Humor me...I talk to myself
- Thread starter Vester
- Start date
I run AMD Radeon HD 7990 video cards. Only one of 5 was new when purchased. I found various problems with all four of the used cards which would explain why I got them so cheap, $125 to $200 each. I reseated the heatsinks, two per card, on all five video cards.
The worst case was a Diamond card that someone had worked on and broken one of the four wires on a fan connector. The fans all ran, but there was no speed control for two fans of 3 and the card would overheat. I repaired the broken wire, but ordered a new replacement fan assembly on eBay for about $20. The affected GPU's heatsink was badly crowned and black from overheating. Lapping with 320 grit, 800 grit, and 1500 grit wet or dry sandpaper restored the heatsink.
It pays to remove the backplates, too. I have found dried or missing thermal pads. Most all video cards use 1.0 mm thick thermal pads for the memory. The Nvidia 1080Ti is one notable exception that uses 0.75 mm thick pads, so check the specs before you try this at home. DuckDuckGo is my friend.
Buying used components can save money, but one has to be willing to take chances and do some work. I even disassembled the new video card and reseated the GPUs after testing it and finding that it ran hotter than some of the other cards.
The worst case was a Diamond card that someone had worked on and broken one of the four wires on a fan connector. The fans all ran, but there was no speed control for two fans of 3 and the card would overheat. I repaired the broken wire, but ordered a new replacement fan assembly on eBay for about $20. The affected GPU's heatsink was badly crowned and black from overheating. Lapping with 320 grit, 800 grit, and 1500 grit wet or dry sandpaper restored the heatsink.
It pays to remove the backplates, too. I have found dried or missing thermal pads. Most all video cards use 1.0 mm thick thermal pads for the memory. The Nvidia 1080Ti is one notable exception that uses 0.75 mm thick pads, so check the specs before you try this at home. DuckDuckGo is my friend.
Buying used components can save money, but one has to be willing to take chances and do some work. I even disassembled the new video card and reseated the GPUs after testing it and finding that it ran hotter than some of the other cards.
Last edited:
I have requested a trade-in for my OCZ ZX 1000W power supply from PC Power & Cooling. I am now running just two video cards of 5 on the mining rig. My order for a used 1000W power supply (to run the 5th video card) has been delayed by the weather. I plan to get the Turbo Master 1350W power supply. With the trade-in, it will cost $225. If I am lucky, the trade-in unit could be under warranty.
One of the beta testers, who is more knowledgeable about computers than me (has more test equipment), showed graphs of the WCG Beta work units causing voltage spikes and said the spiking loads could lead to power supply failures. I think my unit has been going bad for several weeks. When it made sizzling noises, the computer would lockup.
One of the beta testers, who is more knowledgeable about computers than me (has more test equipment), showed graphs of the WCG Beta work units causing voltage spikes and said the spiking loads could lead to power supply failures. I think my unit has been going bad for several weeks. When it made sizzling noises, the computer would lockup.
Sizzling is never good...unless it's bacon cooking.
I should name one of my computers "Lake Woebegone." The second CPU arrived today and it is in service in the server...by itself. It turns out that the original processor was an Intel E5-2420, not an E5-2420 v2. That is OK because I quickly ordered another v2 processor. A computer will not run two mismatched processors speeds.
I have not had a response from PC Power & Cooling about my trade-in. The mining rig is still running just two video cards.
I have not had a response from PC Power & Cooling about my trade-in. The mining rig is still running just two video cards.
The mining rig is running all 8 GPUs on Milkyway@home while waiting for more WCG Beta work. My "like new" but used 1000W power supply is handling two Radeon HD 7990s.
One of the five Radeon HD 7990s appears to be for "parts only" after the rig ran 172 WCG Beta work units. The associated power supply is history, too. I suppose it is another chicken and egg story. Did the power supply take out the video card or vice versa?
I've been hauling away the crepe myrtle limbs that I cut yesterday and last Saturday. It is 74 F at our house today. I wanted a warm day, but it feels like June after being 26 F one morning last week with recent highs in the mid 50s.
One of the five Radeon HD 7990s appears to be for "parts only" after the rig ran 172 WCG Beta work units. The associated power supply is history, too. I suppose it is another chicken and egg story. Did the power supply take out the video card or vice versa?
I've been hauling away the crepe myrtle limbs that I cut yesterday and last Saturday. It is 74 F at our house today. I wanted a warm day, but it feels like June after being 26 F one morning last week with recent highs in the mid 50s.
I am running two processors in the server now and it is running 24 WCG work units simultaneously.
If anyone offers you a server and you've not had one before, think twice. The learning curve has been steep and the configuration is limited. I don't have an iDRAC license and iDRAC needs to be updated so that I can update the configuration. Until the configuration is updated, it will not utilize the second bank of memory. I put 24 GB ECC RAM (six slots) on CPU0 and shared the memory cache between the two. I have seen no errors so far on this machine. It is a 2.5 GHz draft horse.
If anyone offers you a server and you've not had one before, think twice. The learning curve has been steep and the configuration is limited. I don't have an iDRAC license and iDRAC needs to be updated so that I can update the configuration. Until the configuration is updated, it will not utilize the second bank of memory. I put 24 GB ECC RAM (six slots) on CPU0 and shared the memory cache between the two. I have seen no errors so far on this machine. It is a 2.5 GHz draft horse.
The APC Back-UPS 600 for the satellite TV tuner and TV failed Monday. Being a good husband, I took two computers off of a Back-UPS Pro 1000 and put it to use in the living room. The replacement for my computers arrived today. It is a Back-UPS Pro 1500 from Amazon for $164.99 plus tax with free two-day shipping. The things I do for my wife!
With the UPS I can see the load in Watts. My desktop computer with Intel Core 2 Quad and an unloaded (no GPU work) Radeon HD 7790 crunches four work units at 150 Watts (minus whatever the modem and router use). The Dell PowerEdge T420 with two Intel E5-2420 processors and onboard video (minus whatever the modem and router use) crunches 24 work units at 150 Watts. Really.
Nvidia's RTX 3080 Ti Could Have a Mining Limiter and could apply to all future RTX 30 cards. This is bad news for GPU crunchers at BOINC and Folding@home.
This is like having a 4000 RPM rev limiter on a Ferrari.
Last edited:
I think this is the link: https://www.tomshardware.com/uk/news/rtx-3080-ti-3070-ti-ethereum-limiter
From what I've read elsewhere, this is specific to Etherereum hashing which has a distinct and easily detectable pattern. Other usage shouldn't be affected, but we'll have to see.
I found a one-year-old article on SSD reliability that you may like. "SSD 101: How Reliable are SSDs?" by Roderick Bauer.
My great-niece, Harper Krimm (16), sings "Skylark" followed by her younger sister, Mallory (12), singing "Anything You Can Do, I Can Do Better."
Last edited:
Not a surprise it didn't work, but surprising that Nvidia broke it themselves.