up the ghost you're looking at a GPU with a lifetime of less than a year when mining. Of course there will be exceptions as always with the silicon lottery, but that will be the general "trend".

So, no, I wouldn't be happy to receive a graphics card from someone whom used it for 24/7 mining for a long time as I know it would have significant component degradation.

Nobody is arguing that mining a card 24/7 is shorting the lifespan of the card, just people are totally exaggerating the effects. And you are doing it too. The example you used is so wrong. Even if you somehow managed to run a capacitor in video card to that that high of temp and ripple current for 10K hours, it won't suddenly just die. That's just a specification it has to reach. It might keep working for double or triple that time, or if it has defect it might not make 10 hours.

Your general trend is so wrong too. The general trend is that most mining cards will go on working for years after they are finished mining.

I don't understand why people think that miners are all idiots, of course there will the few, there always will, but mining isn't easy, it needs a little bit of technical know how. A GPU miner who is doing it right will definitely try to keep temps under control. They don't want to be replacing GPU's. Replacing GPU costs time and money.

And the risk is lessened even further with the thermal protection that modern GPU's have.

You are just as likely to buy a second card from a friend and have that fail as to buy an ex mining card and have that go wrong.

Why would I buy a second card that has mined over one that hasn't? simple, I believe that any card that has mined for a period of time is a good card. The ones with faults tend to fail very early on. So If I see an ad with for a card saying "mining for one year" or an ad that says "never overclocked" Which one is probably telling the truth?

Saying all that, I would never buy a second hand card full stop because there are too many risks full stop no matter if it's been mining or not. At the end of the day you don't know what has been done to any card that you buy second hand.

 

[humbug]

slightly back on track i do agree having intel's gpu's in the stats is a bit out of whack, is there any stats on dedicated gpu's

Ok. back on topic

Usually discrete AIB Figures follow shortly after the overall market figures.

 

[triss]

TY humbug

 

[andybird123]

http://www.jonpeddie.com/publications/add-in-board-report/

Nvidia still around 65% market share, up slightly from Q3

 

[Gregster]

http://www.jonpeddie.com/publications/add-in-board-report/

Nvidia still around 65% market share, up slightly from Q3

It seems the mining craze lost AMD 0.4% from quarter to quarter (35.4% to 35%). Seems strange but JPR have been doing this for years and years.

On the recent poll that was on this site, AMD is in 20% more machines than nVidia (the people who participated). That tells me that either LtMatt is an excellent shill or the UK in general prefer AMD.

 

[CAT-THE-FIFTH]

It seems JPR has conflicting reports now.

The link in the OP says the following:

AMD’s shipments of desktop APUs (heterogeneous GPU/CPUs) jumped 15% from the previous quarter but declined 26.7% in notebooks. AMD’s discrete desktop shipments increased 1.8%, and notebook discrete shipments declined 6.7%. The company’s overall PC graphics shipments decreased 10.4%. Notebook build cycles are specific, and AMD was late with its new parts.

So,AMD sold more desktop cards then last quarter.

Honestly its time we had actual numbers of cards shipped instead of percentages,and also the breakdown of cards sold.

Plus AMD APU shipments on desktop jumped 15% meaning they sold less non-APU systems. Those non-APU systems would have shipped with low end cards.

AMD APU sales are cannibalising low end AMD graphics cards sales.

This has been happening for a while now.

 

[Gregster]

Yep Cat, that doesn't coincide with each other.

 

[CAT-THE-FIFTH]

I think at least on desktop there is positive news for both. Nvidia shipped more cards,and AMD shipped more cards and APUs.

I think the major worry is in laptop for AMD and that is because Kaveri was delayed and is still not in laptops. The improving AMD APUs are eating up low end cards sales though.

Sadly,AMD is using the second best company in the world* for making its CPUs,ie,Global Foundries,but they are always massively delayed when compared to Intel.

*IBM technically would be second best,but the 22NM node they developed is lower volume and very expensive and is meant for their POWER chips.

 

[TAZOO7]

What is it you know m8?

Yes, but only because people seem to do well stupid things, like put 4 cards all right next to each other, with zero airflow, they don't try and configure the cards or undervolt them, they simply mine on them at stock speeds, with no fan profile until they die. We've seen a lot of burnt out cards which we've also seen from gaming crossfire machines too. The returns are reducing as more and more people are getting smarter and being educated, to ensure the cards have airflow and to setup fan profiles, lower voltages etc. etc.


P.S. Got a 7970 which has seen similar use since launch as well, again it works without issue and its being battered 24/7 without issue.

Gibbo, maybe when you get the time you should add a mining hardware guide to put on the website with tips and practices and do's and don'ts on how to set up the hardware correctly before they even start learning about mining, maybe help further to reduce RMA's

This Amd vs Nivida thing is so boring, i have had my fair share of both and not got anything bad to say about any of them other than about the price, AMD and Nvidia both doing ok, both have there own agenda,s both have other aspects of there business in which the make money from other than just GPU's for the enthusiast, i have just gone from a 780ti to a R290 and im equally as happy if not happier given the £200 price difference i paid between the two cards, short version both make good products, both make money over the year

 

[PhysicsMan94]

Except it's wrong.

That 10K hours is measured at maximum ripple current and highest temp inside the cap itself. For every 10 degrees drop in temp you get 4-7 times the lifespan. Those caps he mentions are rated for 105 degrees.

So at 95 degrees, your capacitor will last 40K hours. But I doubt that they will be running at anywhere near that temp, even if the GPU is running at 95 Degrees, unless there is a problem with the capacitor itself.

I feel that I made it quite clear that the figure of 10K hours was in reference to operating at full capacity, perhaps I should've explicitly stated that I meant that this was at maximum operating temperature - and given that a lot of cryptocurrency miners have whole arrays of graphics cards bundled together it is just a fact that the temperature of each GPU is going to be extremely high if they do not have adequate cooling in place (as I said before, I don't trust people on the whole and thus I would not be comfortable buying a second hard graphics card from someone who has been mining as I have no idea whether they have implemented good enough cooling in their systems or not).

But of course, do not get me wrong, I accept the fact that people are the major trouble in this equation (I even emphasised that myself), it's just the mining side of things which would make me even less accepting of buying a second hand GPU in this regard - as I'd know that the user was running it 24/7 at full usage, and as I mentioned before I have no idea whether they have cooled the card properly or not.

Also, out of curiosity, could you give me your source for those scaling figures relating temperature changes to corresponding lifetimes?

 

[LtMatt]

This Amd vs Nivida thing is so boring, i have had my fair share of both and not got anything bad to say about any of them other than about the price, AMD and Nvidia both doing ok, both have there own agenda,s both have other aspects of there business in which the make money from other than just GPU's for the enthusiast, i have just gone from a 780ti to a R290 and im equally as happy if not happier given the £200 price difference i paid between the two cards, short version both make good products, both make money over the year

Agreed.

 

[andybird123]

It seems JPR has conflicting reports now.

The link in the OP says the following:



So,AMD sold more desktop cards then last quarter.

Honestly its time we had actual numbers of cards shipped instead of percentages,and also the breakdown of cards sold.

Plus AMD APU shipments on desktop jumped 15% meaning they sold less non-APU systems. Those non-APU systems would have shipped with low end cards.

AMD APU sales are cannibalising low end AMD graphics cards sales.

This has been happening for a while now.

the gain of 1.8% in the first report is year on year, the loss in the AIB report is qtr to qtr
it isn't explicit in the paragraph you quoted, but if you read that entire section they talk about year to year in the market watch report, where as everything is from last quarter in the AIB report

I do agree though, it would be nice to have actual numbers (though that is what they want you to pay thousands of dollars for)

 

[PhysicsMan94]

Depends how much longer this mining craze lasts, but if Nvidia do start breaking into the market with low power cards which have the potential to offer great performance-per-watt then AMD could start losing a fair amount of business in the crypto currency mining industry. We shall see I guess...