Re: BIOS v. 3.92 Disaster = ! and ? - Page 2 - HP Support Community

MtothaJ · ‎05-22-2018

@BambiBoomZwrote:
MtothaJ,

This may be the exact information I need.

I'd appreciate a clarification on a key aspect: what you mention "It is a fact that performance using the updated Spectre microcodes takes a hit." and "On both I have reverted to the pre-Spectre fix codes." In reference to "microcodes" to which specific "pre-Spectre fix codes" are you referring?"

The comments referring to a greater negative effect on overclocked systems agrees with my observation of certain current systems such as i7-7820X and i7-8700K running at 5+GHz but that appear to have especially low 2D benchmark results- CPU -related , while the 3D- GPU-related is closer to standard.

All along I've assumed this is related to the 3.92 BIOS, but if it is firmware of chipset driver-related that changes everything.

Thanks!

BambiBoomZ

Below some info on the Ivybridge-E microcodes for the Xeon E5 v2 CPU's:

Z420 / Z620 3.91 bios (pre Spectre Fix):

║#│CPUID│ Platform ID │Revision│ Date │Type│ Size │ Offset │
╟─┼─────┼────────────────┼────────┼──────────┼────┼──────┼────────┼
║│306E4│ED (0,2,3,5,6,7)│ 427 │2014-04-10│PRD │0x3000│0x580060│

Z420 / Z620 3.92 bios (post Spectre Fix):

║#│CPUID│ Platform ID │Revision│ Date │Type│ Size │ Offset │
╟─┼─────┼────────────────┼────────┼──────────┼────┼──────┼────────┼
║│306E4│ED (0,2,3,5,6,7)│ 42C │2018-01-25│PRD │0x3C00│0x580060│

Last pre Spectre Ivybridge-E microcode - this is the one I use on Asus P9X79 Pro:

# │CPUID│ Platform ID │Revision│ Date │Type│ Size │ Offset │
╟──┼─────┼────────────────┼────────┼──────────┼────┼──────┼────────┼

║ │306E4│ED (0,2,3,5,6,7)│ 428 │2014-05-29│PRD │0x3400│0x790460│

Changing microcodes is simple enough, but with the various bios protections HP has in place you need to have a failsafe way to recover in case things go wrong. I would think getting the bios socketed is your best bet if you still want to get a few more years usage out of your machine - could then add NVMe and do a few other mods.

BambiBoomZ · ‎05-22-2018

MtothaJ,

Much appreciated.

The rewriting of microcodes is a new realm for me and after the failure of a conventional BIOS update - one that I'd done many times, and the repeated failure to revert the BIOS from 3.92 to 3.91, I'm reluctant to try this appraoch.

One last try: Can anyone here verify that BIOS v. 3.92 may be changed back to v. 3.91 conventionally?

I suppose my fundamental question remanins as to whether it is possible to simply revert to 3.91 and in that, restore the performance? The 2D (Passmark) on the E5-1680 v2 had a 16% reduction and that was my primary concern.

The internals of z620_2 have been transferred to z420_3 running 3.91 and the results so far seem acceptable. I'm condidering cutting my losses and selling z620_2 as a barebones system- case, chassis, and motherboard only- very few workstation users I think would be overclocking, but that means a loss plus having to buy +48GB of ECC unbuffered DDR3-1866 as z420 's don't run registered memory.

Another lesson- one I expected- of this episode is that, with regards to the GTX 1070, even running the system at a good rate of 4.3GHz, the 3D applications - basing performance on the single-thread rates hold the performance at the same level regardless of the GPU. Based on experience running a single thread mark of 2275- 2330, I think I would need a single thread mark of 2700+, in the world of the i7-7700K or 8700K at 4.7 or 4.8Ghz to fully excercize the GTX 1070. I am looking into changing to GPU rendering, which has improved quite a bit in the last couple of years and the 1070 with 1920 CUDA cores as compared to 1024 for the P2000 would liven things up.

BambiBoomZ

MtothaJ · ‎05-23-2018

Z420 runs registered ecc just fine - had 8x8gb ecc reg in mine. But this is something you can easily verify for yourself by moving the parts around between your machines.

BambiBoomZ · ‎05-24-2018

MtothaJ,

Yes, I had thought that if z620_2 were to be sold, it would include the 64GB registered ECC, but after consideration decided to use it in z420_3.

So as to demonstrate the reltivce effect of 3.92 as compared to 3.91, z420_3 replicates as nearly as possible z620_2:

Xeon E5-1680 v2 8C@ 4.3GHz/ z420 Liquid cooling / 64GB DDR3-1866 ECC Reg / GTX 1070 / HP Z Turbo Drive 256GB + Intel 730 480GB

As the z420 and z620 motherboards are nearly identical, the difference being the sockets for the CPU riser, the only difference between the two system was the system BIOS.

Passmark results:

z420_3: Rating = 5976 / CPU=16690 / 2D= 779 / 3D=11476 / Memory= 2823 / Disk=14290

> which makes z420_3 the highest rated z420 on Passmark.

In all these comparisons, the focus is on CPU-related performance- the CPU and 2D marks and the above test which raised the 2D of the GTX 1070 from 579 to 779 describes and iportant effect of 3.92 in comparison to 3.91.

However, I was unhappy with the CPU rating and the Single Thread Mark of only 2204 as compared to 2339 on z620_2. z420_3 , after changing tot the E5-1680 v2 had a startup error, "Front case fan not detected". Neither Z420_1 nor z420_3 have a front case fan and z420_3 It appears that the more power hungry E5-1680 triggered the error.

There was a possibility that the the interruption of the startup by the fornt fan error was affecting performance, so yesterday I added a front case fan to z420_3. The MSI GTX 1070 very nearly doesn't fit in the case- so when it was out I decided to replace the Quadro P2000, as that is my primary GPU. I also considered that the Z Turbo Drive had problems, having been run quite a lot, changed a number of times and tried in four systems, that the freshly configured Samsung 860 EVO 500GB was installed.

Thus configured in the form nearest to the working state:

Z420_3: Xeon E5-1680 v2 8C@ 4.3GHz/ z420 Liquid cooling / 64GB DDR3-1866 ECC Reg / Quadro P2000 5GB / Samsung 860 EVO 500GB + Intel 730 480GB

The overall rating is lower as the because of the relative transfer rates of the Z Turbo Drive and the 860 EVO, but the key parameters: CPU, single Thread rating, 2D (679 to 807) and 3D ( increased from 7795 to 8464) performance is now in the acceptable range. The single thread rating of 2357 is particularly welcome, surpassing the earlier top mark of 2339. Even the memory mark improved 2823 to 3018, the best I've seen.

A couple of hours using Sketchup and AutoCad have shown z420_3 is back to form.

I'm installing a Xeon E5-1607 v2 , 1X 4GB of RAM, and the original, slowest on record Seagate 500GB HD- Disk mark 574 ! into z620_2 and can fuss with trying to revert it 3.91.

Thanks!

BambiBoomZ

ShadowX360 · ‎05-24-2018

I updated my bios on a Z420 to 3.92 and just got the same problem. The HPFlash program froze and restarted. Got no video output, no power to USB devices. Fans kicked on. Left it there for 60 minutes before I hit the power button. There's an issue with this update. Now I'm stuck trying to recover this mess.

EDIT: There's another report of same BIOS version on an HP Z420 with same results on experts-exchange. No one figured out how to actually get the crisis jumper working on a Z420 - the jumper is there but zero reports of anyone successfully using it to re-flash the BIOS despite many attempts. Looks like I'm screwed. The sudden power loss, then black screen with no output during the bios flash - looks like it was bricked by the time it went black, and waiting the 60 minutes did nothing.

MtothaJ · ‎05-25-2018

@BambiBoomZwrote:
MtothaJ,

Yes, I had thought that if z620_2 were to be sold, it would include the 64GB registered ECC, but after consideration decided to use it in z420_3.

So as to demonstrate the reltivce effect of 3.92 as compared to 3.91, z420_3 replicated as nearly as possible z620_2:

Xeon E5-1680 v2 8C@ 4.3GHz/ z420 Liquid cooling / 64GB DDR3-1866 ECC Reg / GTX 1070 / HP Z Turbo Drive 256GB + Intel 730 480GB

As the z420 and z620 motherboards are nearly identical, the difference being the sockets for the CPU riser, the only difference between the two system was the system BIOS.

Passmark results:

z420_3: Rating = 5976 / CPU=16690 / 2D= 779 / 3D=11476 / Memory= 2823 / Disk=14290

> which makes z420_3 the highest rated z420 on Passmark.

In all these ocmparisons, the focus is on CPU-related perofrmance- the CPU and 2D marks and the above test which raised the 2D of the GTX 1070 from 579 to 779 describes and iportant effect of 3.92 in comparison to 3.91.

However, I was unhappy with the CPU rating and the Single Thread Mark of only 2204 as compared to 2339 on z620_2. z420_3 , after changing tot the E5-1680 v2 had a startup error, "Front case fan not detected". Neither Z420_1 nor z420_3 have a front case fan and z420_3 It appears that the more power hungry E5-1680 triggered the error.

There was a possibility that the the interruption of the startup by the fornt fan error was affecting performance, so yesterday I added a front case fan to z420_3. The MSI GTX 1070 is very large- very nearly doesn't fit in the case- that when it was out I decided to replace the Quadro P2000, as that is my primary GPU. I I also conidered that the Z Turbo Drive had problems, having been run quite a lot, changed a number of times and tried in four systems, that the frshly configured Samsung 860 EVO 500GB was installed. the 860 EVO 500GB was chosen as this system will be used for music production and it provides capacity for the MIDI samples can be on the main drive to reduce latency.

another facor I've noticed over a week' s use of the GTX 1070 is that it's behavior in AutoCad is a bit odd and in all uses, it is not perceptually faster than the P2000.

Z420_3: Xeon E5-1680 v2 8C@ 4.3GHz/ z420 Liquid cooling / 64GB DDR3-1866 ECC Reg / Quadro P2000 5GB / Samsung 860 EVO 500GB + Intel 730 480GB

Thus configured in the form nearest to the working state:

The overall rating is lower as the becuase of the relative transfer rates of the Z Turbo Drive and the 860 EVO, but the key parameters: CPU, single Thread rating, 2D and 3D ( increased from 7795 to 8464) perfomnace are now in the acceptable range. The single thread rating of 2357 is particularly welcome, surpassing the earlier top mark of 2339. Even the memory mark improved 2823 to 3018, the best I've seen.

A couple of hours using Sketchup and AutoCad have shown z420_3 is back to form.

I'Ym installing a Xeon E5-1607 v2 , 1X 4GB of RAM, and the original, slowest on record Seagate 500GB HD- Disk mark 574 ! into z620_2 and can fuss with trying to revert it 3.91.

Thanks!

BambiBoomZ

Its a nice set of results - congrats.

For Z620_v2 system I guess selling it is also an option - after all it has a fully functional board with the latest bios and the E5 v2 capable bootblock date.

If on the other hand you want to tinker with it, check out the below links:

Socket:

https://www.aliexpress.com/item/SOK-SPI-16W-test-socket-300mil-SOP16-SOIC16-SO16-ic-socket-Pitch-1-2...

Programmer:

https://www.ebay.com/itm/CH341A-24-25-Serie-EEPROM-Flash-BIOS-USB-Programming-Black-Software-Driver/...

BambiBoomZ · ‎05-25-2018

Shadow X360,

The crisis recovery jumper did work on the first try on the z620 which uses the identical BIOS to the z420, using the sequence as listed in:

https://h30434.www3.hp.com/t5/Business-PCs-Workstations-and-Point-of-Sale-Systems/Crisis-Recovery-Ju...

As you may know, the motherboards of the z420 and z 620 are nearly identical, the only difference I can see being the 2nd CPU riser socket mounted on the z620 board.

Unfortunately, as mentioned, in my example the BIOS was recovered to 3.92, which has an effect of reduced CPU / 2D performance, and so far, there is no sign that it can be reverted to 3.91. As I saw when using the E5-1620 v2 in place of the overclocked E5-1680 v2, the performance reduction is not as noticeable on systems that are not overclocked by the way.

Print out the instructions from above link and try the method as described. Given that the crisis recovery jumper is present on the z420 board and the BIOS is shared with the z620 it should be possible.

The recovery method is a bit quirky as there are so many beeps, points at which seemingly nothing and/or too much is happening, and certain actions have to take place within a certain time frame, but keep at it. - It can't be worse off!

Let us know how it goes.

BambiBoomZ

BambiBoomZ · ‎05-25-2018

MtothaJ,

Thanks for the suggestions and links. I would need to study the way the socket and programmer are used, and those are intriguing possibilities. It seems as though the programmer would be a sort of absolute method as it's possible to completely clear and then see every register after flashing.

Earlier this evening I installed the E5-1607 v2 + 1X 4Gb of RAM + a Gigabyte 256MB GS7100 fanless GPU + Seagate 500GB HD on z620_2 but couldn't make progress as the DVD drive is not recognized for loading Windows. I wanted to have a clean, Windows with no programs and no anti-virus /anti-malware to start with. The optical drive appears in Device Manager as working, and is first in the Boot Order, but does not read in Win Explorer. An odd problem, but some simple setting somewhere.

I may however, see about running the system on a mini DOS on a USB drive and try flashing from the .bin file on a second USB. The .bin file on a USB was the way the BIOS was recovered after 3.92 was corrupted originally.

BambiBoomZ

:0345!

ShadowX360 · ‎06-04-2018

@BambiBoomZ wrote:
Shadow X360,

The crisis recovery jumper did work on the first try on the z620 which uses the identical BIOS to the z420, using the sequence as listed in:

https://h30434.www3.hp.com/t5/Business-PCs-Workstations-and-Point-of-Sale-Systems/Crisis-Recovery-Ju...

As you may know, the motherboards of the z420 and z 620 are nearly identical, the only difference I can see being the 2nd CPU riser socket mounted on the z620 board.

Unfortunately, as mentioned, in my example the BIOS was recovered to 3.92, which has an effect of reduced CPU / 2D performance, and so far, there is no sign that it can be reverted to 3.91. As I saw when using the E5-1620 v2 in place of the overclocked E5-1680 v2, the performance reduction is not as noticeable on systems that are not overclocked by the way.

Print out the instructions from above link and try the method as described. Given that the crisis recovery jumper is present on the z420 board and the BIOS is shared with the z620 it should be possible.

The recovery method is a bit quirky as there are so many beeps, points at which seemingly nothing and/or too much is happening, and certain actions have to take place within a certain time frame, but keep at it. - It can't be worse off!

Let us know how it goes.

BambiBoomZ

You rescued my motherboard! I was so close to just giving up and buying a new motherboard, the key that I suspect most people are doing wrong is using too large of a flash drive. I did this procedure over a dozen times, and then eventually I saw one of your other posts about the partition size, and dug through all my drawers for the oldest flash drive I could find - a 1GB stick from 2008 and it magically worked!

I'm going to add it to all the threads I find about HP Z420's. I believe I am the first user to have ever successfully flashed and rescued a Z420 model - everyone else before couldn't get it to work right and they may have been missing that small piece.

OmTatSat · ‎10-05-2018

@MtothaJ wrote:

PS. for 'proper' overclocking you need to unlock long turbo duration, otherwise you are still limited to 130W TDP i.e. that 43X all core boost is good for short bursts, but on any longer duration tasks you will downclock considerably - try Prime95 with small FFT max heat settings and watch what happens with your clocks.

Hello, please help me to figureout how to unlock long turbo duration?

Maybe thereis some z420\z620 overcklock FAQ?

BIOS v. 3.92 Disaster = ! and ?

Create an account on the HP Community to personalize your profile and ask a question