Turns out the RTX A4000 wasn't a good daily driver GPU
Вставка
- Опубліковано 9 лют 2025
- In this video I talk about my experience with a single slot workstation card, and why I'm switching back to a normal consumer gaming card. I also pop it open and do some investigation into why she so dang hot!
*Turns out the RTX A4000 wasn't a good daily driver GPU*
>Guys its toasty as hell I regret buying this crap
>I got it a little cleaning
>Turns out the RTX A4000 is a great card I will keep it
EDIT: ok I got a little angry cause at the end I felt like watching it was totally wasted time as there was no message or reliable info other than "Check your vents". Besides this particular example, you got that something that kept me watching. Just don't make it 100% filler and you will do just fine ;3
Thanks for the feedback
The a4000 has a job to do, above your pay grade.
I’m sure it’s in some ai render farm now 😆
🤣
2:25
If the card gets this hot at idle, turning the max power limit down does absolute nothing, since it is not even close to being power limited at idle ...
Right you are, I didn't even think of that!! Man if this thing was getting THAT hot pulling like 15w at idle there was a major issue. Thanks for pointing that out 🤦♂️
I did an experiment with the RTX A4000, I put liquid cooling on it, it stays at 35 without doing anything in game and rendering does not go over 55.
Hilariously small heatsink. No wonder the fans need to exceed the speed of sound to make that heatsink work
I think one of the main issues too with this particular card is the opening in the heatsink for air to flow through is small and very prone to getting blocked by dust.
Well, its meant to be stacked side by side in a row and pushed by server fans.
It's probably meant to be in a workstation or server that has alot of airflow already, not in a water-cooled enthusiast PC.
Also, 80C, while high for a GPU, is still below the maximum safe temperature for this type of GPU. NVidia GPU's can go up to 95C.
@@diazrocks Not really. A2000 to A6000 are for workstations on the desktop. With a blower GPU you just have to embrace the bake. They are made to operate at temperature. For 4000 I'd say is meant for GPU accelerated workloads that are not graphics intensive. Using CUDA to do calculations that benefit from it. The issue is the most consumer like GPU's in the A series A5000 and A6000 will cost you a kidney.
not a desktop card
uh, turning the power limit down during idle isn't gonna do shit, thats not how things work.... you need to drop the voltage for that
Absolutely. I overlooked that originally. Comments like yours are one of the reasons I enjoy making these videos to have people smarter than me chime in and catch when I make an error. That way we can all learn and grow!
@@ultraprecisiontechnologies Thanks for spreading disinformation for your own gain, we don't need more of you.
Many thanks for this. Exactly the same symptoms as my workstation with RTX A4000 card.
I'd be super interested in knowing how it did in games. I use my computer for solidworks during the day and gaming at night. With the cost of a good pro grade card not being that much more than a gaming card....
I have used a Fermi based Quadro 4000 and a Kepler Quadro K4200 as daily drivers they were fine. I still have them for back up to my 1070ti 8gb
That's the server SKU? It's designed to be ran with server style thermals - front to back with a large volume of air being moved.. That's probably why it's still running a bit hot.
Help me understand what SKU is.
a SKU number is a number associated with a specific model or item. Technically it doesn't apply here as A4000 is the GPU model and not the SKU number. The terminology "SKU", "model", and "part" are sometimes used interchangeably. (Example, "That's the server SKU?" is basically saying 'That's the server model?')
Hope that helps clarify how SKU is thrown around in the PC world.
Also, SKU stands for “stock keeping unit”. Typically 8 characters long. For what it's worth, I found on PNY's website (a manufacturer of the A4000) that their SKU # is: VCNRTXA4000-PB for this GPU.
What is the power consumption at idle and at full load?
I have a 4070 Ti. Would you say that my regular consumer GPU works better than this A4000 Quadro?
100%
Oh 100% even ignoring the inadequate cooler on this, the a4000 is between a 3060ti and 3070 with more ram
i bought mine, because no one wanted them, initially during the mining craze, and B&H didn't mark it up. it is just for my capture rig 2nd editing computer. in a X299 system.
mine is 38c no changes at all.
RTX A4000 is amazing card. With heatsink swap it fits even in smallest ITX cases. Stock heatsink is terrible anyway.
Quadro's are great, Just strap on the RTX / GTX cooler and you have a GPU which will last often quite a bit longer.
Seems I have the same problem with my Quadro RTX 4000, I bought it new, and it was always running hot!
8:26 "undervolted to 71% power target"
Makes absolutely zero sense.
Huge temp difference
There is no indication that the KPX won't be squished out after an arbitrary number of thermal cycles. KPX is a LN2 paste which does not necessarily make it good for ambient use. Liquid metal is best for scenarios where there is little mass to soak up the heat but the heatsink is not Ni coated so you will have faster amalgamation. So thermal paste will be "less better" than LM on a 3090 FE for example than on an A4000.
All modern thermal pastes are subject to the dreadful pump-out effect it seems and I found the stock pastes to be far superior in this regard than something like MX4 or MX6. I didn't mention MX5 because it simply disappears before it has any chance to squeeze out. I recently repasted my 3090 FE with MX6 and it pumped out in less than 2 months and started performing worse than the dried up stock paste of 4 years. Then I tried PTM7950 and it has been perfect so far and should stay perfect for far, far longer than the service life of the card.
I used LM on my previous GPU, a 1080ti Gaming X and I used it like that for more than a year, sold it to a guy and he has been rendering stuff on it even today with no performance degradation whatsoever. I'd use the PTM if I had it at the time though. Does LM perform far better than PTM7950 thermally? Hell yeah. Would that have mattered for the 1080 ti Gaming X? No. Would it have mattered for the A4000? Very much yes. Would I prefer it to PTM7950? No, but only because the oxide layer on the Cu heatsink is more permeable than the oxide on the Ni coating.
The moral of the story is that any particular thermal paste would perform differently on different surfaces. Just because it's good on the CPU IHS it wouldn't necessarily be good on bare silicon. Also, I kind of find it funny that LM is not really more useful on an IHS, I mean it works and the performance is still superior than a with a thermal paste but it's not worth the damage on the IHS at all.
I recently found out first hand that kpx doesn’t last long term due to pump out, as I’ve been using it for a couple years now. I hear Thermal Grizzly is launching 2 new thermal pastes that specifically focus on longevity instead of maximum performance, so I’d be interested to try those out.
We should also mention Thermal Grizzly Carbonaut pads, as they offer great temps and longevity, but they are expensive per application.
That is false. How much heat the heatsink can soak has nothing to do with the TIM.
The TIM only cares about the Watts per area per temperature difference. This is a big chip and low power. Meaning the difference between LM and regular paste is small. Think of it like the heat transfer between the heatsink and air is the bottleneck, not the TIM.
@@Violet-ui you are missing one point: having less surface area on the interface limits the amount of heat that it could give away at any unit of time. This therefore makes the efficiency of the thermal interface material that much more important. This sounds like it doesn't make sense but you have more time to take the heat away with a larger heatsink with more surface area.
I had a 7700K that I had to delid, I delidded two of them actually, and I tried both mx-4 and TG conductonaut between the chip and the IHS, then mx4 between the IHS and the heatsink. The difference was more than 15 degrees C. The results didn't change across multiple applications of mx4 and conductwhatever. 7700K had a very small die. Keep this in your mind.
Then I did the same with a 1080 ti gaming X. GP102 was a huge chip with like 5 times the surface area of 7700K. The difference between mx-4 and conductwhatever was around 5 degrees C. The main problem was the pump-out effect so I had to use LM regardless.
Visualise this:
There is a small die, then there is an IHS on top of it which is for all intents and purposes a heatsink then there is a larger heatsink on top of the IHS. 7700K produces X joules of heat, then this heat has to be transferred from the surface of the silicon which is Y mm^2 to the IHS with around 6Y mm^2 surface area only to be transferred to another surface with a bit larger area. The bottleneck is between the silicon and the IHS.
It is always more difficult to extract heat from a smaller surface. This is exactly why 100 and 102 dies have considerably larger surface areas than their brethren with far less heat output. Engineers leave empty spaces of the silicon only to make it possible to take the heat away.
@@ultraprecisiontechnologies that thing is the most electrically conductive material in the world (in room temps). It's far more dangerous than LM because the shit chips and crumbles to produce little circuit shorters that could travel anywhere with some airflow. PTM7950 seems to be the better choice.
@@ahmetrefikeryilmaz4432 Read my comment again. You are missing how the surface area is big for the amount of power and the power density is low.
Your 7700K will have far higher power density, which is why the LM made a big difference. The 1080 Ti has a lower power density and the A4000 has an even lower one than that. So by your numbers it would probably be 3 to 4 C better than MX4.
Thanks for this. i too have an RTX a4000, and I will keep an eye on the temperatures. So far so good after 3 years. Cheers. BTW, funny that I too bought an Asus RTX 4070 two weeks ago. but that's for my new build. Ha
The a6000 would had been the one to get, and u want a case wirh proper controlled airflow tho may be overkill I’m terms of price 😅
Nah you have to change the memory pads and the paste on the chip... You have to put memory pads on the memory pads and also paste on the heatsink to above the memory.
good video
Do you still have a4k or getting rid of it
Sold it, but wishing I didn't as it would have been perfect in my virtualization server lol
@@ultraprecisiontechnologies oh rip
These temperatures drop are absolutley shocking! I really think PC owners should vaccum dust their environment at least 2 times per week. I have an RTX 4090 and it went around 74-78 Celsius in full load while in games like Alan Wake II but I am talking about this summer when the average in the house was 34-36. So I think good cooling is a combination of a well designed shroud, good thermal paste or pads and thick heatsink. A good proper air collled case with lots of fans also works but as I said, don't forget to vacuum dust two times per week. Especially in big cities with plenty of construction yards nearby.
clickbait :/
What were you expecting out of the video vs what didn’t I deliver?
Stop the 🧢
Maybe it is a bit late to know this, but the problem is that using the workstation card in a consumer hardware it will force the card to work in a way that is not 100% coeceive for, it works the other way around too, consumer GPUs don't perform well in workstations, it has to do a lot with the bond between the CPU - GPU duality. havve a good day
It simple Cyts Only Video no FUSION maybe a simple Color Correction. What Gives.
3:57 🤣🤣🤣🤣
almost buy trap gpu 😂😂 , im fine with rtx2000
it was the paste being trash. When your paste starts going bad again in a couple months replace it with the thermalright Heilos AMD paste (its AMD IHS shaped, read square, so itll fit your gpu die)
KPX.
I5⏰4460 🧒🏻ADD OIL！BECAUSE A4000？HUH