PCIe Slot Life Cycle(Insert and Removal gpu)

[Knight47]

So I was wondering about the PCIe slot mating cycles.

TPU tested like 100 GPU's after they switched to EVGA Z790 Dark mobo, was that in one board or did they replace it multiple times?

 

[arbiter]

I won't hazard a guess but a lot of testing sites use same board for years of testing gpu's swapping out gpu's all the time. So i would say lifetime is not something you need to worry about. If issues with it come up reviewers with gpu test benches would have had them.

 

[DemonicRyzen666]

So I was wondering about the PCIe slot mating cycles.

TPU tested like 100 GPU's after they switched to EVGA Z790 Dark mobo, was that in one board or did they replace it multiple times?

If I recall correctly, I remember reading it in another thread here. it's very low something like 15-20 is the expected cycle lifetime for the slot.
I believe it is listed in the thread about the New 16 pin power adapter.

 

[Fouquin]

The official limit is quite low. The practical limit is in the thousands.

I have an AM3 board from 2010 that I have tested no fewer than 400 cards on in the last decade+, and done reinsertions of a few of those cards at least a dozen times each when swapping positions for SLI/CrossfireX setups, or removing and then reinstalling two different cards for a head-to-head.

 

[GerKNG]

my z490 Unify had at least a hundred installation cycles. the slot gets a bit "softer" but it works just fine.

 

[iameatingjam]

tbh I've been wondering the same thing about CPUs. Like I heard somewhere that it is common for CPU sockets to only be supported for the first... 10 or 20 mating cycles, can't remember exactly what the number was but I was shocked at how low it was for sure. Yet when I go googling its hard to find actual numbers for modern CPUs... Is this a problem that has been solved? Or the opposite? A problem that exists that the manufacturer wants to hide?

Anyway sorry no intention to hijack the thread.. figured the questions were related.

 

[arbiter]

tbh I've been wondering the same thing about CPUs. Like I heard somewhere that it is common for CPU sockets to only be supported for the first... 10 or 20 mating cycles, can't remember exactly what the number was but I was shocked at how low it was for sure. Yet when I go googling its hard to find actual numbers for modern CPUs... Is this a problem that has been solved? Or the opposite? A problem that exists that the manufacturer wants to hide?

Anyway sorry no intention to hijack the thread.. figured the questions were related.

It sounds like a low number then when you think about the avg user how often do they change cpu's? how often do they pull out gpu and replace them vs when upgrading of MB/cpu happen. I had an old i7 4770k i used for 10 years i probably barely went pass that 20 in whole time i owned it. I had i think a gtx780 which was replaced with a 980 then a 1080. As low as it sounds as what they put in you then look at how a lot of reviewers are pulling cpu's out on the reg and swaping gpu's and say never problems with 100's of swap outs. Shouldn't need to worry about anything.

 

[ir_cow]

Wiz swaps out a lot of cards, but I've done 300+ personally on the same motherboard. These newer slots are more durable, but the only thing I ever broke was the latch.

NVIDIA says you cannot only plug in the 16pin connector 20 times. Yeah i'm well past that now

 

[LabRat 891]

tbh I've been wondering the same thing about CPUs. Like I heard somewhere that it is common for CPU sockets to only be supported for the first... 10 or 20 mating cycles, can't remember exactly what the number was but I was shocked at how low it was for sure. Yet when I go googling its hard to find actual numbers for modern CPUs... Is this a problem that has been solved? Or the opposite? A problem that exists that the manufacturer wants to hide?

Anyway sorry no intention to hijack the thread.. figured the questions were related.

I've experienced an LGA 775 and an LGA 1155 socket 'fatigue'.
The LGA 1155 had probably 6-8 replacement/reseat cycles.
The 775 was an eWaste-pull EP45 board.
-so, it could've only had the 1 CPU before me, or many. I used it for testing my LGA 775 eWaste-pull CPUs; it probably had a dozen+ cycles on it before I had any issue.

In both cases, a simple (and VERY GENTLE) 'bending back/up' of the LGA 'fingers' resolved the issue.

Wiz swaps out a lot of cards, but I've done 300+ personally on the same motherboard. These newer slots are more durable, but the only thing I ever broke was the latch.

Definitely the metal-reinforced PCI-e slots. IMO, 100% a necessity as cards got so dang heavy.

I have had 1 PCIe x16 'metal fatigue' on me, though. IIRC, on my Z77 Extreme4 I had to take a sewing needle to the contacts, and bend them back out. Seemed fine after that.

NVIDIA says you cannot only plug in the 16pin connector 20 times. Yeah i'm well past that now

Semi-related: my Rosewill HIVE 1kW's PCIe 6+2-pins are very worn.
(mining PSU before main rig PSU, swapped a dozen+ cards 2-3 times ea. w/ these plugs)

They slide in too easy, and like to 'rest' on the "stop-latch" of the retention clip. Thankfully, w/ the design 'overhead' of the 6- and 8-pin PCIE power receptacles, I can continue to use them.
Recently pulled 300-350W (6+8 input) through the 2 lines out of my PSU, with no issues whatsoever. So, even noticeably worn, PCIe 6 and 8 pins are G2G, it would seem.

 

[Knight47]

It sounds like a low number then when you think about the avg user how often do they change cpu's? how often do they pull out gpu and replace them vs when upgrading of MB/cpu happen. I had an old i7 4770k i used for 10 years i probably barely went pass that 20 in whole time i owned it. I had i think a gtx780 which was replaced with a 980 then a 1080. As low as it sounds as what they put in you then look at how a lot of reviewers are pulling cpu's out on the reg and swaping gpu's and say never problems with 100's of swap outs. Shouldn't need to worry about anything.

Unless you need to troubleshoot something and the gpu is in the way.

First I ran the PC outside of the case(1 times). I removed my gpu to replace sata cables twice(got write errors) and remove the cables after(total 3 times). Insert and swap NVMe drives, it's under the gpu and my hands are too big(total 4 times).
So I that's total of 15 (8 insert, 7 remove). I usually remove the gpu twice a year to blow off dust and clean the fan blades with Q-tips, at least I used to do it. But I gues i'm already at 15 and it's the end of it's cycle if 10-20 insertions(what iameatingjam said) is true.

 

[Outback Bronze]

but I've done 300+ personally

I suppose where you're coming from, installing memory would be very similar to GPU's and I'm guessing you've done a shit ton of memory installs on the same mobo : )

 

[evil Kelly]

I get nervous when components stray from defined industry standards like ATX with no innovative or discernable reason.

Like when some board manufacturers where omitting the third fastener on the IO side of mainboard {the one under the IO closest to the main PCIE slot} as graphics cards were getting larger and heavier and then when the cards themselves got so Un optimal PCB huge that they required little plastic jacks to support them, I think I purchased the last graphics card sized graphics card recently and the ATX standard may be gone forever..

Picture stolen from internet..

 

[iameatingjam]

It sounds like a low number then when you think about the avg user how often do they change cpu's?

Thats true... but for me anyway, I've already upgraded the cpu once that actually included a number of inserts because of a tiny spec of thermal paste blocking a pin and the troubleshooting required to find the problem...

I've put in a temporary cpu in while I waited for an rma.

Then put in the rma replacement itself.

And the board is only a year old! I sure hope its not causing that much wear on the connectors to have already like 7 inserts on it.

Good news is that I see it unlikely this will get another upgrade unless bartlett lake brings something to the table which I see as unlikely.

 

[freeagent]

The slot itself should be good for hundreds of insertions, same with a dimm slot. But if you swap your own hardware around a lot then you will see wear on the device fingers. Asus boards that have one latch on the dimm slot will show a hook pattern on the fingers after awhile. The same with your PCIE device..

 

[Assimilator]

Why does it matter?

 

[bonehead123]

Why does it matter?

For sure

For pcie/ram slots, I don't know the actual answer in terms of # of insert/removal cycles, but I would think that if it was really an issue, the reviewers here & elsewhere would have reported on it long ago....those folks typically use the same mobo's for a good while and test a large # of devices with them, at least until they move to a new generation and then the cycle starts again...

But the same would go for folks who typically upgrade their GPU's & ram frequently, either for real reasons like performance, features etc, or merely to have the "latest & greatest" thingies but keeping the same mobo as long as it continues to work as needed...

CPU sockets however, are a different story IMO, with all those teeny tiny pins and their corresponding connectors, which I would think are way more fragile than pcie or dimm slots

 

[Shrek]

I use silicone oil on mine, which should greatly extend the number of cycles it can tolerate (applied sparingly to the card connector and transferred that way).

 

[LabRat 891]

Semi-related: my Rosewill HIVE 1kW's PCIe 6+2-pins are very worn.
(mining PSU before main rig PSU, swapped a dozen+ cards 2-3 times ea. w/ these plugs)

They slide in too easy, and like to 'rest' on the "stop-latch" of the retention clip. Thankfully, w/ the design 'overhead' of the 6- and 8-pin PCIE power receptacles, I can continue to use them.
Recently pulled 300-350W (6+8 input) through the 2 lines out of my PSU, with no issues whatsoever. So, even noticeably worn, PCIe 6 and 8 pins are G2G, it would seem.

Sorry to be off-topic, but I do need to correct my statement:

Just discovered I had a bad lead.
(would've edited old post, if I could've)

New 7900 GRE wasn't 'clocking right' and was stuttering. Checked the 2 8-pins, and one of the two is noticeably warmer. Disconnected that 1, and plugged in a spare lead; all better.

 

[W1zzard]

I've swapped GPUs thousands if not ten thousands of times and the slot was never an issue. What happened once was that repeated positioning of the card scratched the motherboard in the area between the end of the connector and the end of the motherboard

https://imgur.com/QmNFbNt

Like here. Since then I put some tape over that area, unless the mobo has some shield coverage in that area

I ever broke was the latch.

I break that off intentionally on my GPU test systems because it's such a waste of time fiddling in there, possibly with a screwdriver, which has a significant chance to slip and damage the board. Love the ASUS quick release mechanism though on some other systems. But for GPU testing I'd remove that too, to save a few seconds

 

[de.das.dude]

I've swapped GPUs thousands if not ten thousands of times and the slot was never an issue. What happened once was that repeated positioning of the card scratched the motherboard in the area between the end of the connector and the end of the motherboard

https://imgur.com/QmNFbNt

Like here. Since then I put some tape over that area, unless the mobo has some shield coverage in that area

the master has spoken