HP Z4 G4 replaced Quadro p4000 with RTX2080

[Bolis]

Hello fellow HP product users.

I have an HP z4 g4 with a 7940x, optane 900p, 1x 1TB nvme, 1x 512GB nvme, 1x 256 nvme, 2x 6TB RAID0 hdd's & a Quadro p4000. The PSU is a 1000watt unit.

I swapped the p4000 for a RTX2080. Sadly the PC hardlocks when I put the RTX under load. If i put it under small load (<60% TDP) I dont get a hardlock. When I put a very heavy workload (Furmark) on it it hardlocks in one second.

The PSU has 4x 6+2 for GPU power, and only two of those are connected to the RTX2080 (6pin and 8pin). I tried all different combinations of the 4 6+2 cables, but it hardlocks on all. One time after a hardlock I got a PCIe 928-Fatal error.

Could the RTX2080 be faulty, or could it be a power related issue? 

When I check the slot settings  I see the auto setting for the RTX2080 is Gen 1 2,5.

I have tried:

Bios reset (bios is updated)

Tried different PCIe slots

Tried changing PCIe limit from Auto to Gen 3. 

Thanks for your advice.

Bolis

[BambiBoomZ]

Bolis,

The RTX 2080 required PSU is 650W and as there are a number of Z4 G4's listed in Passmark benchmarks using RTX 2080 Ti's - which draw more power than the the non- Ti, the 1000W PSU capacity is not an issue.

1. As to the 928 PCIe Fatal-Error Completion-Timeout, see this post:

https://h30434.www3.hp.com/t5/Desktops-Archive-Read-Only/928-PCIe-Fatal-Error-Completetion-Timout/td...

"I just wanted to post a solved issue with a Z440 and a recently occurring 928 Fatal Error. I believe the problem was the small power plug going into my AMD W7100 Video Card from the power supply. It had not been fully plugged in properly from the factory and was loose. It popped right off when I touched it. It was making enough of a connection for the card to work but when the card needed to pull some power while running a video demanding program. It caused the computer to crash."

As the RTX 2080 has been tried in different slots in the Z4, the above raises the idea that any added connector/adapters may not be properly attached at the GPU end, the connectors/ wires are faulty, there is a fault in the connection of the PCIe power cables at the power supply end. 

2.  Only to eliminate a residual driver conflict as a possibility. If it appears that the power supply and connection is proper and If the BIOS is up to date, consider reinstalling the RTX 2080 driver, insuring that the "Clean Install" option is checked. Also consider a preliminary and careful run of CCleaner,  not letting it delete too much.

BambiBoomZ

[DGroves]

earlier HP workstations used a multi rail power supply instead of the more common single rail supply found in consumer systems the main difference is that on a multirail supply it is possible to overload a power rail (GPU in this case) while still under the supplies total wattage

i have not used this model so i have no personal experience with it but i would look at the HP quickspecs for your model note the wattage listed for the GPU rail and compare it to the 250 plus watts that the nvidia 2080 can draw and if over the spec you can look for a aux power supply that fits in the 5.25 bay, but be aware these units are not cheap but they might solve your problem

https://www.amazon.com/450W-Graphics-Power-Supply-Juice/product-reviews/B001AO4O98

[BambiBoomZ]

DGroves,

The RTX 2080 is rated at 215W/225W (Founders Edition) and the NVIDIA required minimum power supply is 650W. 

https://www.nvidia.com/en-us/geforce/graphics-cards/rtx-2080/

As the OP's Z4 G4 has a 1000W PSU, the power capacity is not in question.

BambiBoomZ

[DGroves]

BambiBoomZ, please reread my comments..... if the Z4 G4 does use a multi rail power supply instead of a single rail supply then overloading the GPU rail is a real possibility as the 1000 watt supply only has two six pin aux GPU connections and i don't know if each gpu connection is on a seperate rail  or what the gpu rail(s) are specd at because as i said i don't own this system and have no experience with it

http://www.dynapowerusa.com/difference-between-single-rail-and-multiple-rail-power-supplies/

HP does spec the 290 watt Quadro RTX 8000 card as a option, but may include a gpu power adapter to take power from another rail besides the aux GPU one or they may just include the HP 6 to 8 pin adapter that is a overengineered/rated adapter that uses thicker wiring and better quality molex connectors compared to most of the generic 6 to 8 pin adapters

i do know that adding a high power card also requires the optional HP front cooling fan kit

it's also possible the "OP" is also using poorly made 6 to 8 pin gpu adapters

[Bolis]

"1. As to the 928 PCIe Fatal-Error Completion-Timeout, see this post:"

I will try to clean the PCIe slot and graphics card and retry different slots again to completely eliminate a PCIe slot problem. The 4 PCIe power plugs were properly seated and connected. The faulty PCIe power plugs could be the issue indeed.

"2.  Only to eliminate a residual driver conflict as a possibility."

I'll go ahead and do a driver sweeper and reinstaller the driver too.

The 1000watt PSU should be built for higher wattage multi GPU arrangements, since the CPU uses a locked max power usage of 165 watts combined with and all other peripherals + motherboard (PCIe onboard power) shouldn't use more than 400 watts. 4 PCIe 6+2 pins on a modern watt 1000 watt power supply should be atleast power a regular 250watt videocard that indeed can be ordered with the PC itself.

[BambiBoomZ]

Bolis,

Yes, I completely agree with the stated evaluation of HP WS design; these systems are designed generously so as to accommodate the highest performing components with high reliability.   

Before making any further changes,  consider starting the system and running HWMonitor or similar on top, monitoring the GPU  temperature so as to eliminate the possibility that there is a thermal shutdown, and initiate a medium GPU-intensive program- not the program that so quickly halts the system.  Monitor the CPU temperature secondarily. If the RTX 2080 is a multiple fan design- which puts the GPU-generated heat into the case- that might trigger a system thermal protection. It's not highly likely, but anyway, with a new, higher power GPU, an idea of the system operational temperatures is worthwhile.

BambiBoomZ

PS: I've found Passmark Performance Test very useful in researching and evaluating upgrades and changes. There's a free 30-day trial.

HP z620_2 (2017) (R7) > Xeon E5-1680 v2 (8-core@ 4.3GHz) / z420 Liquid Cooling / 64GB DDR3-1866 ECC Reg / Quadro P2000 5GB _ GTX 1070 Ti 8GB / HP Z Turbo Drive M.2 256GB AHCI + Samsung 970 EVO M.2 NVMe 500GB + HGST 7K6000 4TB / Focusrite Scarlett 2i4 sound interface + 2X Mackie MR824 / 825W PSU /> HP OEM Windows 7 Prof.’l 64-bit > 2X Dell Ultrasharp U2715H (2560 X 1440)

[ Passmark Rating = 6280 / CPU rating = 17178 / 2D = 819 / 3D= 12629 / Mem = 3002 / Disk = 13751 / Single Thread Mark = 2368 [10.23.18]

HP z420_3: (2015) (R11) Xeon E5-1650 v2 (6C@ 4.3GHz) / z420 Liquid cooling / 32GB (HP/Samsung 4X 8GB DDR3-1866 ECC registered) / NVIDIA GeForce GTX 1060 6GB/ Samsung 860 EVO 500GB + HGST 4TB / ASUS Essence STX / Logitech z2300 2.1 / 600W PSU > Windows 7 Professional 64-bit (HP OEM ) > Samsung 40" 4K

[Passmark System Rating: = 5644 / CPU = 15293 / 2D = 847 / 3D = 10953 / Mem = 2997 Disk = 4858 /Single Thread Mark = 2384 [6.27.19]

[Bolis · Accepted solution]

I did a few tests in changing slots and PCIe configurations and found the brand new RTX2080 was the faulty product here. After swapping slots a few times the card failed to even be recognised by the Nvidia driver and those lovely green RTX artifacts were everywhere. I swapped the RTX2080 for a different RTX2080 from one of my partners and everything is running stable again. The RTX2080 is going for a RMA.

Thanks for troubleshooting my problems, especially BanbiBoomZ.

Cheers.

Bolis