Crazy Fire Sale on 290x's

P

pingjockey

Limp Gawd
Joined
Nov 14, 2006
Messages
177
Since the price cut there has been a glut of used 290x's both in a oem cooling and aftermarket cooling set-up. Has anyone had luck buying any of these cards like this?

My concern is that these cards have been used for mining 24x7 and I really don't want to buy something that is going to die in a few months.

I have seen a bunch of Sapphire TRI-X on the various auction sites....
 
Ebay looks like it's around $250.... that's not low enough in my opinion. Not when a brand new 970 is $330. And like you said, you don't know where they have been or what they were used for.
 
Just get one that does RMA's via serial number verification.

Whether a card was flogged to death as a miner or just used for typical gaming it could still go at any moment. Statistically speaking though if they don't die in the first week they tend to last years before going on you.
 
They were less than $200 a couple of days ago. Just wait a few days and check often.
 
I picked up a used reference R9 290x a month back for $250, that was used for mining and it's running flawlessly. I'm thinking of picking up a second one, but it might be overkill @ 1080p :D
 
Has there been any proof that running these 6-12 months 24/7 at max speeds degrades them when cooled properly ?
I can't believe GPU's have a limited life-span that suddenly is reached by scrypt mining when Folding@Home has been possible on the GPU for such a long time.

I say, buy these as long as it has atleast a month of warranty on it and get one with proper cooling, not the garbage reference cooler.
 
Just trying to decide if it's worth the effort to get a 290x to replace a gtx 770 or not. I only play at 1080p but some of the titles I play could really benefit a 4 gb frame buffer.
 
I'm trying to decide if I want to replace my sli 670 setup with 2 of these. Would a Seasonic 750 gold be able to handle 290x crossfire or would I need to upgrade that as well? Which brands carry serial based warranties?
 
pingjockey said:
Just trying to decide if it's worth the effort to get a 290x to replace a gtx 770 or not. I only play at 1080p but some of the titles I play could really benefit a 4 gb frame buffer.
Click to expand...

Although I think you would be incredibly happy with an R9 290x, I also think you would be very pleased with the 970 as well. However, I would just stick with the 770GTX you have already at least until the dust settles.
 
Phuncz said:
Has there been any proof that running these 6-12 months 24/7 at max speeds degrades them when cooled properly ?
Click to expand...
No but there is little in the way of reliable public data on the longevity of these cards at any work load.

Best information that I know of on the subject:


From here:
http://www.pugetsystems.com/labs/articles/Video-Card-Failure-Rates-by-Generation-563/

Personally I've crunched GPGPU projects like MilkyWay for years on my AMD cards and none of them died with in a year or out of the box on me. I've had 2 failures (1 4890 and 1 7970, both XFX models that were fairly cheap for their time but had lots of complaints of heat and high failure rates) of AMD video cards. The 4890 died suddenly on me a little after 2yr of crunching. The 7970 started to go and do weird stuff after a year and a half or so. My other 4890 works fine to this day and was only recently retired as a back up video card. But that is just my personal experience and anecdotal info. isn't worth a whole lot vs the above links which are from a retailer.

Looking at the graph if your R9 290/x doesn't die out of the box it tends to be a fairly reliable card.
 
Yeah, I have heard the same on other sites that if a card is going to die it will do so fairly quickly out of the box. I think I am going to keep watch on ebay and see what happens over the next few days with prices.
 
Briguyatx said:
I'm trying to decide if I want to replace my sli 670 setup with 2 of these. Would a Seasonic 750 gold be able to handle 290x crossfire or would I need to upgrade that as well? Which brands carry serial based warranties?
Click to expand...

Do you plan on OC'ing the hell out of the cards? If not, it should be fine, but my sig rig with my 290 (1.275Ghz core/1.375Ghz mem) and cpu loaded 100% pulls over 670W..Granted this is with a WC pump running full speed (~15W) 4 GT AP-15 120mm fans and 2 200mm fans on my front rad..I am getting ready to add another card to my setup and am hoping I can still clock both to 1.1~1.2Ghz without my X-850W Gold screaming like a 3 year old in the candy isle!:p
 
Ooops didn't notice this until the above poster quoted it, sorry.

Briguyatx said:
Would a Seasonic 750 gold be able to handle 290x crossfire or would I need to upgrade that as well? Which brands carry serial based warranties?
Click to expand...
Your PSU should be OK. Going for CF'd 290's would give you some more leeway power wise, be cheaper, and get nearly the same performance though.

Gigabyte, MSI, and I believe ASUS all RMA through the serial.
 
I just bought one of these 290x's to replace my loyal 4 year-old 6950 (modded to 6970). The new card is only a few months old and it looks like it was used in a professional data center, so i figured I role the dice. Paid $240. I'll let you know how it goes when it gets here later this week.
 
vikingod said:
I just bought one of these 290x's to replace my loyal 4 year-old 6950 (modded to 6970). The new card is only a few months old and it looks like it was used in a professional data center, so i figured I role the dice. Paid $240. I'll let you know how it goes when it gets here later this week.
Click to expand...

You shouldn't have after I just read this from that ebayer
“Used for five months or less in a professional datacenter as part of custom Litecoin/Scrypt mining ”
Click to expand...

Those cards are run hard 24/7 for bitcoin. Newegg has the HIS R9 290X new for 359 with a $20 rebate card. Would be better to save a little longer and know what you are getting or wait for the next price drop as it seems these are dropping faster and faster.
 
SnowBeast said:
You shouldn't have after I just read this from that ebayer

Those cards are run hard 24/7 for bitcoin. Newegg has the HIS R9 290X new for 359 with a $20 rebate card. Would be better to save a little longer and know what you are getting or wait for the next price drop as it seems these are dropping faster and faster.
Click to expand...

What ebayer? He did not provide a link for the one that he had purchased. Also, he could have bought it through Amazon which is what I did recently with an Amazon Fulfilled vendor. If he bought a Gigabyte one, it would still be under warranty as well.

I understand what you are getting at but, it depends on where he bought it from and what Vendor the card is from.
 
SnowBeast said:
You shouldn't have after I just read this from that ebayer

Those cards are run hard 24/7 for bitcoin. Newegg has the HIS R9 290X new for 359 with a $20 rebate card. Would be better to save a little longer and know what you are getting or wait for the next price drop as it seems these are dropping faster and faster.
Click to expand...

Cards are either operated within their thermal ranges and work fine, or they are forced not to and they fail. You have an outlier that are defective out of the box. I would happily save a large percentage of money on a card that was mined with, especially one out of a professional setup.
 
SilverSliver said:
Cards are either operated within their thermal ranges and work fine, or they are forced not to and they fail. You have an outlier that are defective out of the box. I would happily save a large percentage of money on a card that was mined with, especially one out of a professional setup.
Click to expand...

That's exactly my rationale for the purchase. Honestly, getting anything used is a crap shoot. I'd rather buy something that appears to be well cared for in a data center than something stuck under a dusty desk. Aside from the fan (which I expect to replace), I can't see what would fail that wouldn't have already gone bad in the first few weeks of ownership.
 
mesyn191 said:
No but there is little in the way of reliable public data on the longevity of these cards at any work load.

Best information that I know of on the subject:


From here:
http://www.pugetsystems.com/labs/articles/Video-Card-Failure-Rates-by-Generation-563/
Click to expand...
That's just it, it's Puget's "burn-in test", not endurance testing. They only test FurMark for 5 minutes, I've even used it longer than that and I even don't care about benchmarking my own gear. So the only thing this graph shows is how much DOA cases they have and how much short-term (1-365 days) issues they have with each GPU family. But it doesn't show what happens to a GPU when it's being loaded at 100% 24/7 for 3-6 months.
 
SilverSliver said:
Cards are either operated within their thermal ranges and work fine, or they are forced not to and they fail. You have an outlier that are defective out of the box. I would happily save a large percentage of money on a card that was mined with, especially one out of a professional setup.
Click to expand...

Mining puts an unusually high load on a card. It's not unusually for the cards to consume 300-350watts when used for computing purposes. And considering its was used 24/7, the age of the card is drastically higher and so is the wear.

The best way to look at this is with a car analogy.

Say you have a car that 6 months old.

And you have another thats 3 years old with 60,000km. Which engine should be the worse for wear?

You can't tell from age alone.

Hypothetically, what if that 6 month car had been driven nonstop at 150km an hour, non stop continuously. The car would have 648,000km. Sure the car is still functioning, but there is a far greater risk something is going to go.

So similarly, if someone mined a card for 5 months 24/7, its the equivalent of someone gaming overclocked(to account for the increased power consumption), 40 months, 3 hours a day.

Yes card will run that are over 3 years old, but their chance to fail skyrockets, hence why most warranty even extended ones on computer components are 3 years.
 
tajoh111 said:
Mining puts an unusually high load on a card. It's not unusually for the cards to consume 300-350watts when used for computing purposes. And considering its was used 24/7, the age of the card is drastically higher and so is the wear.
Click to expand...

Exactly. Thank you. Gaming on these cards isn't going to kill it as its not even being 100% utilized. In mining, that card is balls to the wall full on out, every component is being maxed out. You are taking a $260 dollar gamble which IMO, is nuts when you could save a week or two and get a new one near the $300 mark.(With the way prices are falling, that could happen.)
 
ManofGod said:
What ebayer? He did not provide a link for the one that he had purchased. Also, he could have bought it through Amazon which is what I did recently with an Amazon Fulfilled vendor. If he bought a Gigabyte one, it would still be under warranty as well.

I understand what you are getting at but, it depends on where he bought it from and what Vendor the card is from.
Click to expand...

http://www.ebay.com/itm/Sapphire-AM...099?pt=PCC_Video_TV_Cards&hash=item23429a49fb

It was the first one to come up while I was looking on Ebay. Sapphire are the cheapest built ones out there, mostly reference. I didn't see any DATACENTER USED cards on Amazon at all. Sorry. Still think he should save his money an extra week.

AMD is giving out Civ 5 on top Alien Isolation and Star Citizen.

So lets rethink this. Instead of getting a new card, with 3 top tier games for free, he is getting somebody else's junk that been running full tilt for 5 months or more (You know, they could be lying about that also, how the fuck do you know?), no games and a very limited warranty? Hey you're right. Its his money. But don't come here asking "why doesn't this work, or why do I have a black screen?" ;)
 
Last edited:
As an eBay Associate, HardForum may earn from qualifying purchases.
Most miners undervolt their cards so they're not usually put thru the stress gaming does on them. Also miners usually have some elaborate fan setup to cool the cards which we as gamers don't do that.
 
They underclock the GPU itself, but actually, they OC the hell out of the memory, which is generally the first thing to go in a GPU.
 
The Mac said:
They underclock the GPU itself, but actually, they OC the hell out of the memory, which is generally the first thing to go in a GPU.
Click to expand...

Another fine point. Which is really saturating the 512bit bus 24/7.

Hey no argument here. You want to chance it, be my guest, its your business. Just IMO, it really doesn't make sense to me.
 
Phuncz said:
That's just it, it's Puget's "burn-in test", not endurance testing. ...But it doesn't show what happens to a GPU when it's being loaded at 100% 24/7 for 3-6 months.
Click to expand...
Like I said its the best info. I can find. There is basically nothing else out there on the subject that is public that I know of so if you've got better post it.

You should note though that it does include failures 'in the field' some of which will be from miners. And those rates aren't bad at all.
 
SnowBeast said:
Another fine point. Which is really saturating the 512bit bus 24/7.

Hey no argument here. You want to chance it, be my guest, its your business. Just IMO, it really doesn't make sense to me.
Click to expand...

I don't buy used cars that were driven over 60mph so, makes sense.
 
mesyn191 said:
Like I said its the best info. I can find. There is basically nothing else out there on the subject that is public that I know of so if you've got better post it.

You should note though that it does include failures 'in the field' some of which will be from miners. And those rates aren't bad at all.
Click to expand...
I'm not saying you shouldn't post info but it just isn't relevant to the the topic.
No one buys a Puget system to mine, they buy one to game.

Next to the hear-say and conjecture, does anyone have any tangible data ? Because otherwise you are just finding reasons to justify it to yourself.
 
Being that its the best and about the only tangible data out there on the subject it sure seems highly relevant to the topic. Just because it doesn't have the failure rates of miner only cards doesn't make it useless or conjecture.

Lots of gamers did mining too with their systems. Plenty of people here on the forums for a while would mine while they were at work or asleep on their gaming systems up until things got crazy and the difficulty got to high to make it worthwhile.
 
It's just not "the best and only tangible data" because it doesn't have the right criteria.

Firstly, there is no mention what the problems were. So it could be core, VRM, memory, coil whine, fan, manufacturing defect, software issue, user error, whatnot.

Secondly, a few miners out of the mass of gamers is not going to give you an accurate statistic. It might have shown that 0% or 100% would have failed, without knowing how many people mined with the cards, the statistic says nothing about mining.

Thirdly, it doesn't differentiate between GPU models, just series. So without knowing which type of memory is used for instance, we can't extract if it would be a mayor factor in mining reliability (I would believe it is).

I'm actually in the same camp as you, thinking mining has little influence on the card's practical lifetime (3-5 years). But I won't just take any unrelated info as evidence. You can't either, if you are rational and neutral.
 
I wonder if the folding guys have seen 'premature' component death. Some of those rigs have been 100%, 24.7 for years now.
 
Lots of speculation and anecdotal evidence flying around here. Does anyone here actually work at a coin mine? I find it hard to believe it makes any sense for a professional organization to abuse their (extremely expensive) equipment half as much as is being detailed here and elsewhere. Then there is the matter of it actual effect on the card itself, which there doesn't seem to be much information on at all.

So far the feedback on the lot of cards I purchased from has been very positive. I'll know more tomorrow evening. I shoud have something posted here this coming weekend after I've had a chance to put it through it's paces. If anyone wants to see any specific tests, let me know.
 
One thing I can say about 24/7 operation is one had oc'd 1x 5870@1ghz & 1x 5830@900mhz run practically 24/7 since their release many years ago. They've done just about every DC project, btc, ltc...etc...

And with regular cleaning and greasing the fans they still run just as well as the day I got them.

Small sample size, but they work fine.
 
Yakk said:
One thing I can say about 24/7 operation is one had oc'd 1x 5870@1ghz & 1x 5830@900mhz run practically 24/7 since their release many years ago. They've done just about every DC project, btc, ltc...etc...

And with regular cleaning and greasing the fans they still run just as well as the day I got them.

Small sample size, but they work fine.
Click to expand...

Precisely. Dogging on someones decision to save 30-50% on a card that was mined with in a datacenter because they 'feel' like the card was somehow 'used up' is silliness. If they want to claim the usable life span of a component is reduced by usage, prove it.
 
Phuncz said:
It's just not "the best and only tangible data" because it doesn't have the right criteria.
Click to expand...
Knowing what the exact cause of the RMA would be nice but isn't really necessary since the chart gives you a 'ground floor' of RMA's both in the field and out of the box.

Knowing the models would be nice too but isn't that big of a deal either since manufacturers use mostly the same reference design and components. Therefore its reasonable to assume the model to model variance will not be large and possibly not even significant.

The only real argument against the chart data being relevant is that it doesn't give information on cards that have been mined non stop. However not all cards for sale on ebay or these or other forums have been mined with though. Most of the miners have already sold their cards and moved on to dedicated ASICS a couple months ago due to the difficulty increase of hashing. That was the first firesale situation which dropped the ebay price to $250-300-ish and this most recent one is due to people offloading their 290/x's to buy nV 970/980's which has dropped the price to around $200.

Phuncz said:
But I won't just take any unrelated info as evidence. You can't either, if you are rational and neutral.
Click to expand...
Evidence that isn't ideal is still evidence that can be useful. You have to take what you can get sometimes because the alternative is you're stuck with nothing at all.
 
SilverSliver said:
Precisely. Dogging on someones decision to save 30-50% on a card that was mined with in a datacenter because they 'feel' like the card was somehow 'used up' is silliness. If they want to claim the usable life span of a component is reduced by usage, prove it.
Click to expand...

I tend to agree.

You could use the same "car analogy" to argue for only buying mined cards when buying used. Would you rather buy a car that was constantly used and maintained in a professional environment or one that was allowed to heat up cool down, heat up, cool down, etc. usually without any maintenance in a dust filled case?

Personally, I've bought a used 290X and 7950 for my two brother's computers that were used for mining and they both work perfectly fine.
 
Looks like big price drops on reference R9 290X models but not the the non reference ones.

Lightning or DirectCu versions are still all above 400 buckaroos, too much for my taste.
 
You can get 290 XFX models with better HSF's for $250-ish. They also RMA through the serial. None of the 290X models are worth the price premium IMO.
 
You must log in or register to reply here.
Back
Top