Jump to content

Laughing about used server hardware pricing.

Does anyone else like to build with old server hardware then get blown away when they look at the prices at launch?
I'm building a system for self hosted ai and machine learning. I know nothing about it and thought it would be fun to learn.
 
Core specs:
Dual E5 2686 V4 (32 cores, 72 threads total)
128GB ECC RAM
2TB Gen 4 NVME SSD
(4) 1TB SATA SSDs in RAID 0
(4) Tesla P40 24Gb cards (uses the GP102 chip, same as the Titan XP and 1080TI)
Total price for above hardware was (except SSDs) less then $1,100.00.
 
The prices new in 2016 when they launched.
E5 2686 V4 $2,141.00 each
128GB ECC DDR4 (varies depending on brand and speed) Appx $750.00 according to a google search.
Tesla P40 $5,699.00 each.
Dual socket motherboard for E5 V3 and V4 CPS $600.00
 
At release the total price to build this would have been $28,428.00, and I bought it all for less then $1,100.00.
Link to comment
Share on other sites

Link to post
Share on other sites

Technology depreciates.  You can buy old previously expensive cars for a fraction of their value.

Also the PC you buy today for $2000, will be scrap value in 10 years.

It's not really funny, it's just reality and there is so much old gear constantly being scraped or sold after it's reached End of Life.

If you think buying stuff that was really expensive but now is cheaper is actually cool and fun, I have a pen that was valued at $10 million dollars and I will sell it to you for $10,000.  What a great deal you are getting!

Link to comment
Share on other sites

Link to post
Share on other sites

There is something here, business hardware is very expensive compared to consumer hardware that can have surprisingly equivalent performance. There are reasons for it, you get business-targeted support and warranties, and theoretically more robust design in general. Also companies are just willing to pay more.

As for the current-day price, it's just come down to being actually reflective of performance. With the exception of having huge amounts of VRAM, the rest of that server is weaker than what you could get in a current gen gaming desktop.

Link to comment
Share on other sites

Link to post
Share on other sites

12 minutes ago, Jeeperforlife said:
Does anyone else like to build with old server hardware then get blown away when they look at the prices at launch?
I'm building a system for self hosted ai and machine learning. I know nothing about it and thought it would be fun to learn.
 
Core specs:
Dual E5 2686 V4 (32 cores, 72 threads total)
128GB ECC RAM
2TB Gen 4 NVME SSD
(4) 1TB SATA SSDs in RAID 0
(4) Tesla P40 24Gb cards (uses the GP102 chip, same as the Titan XP and 1080TI)
Total price for above hardware was (except SSDs) less then $1,100.00.
 
The prices new in 2016 when they launched.
E5 2686 V4 $2,141.00 each
128GB ECC DDR4 (varies depending on brand and speed) Appx $750.00 according to a google search.
Tesla P40 $5,699.00 each.
Dual socket motherboard for E5 V3 and V4 CPS $600.00
 
At release the total price to build this would have been $28,428.00, and I bought it all for less then $1,100.00.

when you buy server hardware new, you are paying for the low volume R&D costs, and you are paying for the risk of failure.
the hardware itself is nothing, it can be measured in price/performance. if the gpu preforms the same as a 1080ti, the gpu is worth as much as the 1080ti used. its simple. no one will pay more for the same performance used.
it is the same with the cpus.
the motherboards can stay high in price because they are rare ish but if they do that, the demand falls to zero. the boards eventually get sold for reasonable prices.

Link to comment
Share on other sites

Link to post
Share on other sites

11 minutes ago, Jeeperforlife said:

Does anyone else like to build with old server hardware then get blown away when they look at the prices at launch?

Not really anything to be blown away by. Hardware loses its value overtime, the older it gets, the faster it becomes worthless.

I own a 5000€ Server that I bought for 45€, because it's almost 16 years old at this point and gets left in the dust by a half decent laptop in a lot of tasks.

That's just how technological advancement goes.

English is not my first language, so please excuse any confusion or misunderstandings on my end.

I like to edit my posts a lot.

 

F@H-Stats

The Folding rig:

CPU: Core i7 4790K

RAM: 16 8GB (2x4GB) DDR3-1600

GPU 1: RTX 2070 Super

GPU 2: GTX 1060 3GB

PSU: Gigabyte P450B EVGA 600BR EVGA 750BR

OS: Windows 11 Home

 

Linux let me down.

.- -- --- --. ..- ...         

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Hello!

Link to comment
Share on other sites

Link to post
Share on other sites

Along with what others have already said, this kind of hardware likely brought the company it used to be on many times its value in revenue, so in the end its actual price is pretty meaningless, and during its lifespan I bet its power usage actually costed more than the server itself.

 

Many companies actually pay for other companies to get rid of those servers when they're upgrading in order to achieve better performance density/reduce their electricity bill.

FX6300 @ 4.2GHz | Gigabyte GA-78LMT-USB3 R2 | Hyper 212x | 3x 8GB + 1x 4GB @ 1600MHz | Gigabyte 2060 Super | Corsair CX650M | LG 43UK6520PSA
ASUS X550LN | i5 4210u | 12GB
Lenovo N23 Yoga

Link to comment
Share on other sites

Link to post
Share on other sites

So euhh not to bring you down but as long as it doesnt run out of vram a rtx 2060 12gb beats all of these systems in ML tasks handily just on its own.

 

Rt cores are an EXTREME performance uplift

 

As for pricing. Keep in mind these things go from the BUSINESS price class to consumer price class and are also quite old by then. Hence the extreme drop in value

Link to comment
Share on other sites

Link to post
Share on other sites

2 hours ago, jaslion said:

Rt cores are an EXTREME performance uplift

 

Tensor cores*

RT cores are useless for that kind of workload, to the point of being non existent on x100 chips lol

FX6300 @ 4.2GHz | Gigabyte GA-78LMT-USB3 R2 | Hyper 212x | 3x 8GB + 1x 4GB @ 1600MHz | Gigabyte 2060 Super | Corsair CX650M | LG 43UK6520PSA
ASUS X550LN | i5 4210u | 12GB
Lenovo N23 Yoga

Link to comment
Share on other sites

Link to post
Share on other sites

14 minutes ago, igormp said:

Tensor cores*

RT cores are useless for that kind of workload, to the point of being non existent on x100 chips lol

Sorry my bad thanks for correcting.

Link to comment
Share on other sites

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×