NOTE: This post indicates the method to roll back BIOS 3.92 in HP zX20 systems to an earlier version. The 3.92 BIOS version is strongly recommended by HP to protect the system from Intel processor Meltdown / Spectre vulnerabilities. Reverting the BIOS version to restore performance therefore has an inherent risk. Anyone rolling back should understand the risk and assume the best security measures possbile including a VPN and continuous high grade malware and virus monitoring. < Good idea anyway these days.
HPeers,
For the zX20 system, BIOS version 3.92 proved extremely problematic to my principal 3D CAD modeling system, a z620. However, after three months, the problems have been successfully solved.
The ostensible purpose of the strongly recommended BIOS 3.92 was to protect systems from the Intel CPU Meltdown / Spectre vulnerability to a kind of back channel memory attack.
I had three HP z-series at the time: z620_2, z420_3, and z420_1, all running v 3.91.
In the first instance, the upgrade on z620_2from v3.91 to 3.92- done in the identical way used at least ten times previously, corrupted the BIOS and the system became unusable, even to diagonise the problem. As the crisis recovery jumper was undocumented in the service manual and not shown on the system board diagram, I assumed that the motherboard was ruined.
A replacement motherboard was purchased and installed, unfortunately again using v3.92. However, the performance was so poor that I decided the motherboard was defective and returned it. This happened again with a second motherboard and that mean that v3.92 was reducung the Passmark CPU score from 17128 to 16290, the 2D on the Quadro P2000 went from 876 to 549, and 3D from 8998 to 7640. I later learned that the performance drop was more pronounced on overclocked CPU's.
Eventually, thanks to a post by forum friend SDH regarding the crisis recovery jumper, the original motherboard was recovered, but to v 3.92 and the poor perofmrnace.
Over almost a full month, several strong appeals for a simple reply from HP as to the possibility of rolling back v3.92 were ignored. At that time, all the internals of z620_2 were transferred to z420_3 (E5-1620_2) which was running on 3.91, and the Passmark results were not far off z620_2 at it's best which had a top rating of 6322. This is z420_3 replicating z620_2:
Becaase z420_3 replicated z620_2 almost exactly except in the BIOS version, these results comfirmed that 3.92 had noticeably degraded system performance. For a 3D CAD system, the CPU Single Threaded is the most important score in this test and 2373 is by a bit, the best result ever for any of the five z-series systems. The average single threaded mark for an E5-1680 v2 is 2110. The very good Disk Mark is courtesy of a first generation HP Z Turbo Drive 256GB M.2 AHCI. The Z Turbo Drive must be highly optimized: for comparison, the Passmark average results for a Samsung 960 EVO 500GB NVMe is 14548. These results, by the way, made z420_3 the highest rated z420 in Passmark baselines.
However, z620_2 remained in limbo. Eight attempts to revert 3.92 to 3.91 in about four variations of technique failed. This gave the appearance that v3.92 must contain undocumented microcode that prevented rolling back the version.
After another query, "KelleyC" in business support relayed information that there was indeed microcode preventing 3.92 version roll back. The good news was that there was also a secret, magic button that could allow the version to be reverted to 3.91; a green jumper next to the main motherboard power connection, just above the SATA ports:
Remove the green jumper, load the new BIOS and replace the jumper. Within about 15 minutes, BIOS 3.92 was reverted to 3.91.
Three months frustration, many hours, and expense, might have been avoided by simply knowing the function of two tiny jumpers.
The test results for the restored z620_2 are quite good:
HP z620_2 (2017) (Rev 4) > Xeon E5-1680 v2 (8-core@ 4.3GHz) / z420 Liquid Cooling / 64GB DDR3-1866 ECC Reg / Quadro P2000 5GB / HP Z Turbo Drive M.2 256GB + Samsung 860 EVO 500GB + HGST 7K6000 4TB / Focusrite Scarlett 2i4 sound interface / 825W PSU /> Windows 7 Prof.’l 64-bit > 2X Dell Ultrasharp U2715H (2560 X 1440)
I think the comparative results for z420_3 and z620_2 which ar nearly as possible the same system except for the motherboard, demonstrate something I've had in mind awhile. The z420 and z629 motherboards appear to identical =- and use the same BIOS, except that the z620 has sockets to mount a 2nd processor riser- and the phantom of that socket is visible on the z420 motherboard. It seems that, all things being equal, without the extra QPI links to accommodate a second processor, a z420 will slightly outperform a z620.
Thanks KelleyC in Support for finalizing the end of the v 3.92 disaster and to forum friends Brian1965 and SDH for their advice and generous patience.
BambiBoomZ