Jump to content

The GPU shortage is over


You are about to reply to a thread that has been inactive for 666 days.

Please take a moment to consider if this thread is worth bumping.

Recommended Posts

The GPU shortage is over. Basically, because crypto mining hardware purchases are way down.

Prices are now running slightly below the manufacturers' suggested retail price.

There are also lots of GPUs for sale on eBay, but many of them have had most of their useful life used up running 24/7 in some crypto mining operation. If you go that route, be prepared to run tests and make sure you can return the thing.

Most of what's available is overkill for current SL, but the viewers are getting new features that will use more GPU capability.

  • Like 5
  • Thanks 1
  • Haha 1
Link to comment
Share on other sites

Interesting that you talk about 'useful life', I wasn't aware that electronics got old by use.  Can you point to info on that?

I sort of thought that computers got slower because the software got more complex...

Link to comment
Share on other sites

11 minutes ago, Anna Nova said:

Interesting that you talk about 'useful life', I wasn't aware that electronics got old by use.  Can you point to info on that?

I sort of thought that computers got slower because the software got more complex...

Discussion for electrical engineers. Below 18nm IC line width, a new wear-out phenomenon appears. NVidia 1060 GPUs were at 16nm. NVidia 3xxx GPUs are at 7-8 nm. 3-5nm in the next few years. Becoming a major problem, especially in automotive. This is the price paid for smaller transistors.

Discussion for gamers. Overclocking bad. Fan bearings wear out. Thermal paste between heat sink and GPU matters. Fans matter a lot. It's use time that matters, not just time sitting around.

Current NVidia warranty period: 3 years.

 

 

 

  • Like 4
  • Thanks 2
Link to comment
Share on other sites

9 hours ago, animats said:

The GPU shortage is over. Basically, because crypto mining hardware purchases are way down.

Prices are now running slightly below the manufacturers' suggested retail price.

There are also lots of GPUs for sale on eBay, but many of them have had most of their useful life used up running 24/7 in some crypto mining operation. If you go that route, be prepared to run tests and make sure you can return the thing.

Most of what's available is overkill for current SL, but the viewers are getting new features that will use more GPU capability.

I checked that link to the article you posted. Has some interesting points.

I also checked the examples (Best Buy) the author listed and it raised a question I have always wondered.

I would like to upgrade to a newer GPU. The one I have now is an Nvidia Geforce card I was given and have been using since 2016. Nvidia is still giving me updates. Sadly, I'm unable to give you the exact card model as I'm on my lappy now and the card is in my desktop (maybe post here an update later when I get on the desk).

I had the stock power unit removed (180w) and a 450w unit installed. Am hoping to get an I-7 processor (I-3 now) sometime. The desk is an HP and can be expanded. My question is how set in concrete are the recommended power supply ratings?

If i purchase a new card that has a recommended power of 500w, would using a 450w power supply make much of a difference?

Edited by Schatzi Timmerman
correcting manufacturer name.
  • Like 1
Link to comment
Share on other sites

https://www.tomshardware.com/news/graphics-card-prices-july-2022

Prices are coming down, big time. The shortage is over in the used market at least. They are going for half of MSRP used when they were above MSRP used. Once we properly enter the recession prices are going to tank hard, people can't afford fuel and food they aren't going to spend $1000 on a new graphics card. They are going to drop further when AMD releases their new graphics cards this fall, they're supposed to be a huge upgrade (Nvidia's will be too) since COVID threw a wrench into product releases and both companies basically skipped a generation.

And yes, smaller nodes (nms) are more fragile and not as strong under use. I had an FX 8350, got it to boot at some insane voltages and made it into Windows long enough at 5.3ghz long enough to get a screenshot. My Ryzen, I can barely overclock. Intel has been getting bad too as it gets smaller. At least the smart miners undervolt and underclock for better efficiency, the smarter ones aren't the ones who are pushing their cards to the max, better to mine with two undervolted and underclocked cards that one overclocked and overvolted (power consumption goes up exponentially when you overclock). I had a 7970 (28nm) liquid cooled, I just set basically everything to the max clocks and voltages and used it for 8 years before it died. Wouldn't think about that on a new graphics card, they don't overclock as well and they die quicker. I remember when Intel went to 14nm people who would overclock their higher nm CPUs were killing 14nm ivy bridges like crazy. I actually used to be a huge overclocker and did it for fun, but new CPUs/GPUs are death to overclocking, I don't even really do it anymore, it's not as fun as it used to be and they all die.

  • Like 1
  • Thanks 1
Link to comment
Share on other sites

8 hours ago, animats said:

Discussion for electrical engineers. Below 18nm IC line width, a new wear-out phenomenon appears. NVidia 1060 GPUs were at 16nm. NVidia 3xxx GPUs are at 7-8 nm. 3-5nm in the next few years. Becoming a major problem, especially in automotive. This is the price paid for smaller transistors.

Transistor aging isn't just limited to smaller node sizes, although in older tech it takes far longer than the devices usable life to present a problem .. much to the chagrin of retro computing enthusiasts.

 

6 hours ago, Schatzi Timmerman said:

If i purchase a new card that has a recommended power of 500w, would using a 450w power supply make much of a difference?

You're thinking about power supplies backwards. A 450W PSU doesn't push 450W down the wires, it has the potential to supply 450W when stuff demands it.

 

A power supplies rated output represents a maximum potential rather than a constant output. Think of it in terms of peak availability, you have the potential to supply 450W but your system might not use even a fraction of that most of the time. But for the brief moments when it does need 450W, the power supply can deliver.

Getting a lower wattage power supply creates a situation where it's run at or above it's rated potential output which will dramatically shorten it's lifespan and cause system instability.

You also need to factor in power usage spikes. Sometimes a component that's rated for an average of 200W might try to pull 300W (or more) for a very brief moment, if your power supply doesn't have that spare capacity on hand then then system will glitch, crash, blue screen or just hard reset on you.

Always buy a much bigger power supply than you need. Overkill is your friend and will save you money in the long run.

  • Like 2
  • Thanks 2
Link to comment
Share on other sites

1 hour ago, Flea Yatsenko said:

https://www.tomshardware.com/news/graphics-card-prices-july-2022

Prices are coming down, big time. The shortage is over in the used market at least. They are going for half of MSRP used when they were above MSRP used. Once we properly enter the recession prices are going to tank hard, people can't afford fuel and food they aren't going to spend $1000 on a new graphics card. They are going to drop further when AMD releases their new graphics cards this fall, they're supposed to be a huge upgrade (Nvidia's will be too) since COVID threw a wrench into product releases and both companies basically skipped a generation.

And yes, smaller nodes (nms) are more fragile and not as strong under use. I had an FX 8350, got it to boot at some insane voltages and made it into Windows long enough at 5.3ghz long enough to get a screenshot. My Ryzen, I can barely overclock. Intel has been getting bad too as it gets smaller. At least the smart miners undervolt and underclock for better efficiency, the smarter ones aren't the ones who are pushing their cards to the max, better to mine with two undervolted and underclocked cards that one overclocked and overvolted (power consumption goes up exponentially when you overclock). I had a 7970 (28nm) liquid cooled, I just set basically everything to the max clocks and voltages and used it for 8 years before it died. Wouldn't think about that on a new graphics card, they don't overclock as well and they die quicker. I remember when Intel went to 14nm people who would overclock their higher nm CPUs were killing 14nm ivy bridges like crazy. I actually used to be a huge overclocker and did it for fun, but new CPUs/GPUs are death to overclocking, I don't even really do it anymore, it's not as fun as it used to be and they all die.

I have no idea which generation of Ryzen you are using but 'overclocking' in the traditional sense is different with these cpus . They are pretty much maxed out of the box unless its a low end Ryzen. They respond very well to 2 things, good cooling, obviously, and undervolting.  It's a process but once you find the sweet spot for undervolting and the PBO settings in bios, they are truly amazing chips.  (using a 5600x myself)

Link to comment
Share on other sites

22 minutes ago, Modulated said:

They respond very well to 2 things, good cooling, obviously, and undervolting.  It's a process but once you find the sweet spot for undervolting and the PBO settings in bios, they are truly amazing chips.  (using a 5600x myself)

Three things - FAST RAM

Although most of the benefit can be obtained by making sure that the rams built in XMP profile is enabled in the bios (and it's usually not by default .. go check!)

  • Like 1
  • Thanks 1
Link to comment
Share on other sites

4 minutes ago, Coffee Pancake said:

Three things - FAST RAM

Although most of the benefit can be obtained by making sure that the rams built in XMP profile is enabled in the bios (and it's usually not by default .. go check!)

Was thinking about that AFTER I posted , fo sho!

  • Like 1
Link to comment
Share on other sites

9 hours ago, animats said:

Discussion for electrical engineers. Below 18nm IC line width, a new wear-out phenomenon appears. NVidia 1060 GPUs were at 16nm. NVidia 3xxx GPUs are at 7-8 nm. 3-5nm in the next few years. Becoming a major problem, especially in automotive. This is the price paid for smaller transistors.

Hmmm? So I need to plan to replace my nVidia 1070 soonish then.  Scary.

Link to comment
Share on other sites

There was a British engineer in the 1940's, Tommy Flowers, who discovered that if you got a Thermionic Valve (I think there is an American word, but I forget it) Computer working, and never turn it off (and on), it kept going almost forever.   Valve's were replaced by transistors in the 1950/60s, because valves were 'unreliable' (Flower's work was subject to Official Secrets Act).  That's Tommy Flowers who built the first stored program computer, and didn't get the publicity that Turing, and of course the Americans, did.

Maybe we need to stop power cycling the little buggers.

  • Like 1
Link to comment
Share on other sites

40 minutes ago, Anna Nova said:

Hmmm? So I need to plan to replace my nVidia 1070 soonish then.  Scary.

Depends. If you do at least some basic maintenance (like dusting, thermal paste replacement, working cooling parts) the stuff can last for years, unless you're constantly running the PC hard for extended sessions. My 4-year-old GTX 1050 laptop still works great with few issues as I dust it out once a year + I clean the cooling pad every 2-3 months. 

I wish we could stop power cycling stuff, but think of the environment in general. And Ye Olde Power Grids... 

Edited by JeromFranzic
Link to comment
Share on other sites

59 minutes ago, Anna Nova said:

There was a British engineer in the 1940's, Tommy Flowers, who discovered that if you got a Thermionic Valve (I think there is an American word, but I forget it) Computer working, and never turn it off (and on), it kept going almost forever.   Valve's were replaced by transistors in the 1950/60s, because valves were 'unreliable' (Flower's work was subject to Official Secrets Act).  That's Tommy Flowers who built the first stored program computer, and didn't get the publicity that Turing, and of course the Americans, did.

Maybe we need to stop power cycling the little buggers.

The degradation mechanisms for vacuum tubes are different than for solid state devices. While there are components in power supplies that are stressed by power cycling, that stress will be compensated for by a good designer. The rest of the system is generally insensitive to power cycling, and ages entirely due to on-time and operating (or even storage) temperature. If you aren’t using it, shut it down. 

  • Like 3
Link to comment
Share on other sites

3 hours ago, Anna Nova said:

There was a British engineer in the 1940's, Tommy Flowers, who discovered that if you got a Thermionic Valve (I think there is an American word, but I forget it) Computer working, and never turn it off (and on), it kept going almost forever. 

Power-on failure is a very common phenomenon, and as Coffee mentioned, it's usually the inrush current that on-board capacitors are demanding that is the culprit. Simply upping the power supply won't stop that happening though.

Most electronic devices have a shorter lifespan when worked at an elevated temperature.

  • Like 2
Link to comment
Share on other sites

56 minutes ago, Profaitchikenz Haiku said:
4 hours ago, Anna Nova said:

There was a British engineer in the 1940's, Tommy Flowers, who discovered that if you got a Thermionic Valve (I think there is an American word, but I forget it) Computer working, and never turn it off (and on), it kept going almost forever. 

Expand  

Power-on failure is a very common phenomenon, and as Coffee mentioned, it's usually the inrush current that on-board capacitors are demanding that is the culprit. Simply upping the power supply won't stop that happening though.

Most electronic devices have a shorter lifespan when worked at an elevated temperature.

Valves, tubes, same.

I saw interesting stories about how our incandescent light bulbs had planned obsolescence for the last century almost.

Link to comment
Share on other sites

On 7/2/2022 at 3:56 PM, Coffee Pancake said:

Always buy a much bigger power supply than you need. Overkill is your friend and will save you money in the long run.

There's also Murphy's law of PSU failure.

It always lets the smoke out on a holiday weekend.

Failure may seem easy to fix, and a PSU has plugs and sockets for connections, but there's still a fair bit of your time spent on taking out the old and putting in the new. How long do you spend on getting a replacement? But if you have a margin, a PSU can last a long time.

Link to comment
Share on other sites

Should I finally bother to upgrade my GPU? My PC seems to run Second Life just fine.


CPU: Intel(R) Core(TM) i7-2600K CPU @ 3.40GHz (3410.02 MHz)
Memory: 32720 MB
OS Version: Microsoft Windows 10 64-bit (Build 19043.1766)
Graphics Card Vendor: NVIDIA Corporation
Graphics Card: GeForce GTX 560/PCIe/SSE2

 

Link to comment
Share on other sites

You are about to reply to a thread that has been inactive for 666 days.

Please take a moment to consider if this thread is worth bumping.

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
 Share

×
×
  • Create New...