Will you be Buying an RTX 3080?

[RasmusP]

When doing this do you think i should enable XMP first or change voltage or anything?

I'd start with the xmp profiles and go from there!
There is some tool for theoretically possible timings with your kind of ram but I forgot the name..
Ryzen ram timings calculator should give you some more info!
Sadly I'm really no ram expert so overclocking forums are probably the way to go from here on

 

[ko41]

Built my new computer in September 2020 in anticipation of the Nvidia RTX 30 series release..............
A year and a half later I get to finish it, I am picking up a RTX 3080 Ti today

My old 1080 Ti has been performing well running triple screens at 7680x1440 144Hz, I hope I am not underwhelmed by the performance of the 3080 Ti.

KO

 

[BillyBobSenna]

Built my new computer in September 2020 in anticipation of the Nvidia RTX 30 series release..............
A year and a half later I get to finish it, I am picking up a RTX 3080 Ti today

My old 1080 Ti has been performing well running triple screens at 7680x1440 144Hz, I hope I am not underwhelmed by the performance of the 3080 Ti.

KO

Your frame rates should about double. That is what happen for me when I changed from the 2070 to a 3080 Ti.

 

[ko41]

Your frame rates should about double. That is what happen for me when I changed from the 2070 to a 3080 Ti.

ACC went from high 50s and 60s fps to 90s and a little over 100 fps with the same mid to high settings that I had for the 1080 Ti. With everything set to Ultra with balanced DLSS, I get mid 90 fps, and smooth running without trees and things popping in as you go down the track, how cool is that!
3Dmark Time Spy score was 10600 for the 1080 Ti and 19230 for the 3080 Ti, CPU is a i9-10850K.
I am happy with the performance of the 3080 Ti.

KO

 

[TedBrosby-]

Just another reminder that (I've said this numerous times before, not sure if I said it here) Ethereum merge is like 95% (that's a conservative estimate) happening on time. The testnet has been going well and the dev team has already started warning GPU miners to be ready because they won't get block rewards after the merge is complete.

The reason Nvidia started sharply dropping prices is because they're aware that the used GPU market is about to be flooded and they don't want to get thousands of GPU's returned to them when they can't sell them in the summer.

Nvidia knows this stuff months in advance. A lot of "large scale" miners buy GPU's in bulk directly from AIB partners. The buying started drying up which meant that surplus GPU's could go to system integrators and retailers. System integrators haven't been able to sell PC's even with GPU's priced at MSRP for a month or two now, which means now retailers have surplus. So Nvidia slashed prices across the board because they had been sandbagging us, using the pandemic and chip shortage as an excuse for higher prices and everyone bought it (I didn't, I knew what they were doing).

These prices are "not that bad", but this is your warning: If you buy a 3080 based card for $1100 or $1200 now, don't say you didn't get warned when prices come down even harder in a month or two. Miners have been in denial that the merge is happening, but once reality sets in, they'll start dumping their hardware. And I don't think you can easily comprehend the thousands of GPU's (probably hundreds of thousands) that have been purchased for crypto mining. I'm holding out for a $600 3080 and I feel very confident I'll get it.

In before "miners will just mine a different coin"
No they won't.
Short answer: no profitability.
Long answer:
The hashing power with current GPU's in circulation is 100,000x or more higher than what would be needed for alt coins to be profitable. As more GPU's mine a coin, the coin gets harder and harder to receive rewards for. Ethereum has worked its way up to its current value over the course of nearly a decade. If even 5% of the GPU's out now moved to alternative coins, their difficulty would spike and block rewards would dry up. You'd be making pennies per week with free electricity and lose dollars a day if you pay for electricity.
This is different from the 2018 and 2014 GPU price crash. Those crashes were related to crypto value dropping. This is literally going to be the "main" profitable POW crypto that's been the primary source of GPU mining profitability for almost a decade suddenly not being POW at all. This has never happened before. Even when BTC moved to ASIC's, it was still POW.
I can't even predict how bad GPU prices are going to drop... My gut feeling says that even 50% might not be as big of a drop that we'll see. Just look at Reddit's GPU mining community or YouTube videos on GPU mining setups to get an idea of the kind of hardware people are running, and then tell yourself, that's only Reddit and YouTube.

 

[Bob Gnarley]

Great post Ted, thank you.

I would also like to add that though prices are well down on what they were, a lot of places are still holding inventory that they overpaid for when prices were high. As Ted iterates, It will pay a dividend if you can sit on your hands a little longer.

 

[blekenbleu]

still holding inventory

Indeed, there are many prebuilts available with 3070s or 3080s with CPUs from 2 generations ago.

 

[Terry Rock]

Just another reminder that (I've said this numerous times before, not sure if I said it here) Ethereum merge is like 95% (that's a conservative estimate) happening on time. The testnet has been going well and the dev team has already started warning GPU miners to be ready because they won't get block rewards after the merge is complete.

The reason Nvidia started sharply dropping prices is because they're aware that the used GPU market is about to be flooded and they don't want to get thousands of GPU's returned to them when they can't sell them in the summer.

Nvidia knows this stuff months in advance. A lot of "large scale" miners buy GPU's in bulk directly from AIB partners. The buying started drying up which meant that surplus GPU's could go to system integrators and retailers. System integrators haven't been able to sell PC's even with GPU's priced at MSRP for a month or two now, which means now retailers have surplus. So Nvidia slashed prices across the board because they had been sandbagging us, using the pandemic and chip shortage as an excuse for higher prices and everyone bought it (I didn't, I knew what they were doing).

These prices are "not that bad", but this is your warning: If you buy a 3080 based card for $1100 or $1200 now, don't say you didn't get warned when prices come down even harder in a month or two. Miners have been in denial that the merge is happening, but once reality sets in, they'll start dumping their hardware. And I don't think you can easily comprehend the thousands of GPU's (probably hundreds of thousands) that have been purchased for crypto mining. I'm holding out for a $600 3080 and I feel very confident I'll get it.

In before "miners will just mine a different coin"
No they won't.
Short answer: no profitability.
Long answer:
The hashing power with current GPU's in circulation is 100,000x or more higher than what would be needed for alt coins to be profitable. As more GPU's mine a coin, the coin gets harder and harder to receive rewards for. Ethereum has worked its way up to its current value over the course of nearly a decade. If even 5% of the GPU's out now moved to alternative coins, their difficulty would spike and block rewards would dry up. You'd be making pennies per week with free electricity and lose dollars a day if you pay for electricity.
This is different from the 2018 and 2014 GPU price crash. Those crashes were related to crypto value dropping. This is literally going to be the "main" profitable POW crypto that's been the primary source of GPU mining profitability for almost a decade suddenly not being POW at all. This has never happened before. Even when BTC moved to ASIC's, it was still POW.
I can't even predict how bad GPU prices are going to drop... My gut feeling says that even 50% might not be as big of a drop that we'll see. Just look at Reddit's GPU mining community or YouTube videos on GPU mining setups to get an idea of the kind of hardware people are running, and then tell yourself, that's only Reddit and YouTube.

Months ago, I predicted many of those mining farm GPUs would find their way back to market as new products.
I fully expect that trend to continue again and again to unsuspecting buyers in the coming months.
Be warned!

 

[Durge Driven]

AMD 3900X (underclocked- undervolted ) - RTX 2080 and 1440p Gsync

I honestly can't see me replacing my gaming tower for at least 3-4 years


My AMD 3200G I will keep till I kick it ( touch wood ) ..... best everyday PC ever

Mouse click away from getting RTX3080 ( MSI Suprim X 12G ) and 34" Ultra wide ( $650 cheaper )
Then I had quick think ! what else have you always wanted ?

1st sector out happy made this choice

 

[BrunoBæ]

Months ago, I predicted many of those mining farm GPUs would find their way back to market as new products.

Sounds not so good.
But do you have any proofs or something to back such suspicion up?
Or is it pure speculation?

 

[RCHeliguy]

Linus did some testing on used cards and found that there was no degradation in performance, so the question is whether they would last as long as you needed them to before upgrading to something better.

 

[BrunoBæ]

no degradation in performance

I guess nobody is worried about performance degradation but exactly the left lifespan from such 2nd hand g-cards.
And I furthermore guess these mining g-cards has not exactly been underclocked and/or undervolted

 

[RasmusP]

I guess nobody is worried about performance degradation but exactly the left lifespan from such 2nd hand g-cards.
And I furthermore guess these mining g-cards has not exactly been underclocked and/or undervolted

Afaik they are very often underclocked and undervolted to enhance efficiency!
But yeah they still have a tough life running 24/7 at 100% load with cooling probably just at the limit of performance loss.
So even if they only run 1500 mhz and take 230w instead of 1950 mhz at 370w, it's still a lot of torture...

The chips probably don't care but I'd be afraid for capacitors etc.

 

[RCHeliguy]

I agree. I wouldn't consider buying used a first choice. Unless it saved me a LOT of money, like 50% or better. Then I might consider it. I doubt such deals are available at this time. I'm also waiting for the 40 series.

 

[Trebormoore84]

Its the solder joints degrading and cracking, actually more likely on equipment turned on and off every day vs kit on 24/7.

Most cards "die" when the ball solder joints to the GPU crack after a period of time. You won't generally see any performance degradation, just one day get a black screen when it pops. You can get them reballed (a process where someone removes the gpu chip cleans it up and then re-solders it with fresh solder balls) and all solder joints on the back reflowed, i.e. by a specialist PCB electrical engineer if the card still has value but out of warranty, but usually by that point its not worth doing vs the cost of repair.

Mining cards I think will be alright actually despite the worry unless someone has been running them in a humid / external air environment or something, its a risk I guess, but you'd just want to remove the cooler, make sure its clean and replace dried out thermal paste / pads etc. A small amount of cost / effort if the price is right. I think miners are less likely to have dog hair and dust or whatever caked into their GPU coolers etc...

But used cards resold as new is a scam whatever way you look at it however.

 

[Enzo Fazzi]

I'd take an underclocked, undervolted card that spent its life in a well ventilated mining farm over your average GPU owned by someone who smokes/vapes, has pets or any of the other stuff you find in a typical house.

Its the solder joints degrading and cracking, actually more likely on equipment turned on and off every day vs kit on 24/7.

Add this on top of what I already said and I really wouldn't worry about buying a mining card (for a fair price).

 

[jgf]

In general I never buy the first generation of anything - you pay a premium price to be a beta tester. Wait a year, the bugs have been worked out, or at least are known, and the the price has dropped 30%.

 

[BrunoBæ]

Afaik they are very often underclocked and undervolted to enhance efficiency!
But yeah they still have a tough life running 24/7 at 100% load with cooling probably just at the limit of performance loss.

OK I take your word for for the 1st part.
Hehe eventhough its probably a generalisation based on pure guesswork

But I have re-thought my position on the 2nd part.
Because just like RL taxas often are able to drive 500.000 km before any engine issues compared to normal household cars - then I guess its less stressing for a g-card to be running 24/7 (with your uclock/uvolt) instead of being switched on/off 20 times a day.

 

[RasmusP]

But I have re-thought my position on the 2nd part.
Because just like RL taxas often are able to drive 500.000 km before any engine issues compared to normal household cars - then I guess its less stressing for a g-card to be running 24/7 (with your uclock/uvolt) instead of being switched on/off 20 times a day

That comparison lacks a little, sadly.
Because that taxi/taxa will drive the 500k km in let's say 5 years, while the standard City car will do it's 150k km in 10 years.
The time is just as important for the wear as the distance.
So while a well treated mining card might be healthy for 10.000 hours and a normal household card for only 2.000 hours, the household card will still run 2x as long in years in the calendar.

In the end we all have no real clue. The mining cards might be totally fine or a waste of money.
But it's not like they are all tortured into early electronic-deaths like many people think (me included for some time!).

 

[BrunoBæ]

That comparison lacks a little, sadly.
Because that taxi/taxa will drive the 500k km in let's say 5 years, while the standard City car will do it's 150k km in 10 years.
The time is just as important for the wear as the distance.

Thats the first time I have heard that alu and steel in an engine inside a reasonable time as say 10 years fall apart and needs repair because of age.
I guess the downgrading of a motorcar engine is much more dependent of lack of maintance and/or too much switch on/off caused by the many cold starts and short runs that is normal for an avg household car.
So...