cancel
Showing results for 
Search instead for 
Did you mean: 
Master's Graduate
Master's Graduate
480 455 25 122
Message 1 of 26
1,780
Flag Post
HP Recommended

BIOS v. 3.92 Disaster = ! and ?

Fellow HPians,

 

In March, when preparing  z620_2 (Xeon E5-1680 v2 8C@4.3GHz/  z420 liquid cooling / 64GB / Quadro P2000 / HP Z Turbo Drive 256GB + Intel 730) to add music production to it's 3D CAD and graphic design duties, I updated the BIOS from v.3.91 to v.3.92, which was only a few weeks old at the time.  This was done using the .EXE in Windows as I've done  many times before on the five HP z-series I've had.

 

During the upgrade, the progress bar showed the 3.91 BIOS saved and 3.92 flashed successfully. However, quite soon the screen went black and after a pause, the fans began spinning up and I shut the system down on the power switch.

 

Looking in the user manual, there was a crisis recovery jumper mentioned for the z820, but not for the z420 or z620.  Assuming the motherboard had been ruined, I bought a replacement and - stupidly in retrospect-installed 3.92.

 

The system ran, except that the Quadro P2000 that had tested in Passmark Performance Test as high as 9030 in 3D, was now testing as low as 7646 and the 2D that had been 874 was reduced to 549

 

As the Z Turbo Drive Windows installation was high mileage and the configuration of a new UEFI Windows 7 installation was proving elusive without buying a new copy of Windows, a Samsung 860 Evo 500GB was installed with Windows 7 from the orignal HP recovery disks, plus all the software reloaded.

 

The results from the P2000 were improved a bit- 8038, 8126, and etc., but the the 2D remained very poor.

 

The P2000 was exchanged at PNY for a new one who were very good about it, but the results were only very slightly better. TO run while the P2000 was away, I decided to give GTX a try and bought an MSI GTX 1070 Gaming X- which is on the edge of not fitting in the case- and while this had quite good in 3D- 11350, the 2D was also terrible for that GPU on a system tunning at 4.3GHz - 579.

 

Meanwhile, having learned that the z620 has a crisis recovery jumper, I was successful on the first try using the .bin file on a FAT32 USB but- most stupidly of all, recovered the orignal motherboard to 3.92.

 

Susprise! the performance was the same. At the same time, hearing rumors about performance issues with 3.92, I decided to reflash z620_2 to 3.91.  However,  five attempts, from Windows in Safe Mode and from USB, in the first 3 tries had an "unrecoverable error" that was "unknown" and in Safe Mode the flash was shown as sucessful, yet restarting revealed that the BIOS remained v. 3.92. Of course, the graphic results were the same. The subsequent two attempts to revert to 3.91 produced the same result.

 

I decided to try the P2000 in one of the 3.91 (z420_1 and z420_3) systems and upgraded z420_3 from a Xeon E5-1607 V2 (4C@3.0) to E5-1620 V2 (4V@3.7/3.9) / 16GB /changed the GTX 660Ti to the Quadro P2000 / Samsung 860 Evo 500GB + Intel  and the results were revealing:

 

z420_3_E5-1620 V2_P2000_860 Evo_3D TST 1_4796_5.18.18.jpg

 

Remember that the 860 Evo was straight from z620_2.  The 8419 3D and , especially 755 2D were more like it, considering that is comparing the results on 3.9GHz to 4.3GHz.

 

Subsequently, I checked Passmark results for the P2000 and was surprised to see that the 2D results vary so considerably. The top 2D mark is 1354 on i7-7700K @ 5GHz, but the shock is the bottom mark of 133 on a Dell Precision T5820 / Xeon W-2133 3.6/3.8 - a currently available system, a Threadripper 1950X @ 4.1GHz = 267, and HP z440 E5-1630 v4 @ 3.7/4.0GHz:  682- not terrible, but not up to z420_3's E5-1620 v2 of 2013.  A very interesting result was an i7-7820X 8C@3.6/4.5GHz/ 64GB/ Samsung 960 Pro M.2 and the P2000 results are 2D=634 and 3D=7944. That is also not as good as z420_3 with an E5-1620 v2.  Of course, the HP 3.92 BIOS can not be at fault in the non HP systems, but:

 

Q1: What is going on?

 

As the two z620 motherboards using 3.92 have the same results, whereas the 3.91 results were proportionally similar to the original z620_2 @ 4.3GHz with 3.91, I'm convinced that 3.92 is at fault.  This may have some relationship to Spectre/Meltdown considerations.

 

A straight question for HP:

 

Q2: Can BIOS V. 3.92 be reverted to V. 3.91 or NOT?

 

BambiBoomZ

25 REPLIES 25
Distinguished Professor
Distinguished Professor
2,768 2,729 240 538
Message 2 of 26
1,749
Flag Post
HP Recommended

BIOS v. 3.92 Disaster = ! and ?

first of all, if you would bother to look at the bios history (HTML) file that is included with the sp84133

 

you would have had your answer , hp thoughtly provides this file in all of their workstation bios updates

 

this html file is also auto poped up during a flash from within windows, but again you have to actually use the button and read/move/read  to the end of the file in order to read what hp says on bios downgrading

 

from the HP history html file:

 

History

Version 3.92

Updates Intel processor microcode to 0x42C for Ivy Bridge-EP processors, 0x713 for Sandy Bridge-EP C2/M1 processors and 0x61C for Sandy Bridge-EP C1/M0 processors.

Includes enhancement to address CVE-2017-5715.

Improves security of UEFI code and variables.

HP strongly recommends promptly transitioning to this updated BIOS version.

 

Trusted Execution Technology (TXT) Enabled:

TPM1.2 PCR0 = F075B209C79AC56576B95F9B111941E96EE1C11E (SHA1)

 

Trusted Execution Technology (TXT) Disabled:

TPM1.2 PCR0 = 826FED70FD9663F791D9CB001CF01AE20AB465DF (SHA1)

 

Version 3.91

This release addresses “Microsoft Secure Boot Security Feature Bypass” as described in CVE-2016-3287 and CVE-2016-3320. HP strongly recommends promptly transitioning to this updated BIOS version, which supersedes all previous releases.

Updates UEFI Network Stack to resolve issue when attempting to PXE boot with Secure Boot enabled.

 

Version 3.90

WARNING: After installing this BIOS version onto the system, prior BIOS versions cannot be installed onto the system.

This release improves security of UEFI code and variables and includes mitigation for the following vulnerabilities. HP strongly recommends promptly transitioning to this updated BIOS version, which supersedes all previous releases.

Intel-TA DOC #548744, VU#631788, VU#766164, VU#976132

Updates Legacy PXE OROM to 1.5.56.

Adds F10 Setup option to Restrict USB Devices based on a variety of options including the ability to only allow keyboard and mouse.

Adds F10 Setup option to disable error reporting if system has insufficient Option ROM space.

Fixes an issue where system would not boot when cards with large MMIO are installed.

 

 

while it's possible that hp missed documenting that you can't downgrade the 3.91/3.92 release, it's been a while since these files were released and HP is very good at correcting documentation in newer upgrades and revising older documentation so it appears that hp allows a downgrade which implys that something you are doing is the cause of a failed downgrade

 

last, th v3.92 bios update does not really slow the system down (unless all you use/run is benchmarks) in common usage senarios, i know since i have a z820 and i ran a few tests using real life apps and the impact is mimimal for desktop users, on my server however it's a diffrent story, some functions are slower but again unless your server is  at 70% usage or more, the bios fix as a slowdown is rather small overall can rarely be seen in most day to day usages

 

 

 

Reply
0 Kudos
Master's Graduate
Master's Graduate
480 455 25 122
Message 3 of 26
1,726
Flag Post
HP Recommended

BIOS v. 3.92 Disaster = ! and ?

> Firstly, if any Forum member has a proven method of reverting BIOS v. 3.92 to v. 391, including virtual DOS, that would be most appreciated.

 

 

DGroves,

 

As BIOS updates are so importantly fundamental, I do read all avaiable information.  My query centers on the dissociation between there being no warning  that the v. 3.92 update was irreversible (as is the case with v. 3.90), and the fact of all my attempts to revert from 3.92 to 3.91 that were reported as "successful" did not actually accomplish the change.  Also, these attempts were following many successful updates in the five office z-series workstations and done as carefully as possible increased the uncertanty.  There is a method that uses a form of what appears to be a kind of temporary, virtual DOS- similar to WINPE? and I]ll look into that.

 

In terms of perofrmance problems, as I work with large- 200MB 3D models and 120-page documents comtaining photos that are edited on 150-300MB files, any performance reduction is important.  This system was already on the edge of usuability and when I've spent quite a bit on two GPU's, Xeon E5-1680 v2, 64GB RAM and SSDs also not being free, it's a waste not to pursue the best possible performance.

 

Meanwhile, last evening I all the components from z620_2 including the liquid cooler were installed into z420_3, which is running v. 3.91.

 

Passmark results for z420_3: Xeon E5-1680 v2 8C@ 4.3GHz /  Liquid Cooling / GTX 1070 8GB / 16GB RAM/ Samsung 860 Evo 500GB+ Intel 730 480GB / Win7 64:

 

 

z420_3_E5-1680 V2_GTX 1070_860 Evo_2D_43x + 187.50_ 2D TST 3_5498_5.19.18.jpg

 

This experiment is an near to having two systems as similar as possible except for the BIOS bersion that I can quickly accomplish. And, it does show a siginificant improvement in the 2D performance.

 

Z620_2 @ 4.3GHz, using the GTX 1070 produced a Passmark 2D mark of 574 and 3D of 11385, and the low 2D was immediately noticeable in work. The 791 is still not terrific, but 791 is 137% of 574. The 2D and 3D perfomance is related to single thread performnace. Z420_3 is running a Passmark single thread rating of 2261.  Z620_2 had a single thread at up to 2354, but I decided to reduce the O/C voltage to a maximum of 1.4 to reduce stress on the CPU and in will run it 4.1GHz when CPU rendering. For comparison, an i7-8700K, running at 5GHz could have a single-thread of 2900.

 

This configuration has only had a bit of testing with various programs and project files, but is noticeably better and acceptable. 

 

BambiBoomZ

 

 

 

Reply
0 Kudos
Distinguished Professor
Distinguished Professor
2,768 2,729 240 538
Message 4 of 26
1,716
Flag Post
HP Recommended

BIOS v. 3.92 Disaster = ! and ?

a few points to consider

 

1. according to the HP documentation, (which i also posted here) you can downgrade from 3.92 to 3.90 anything below that is not allowed since i am not there to see what is going on in reguards to downgrading your system i have refraned on making any specific comments as to why you are unsucessful. i will also point out that if this is so important to you, you do have the option of paying HP for a support call if out of warranty and speaking with a hp tech about your issue with downgrading

 

2. i also works with video and raw files on my z820 and unlike you i actually used some of my previous projects to benchmark the changes the spectre bios fix caused on my system.  I find that running the actual programs i use on the system makes for a benchmark/test that is indicative of my real world usage rather than a synthetic benchmark and in my case the slowdown is around 1-3 percent. i would bet if you had also done this method of testing you also would have seen the slowdown is not really a concern to go nuts over like you are doing

 

3. if your system was/is borderline in performance for your needs, then you should have been considering a system upgrade i am currently lookinto if it's cost efective to jump from the z820 direct to a G4 based system or just get a z840 to work with the z820

 

4. your test using one program to get a result in 2d performance is not a statically valid test,.. to be a valid test you will need to use several programs, and again benchmarking programs are not the way to get a real world indication of performance. testing using the applications you use day to day is a much better way to get a real world comparison

 

5. i'm currious just what applications do you use under windows that call 2d functions i can't think of any modern application that makes direct use of 2d functions in a "major" way on windows  7/10 short of placing text on the screen

 

6. if you google "passmark 2d video test low"  you will see pages and pages on how the  passmark 2d test is unreliable and using this test as a reasion to go nuts based on it's flawed benchmarking is counterproductive

 

last,...... this is for people flashing their bios,... patience is the key here.....do not interupt the flash by powering off !! let the process complete!,..... as this user found pulling the power resulted in a corupted bios

 

================================================================

from the "op's" first post describing what he did during the bios update:

 

During the upgrade, the progress bar showed the 3.91 BIOS saved and 3.92 flashed successfully. However, quite soon the screen went black and after a pause, the fans began spinning up and I shut the system down on the power switch.

================================================================

 

if the user had simply waited 30/45 sec after the screen went black the bios update would have finished and then the normal boot screen would have shown. when i did my z820 to  bios  v3.92 it also went black and the fans ramped up. I however simply waited till the bios finished updating and returned me to the boot screen. as such i did not have to try recovering from a failed flash.

 

Reply
0 Kudos
Master's Graduate
Master's Graduate
480 455 25 122
Message 5 of 26
1,705
Flag Post
HP Recommended

BIOS v. 3.92 Disaster = ! and ?

DGroves,

 

I appreciate your efforts in replying, but prefer that the tone is less personal.

 

To repeat, I previously moted and am pleased that that HP does not list 3.92 as irreversible,  but what is the method to revert? Has anyone reading changed back? 

 

Benchmarks are analytically useful  in objectively comparitive terms. When I notice reduced performance in the applications I use: AutoCad, Revit, Sketchup, Adobe CS6 MC, Corel Technical Designer, and  WP Office that is subjective, and in that case I check it using bnechmarks. As you mention, the difference has to be quite a bit to be noticeable, but in my work it becomes quite obivous with particular processes, especially in the application of shadows to large 3D  models. This system is to soon add a high-performance MIDI/music production using Ableton Live and low-latency in 64- tracks is demanding.

 

Except for WP Office, these include intense 2D and 3D applications particularly photo effect processing, 3D modeling and CPU rendering in which the best performace is essential. The basis for performance is subjective: I open and manipulate copies of  files of the highest complexity in each of the most demanding programs. The previous benchmarks in comparison with current tests are informative.  The benchmarks provide a link between the subjective observation and supporting, comparative objective data to look for the particular situation.

 

Passmark tests are useful as each category of test: CPU, 2D, 3D, Memory , and Disk are in turn comprised of sub-tests. The CPU single thread rating is relevant to about every application, 2D tests are revealing of CPU processes relating to display functions, whereas 3D identify conditions in 3D.  Any benchmark is the result of experience- knwing what are the average, previous, and current results. I believe that Passmark weights the system ratings to an experiential perofrmance in gaming,  but as gaming. as for 3D CAD emphasizes the effeiicency of polygon positional calculation and rendering of same, I can understand a system comparitively accordingly. I have used Quadros exclusively for fifteen years and undersand that the Quadro approach of image quality as compared to gaming frame rates means lower 3D marks. 

 

As for 2D functions, how about AutoCad, After Effects, and CPU rendering?

 

The BIOS installation failure mode in which the screen goes black and the fans spinning  up is the same I experienced and not the result of premarturely powering down.  The fans and  the liquid cooler were accelerating to maximum- screaming, the power LED flashing and I was concerned that the CPU and/or cooler may be damaged. This is quite different from the relatively peaceful successful installations as well as the "successful" installations I've done that did not actual revert the BIOS version. However, - eventually-the crisis recovery jumper saved the situation.

 

My system was not at all borderline when the Passmark system rating was 6322 but at 5498, it's nominal. Inthe absence of  satisfactory resolution of z620_2, when the i7-9700K and Z390 platform arrive in a couple of months- 8- core @ 5GHz will be a direction to consider. Another consideration is the i7-7820X - more PCI lanes and quad channel memory. If HP makes a future "Z2" with an i7-7820X on Z390, I'd be very pleased to Beta test it, and potentially it could have a higher single thread performance than any Z4 or Z6. 

 

BambiBoomZ

 

 

Reply
0 Kudos
Distinguished Professor
Distinguished Professor
2,768 2,729 240 538
Message 6 of 26
1,701
Flag Post
HP Recommended

BIOS v. 3.92 Disaster = ! and ?

I mean no disrepect to you, and if you thought it was a attack on you it was/is not but i strongly disagree that the spectre fix is a disaster that causes the end of computing.   your posts here make it seem that the bios update ruined your system 

 

while a static bench marking program has it's use they are very limited in actually conveying real time benchmarks that reflect actual usage of a users computer

 

i do some 4k video editing on the z820 and will be doing more, which is why i am comparing the return on investment of buying a new system 

 

based on all of the things you say you are doing/using on the z620 you might want to consider another system and split the applications between them

 

your comments about the fans going full speed/red light flashing  during the bios update to v3.92

 

this is normal for this update,.... ask around and other z620/z820 users will confirm that this update will do this and then after a 30-45 sec pause switch to the normal bios startup screen. I maintain a local network of over 500 HP workstations and this behaviour was the same on all of the z620/z820 systems updated to 3.92

 

i don't think hp is going to change their stance of using only xeons in their workstation line in favor of core i7 based cpu's

 

fyi, i downloaded the latest passmark and on my z820 with a nvidia 1080 and a HP 256GB turbo drive SSD using the sm951 (sata) here's the results

 

passmark rating: 3905.7

 

cpu mark 15129.8

 

2d graphics mark 447.9

 

3d graphics mark  7983.2

 

memory 1861.3

 

disk mark  10736.8

 

 

 

note the 2d mark score this does not suprise me as many 2D operations are not accelerated by the video card

 

 

i know the "OP" knows this but for the other people out there

belive it or not, the intel intergrated HD video will post better scores, but does this mean that you want to run your applications on the intel HD graphics or a dedicated video card? the 2d performance of a video chip is overshadowed by many other factors resulting in the  dedicated card being a much better overall device to use

Reply
0 Kudos
Master's Graduate
Master's Graduate
480 455 25 122
Message 7 of 26
1,682
Flag Post
HP Recommended

BIOS v. 3.92 Disaster = ! and ?

DGroves,

 

1.  To clarify, the internals of z620_2  running BIOS v. 3.92: CPU(Xeon E5-1680 v2 8C), GPU (GTX 1070 8GB), and drives (Samsung 860 Evo 500GB) were transferred to z420_3 running BIOS v. 3.91. 

 

In both test series, the CPU  clock speed was 4.3Ghz. Comparing the results of the two systems replicated as nearly as possible (z620_2 has 64GB of RAM and z420_3 has 16GB) the Passmark 2D mark on the system running 3.91 improved from 579 to 794.

 

Running z420_3  using a Xeon E5-1620 v2 (4C@ 3.6 /3.8Ghz), and BIOS v. 3.91, the 2D mark of a Quadro P2000 5GB improved from 549 to 743.

 

In all tests, the replicated systems demonstrated a significantly improved 2D score using BIOS v3.91.

 

2.  The failed BIOS update of z620_2 was not the result of premature shutting off of power. The extreme acceleration of fans and the liquid cooler is not consistent with a successful BIOS flash.

 

3. The statement, that HP will never, "change their stance of using only xeons in their workstation line in favor of core i7 based cpu's" is incorrect.  See the Processor list for the z240 workstation:

 

https://store.hp.com/us/en/pdp/hp-z240-tower-workstation-l8t12av-1--1

 

Among various Xeon E3's, the z240 could be configured using:

 

Intel® Core™ i5-7500 Processor (3.4 GHz, up to 3.8 GHz w/Turbo Boost, 6 MB cache, 2400 MHz, 4 core) + Intel HD Graphics 630

Intel® Core™ i5-6500 Processor (3.2 GHz, up to 3.6 GHz w/Turbo Boost, 6 MB cache, 2133 MHz, 4 core) + Intel® HD Graphics 530

Intel® Core™ i7-7700 Processor (3.6 GHz, up to 4.2 GHz w/Turbo Boost, 8 MB cache, 2400 MHz, 4 core) + Intel® HD Graphics 630

Intel® Core™ i7-6700 Processor (3.4 GHz, up to 4 GHz w/Turbo Boost, 8 MB cache, 2133 MHz, 4 core) + Intel® HD Graphics 530

Intel® Core™ i7-7700K Processor (4.2 GHz, up to 4.5 GHz w/Turbo Boost, 8 MB cache, 4 core) + Intel® HD Graphics 630

Intel® Core™ i7-6700K Processor (4 GHz, up to 4.2 GHz w/Turbo Boost, 8 MB cache, 4 core) + Intel® HD Graphics 530

 

4. As for benchmarks, a dismissive attitude needs an analytical qualification. No single test is the perfect description of performance, but the sub-routines of various benchmarks evaluate seperate performance components are revealing of relative experiential performance. The weighting of performance measurement in my experience of perhaps sixty or eighty Passmark runs has in all cases corresponded to experience using applications. 

 

These benchmarks have been very helpful in choosing components. In the example of your z820 Passmark results, for example:

 

z820 Workstation: 2X xeon E5-2640 / 64GB / GTX 1080 / Samsung SM951 AHCI

 

Rating = 3905 / CPU = 15130 / 2D= 448 / 3D= 7983 / Memory 1861 / Disk = 10737

 

Immediately, the deficient performance of the GTX 1080 is noted. The average 3D for the 1080 is 12266, but that  is including overclocked and SLI systems.  However, it does indicate that the GTX 1080 in a 2.5 /3.0 GHz system is wasting a significant proportion of it's potential performance.  The memory and ZTD marks are also poor. The GTX 1080 in this example is being throttled by low CPU clock speeds. The E5-2640 should be a superb and ultra-reliable server CPU, but is insufficient for current visualization applications.

 

For comparison, these are the results for a z620 using a pair of E5-2690's (8C@ 2.9/3.8GHz) / 32GB RAM / GTX 1070 /Samsung SM951 NVme

 

Rating = 5248 / CPU = 20345 / 2D= 629 / 3D= 10147 / Memory 2285 / Disk = 14537

 

The relative Passmark average single-thread marks:

 

E5-2640 = 1461

E5-2690 = 1831

 

> is a strong factor in the comparitive rating of 5248 to 3905. And, in every application except CPU rendering, the single-thread performance is primary- far more than thread count.

 

Compare the 3D of 7983 on a GTX 1080 to 10147 on a GTX 1070 and consider that the difference in cost of a 1080 to 1070 would more than pay for the pair of E5-2690's. So, without additional cost:  +4 -cores with a + 34% CPU clock cycle rate, a 25% improvement in single-thread performance, and a 27% 3D improvement is possible.

 

This is ample demonstration of the way in which benchmark analysis can economically correct poor component balance.

 

For almost all graphics and 3D applications, the single-thread performance is the most important feature.  Also, except for CPU rendering and server use, the peak CPU multi-threading efficiency is 5-6 cores, which is the reason for my changing from a 2X 2690 system:

 

z620_1 : 2X E5-2690  8C@2.9/3.8GHz / Quadro K2200 /64GB / ZTD 256GB

 

Rating = 5675 / CPU = 22625 / 2D= 815 / 3D= 3579 / Memory 2521 / Disk = 12639  Single- thread = 2044

 

> to a single 8-core:

 

z620_2: E5-1680 v2  8C@4.3Ghz / Quadro P2000 / 64GB / ZTD 256GB

 

Rating = 6322 / CPU = 17128 / 2D= 852 / 3D= 8992 / Memory 2790 / Disk = 14056 Single- thread = 2339

 

While the CPU mark is indicating that z620_2 is running fewer clock cycles/unit time, the single- thread of 2339 is the critical factor.

 

The benchmarks are meaningful only in comparison, but the configuration of the Passmark test is such that the compartive results are an objective basis for component choices oriented towards a performance experience in day to day use.

 

 

BambiBoomZ

 

Reply
0 Kudos
Highlighted
Master's Graduate
Master's Graduate
336 322 20 99
Message 8 of 26
1,667
Flag Post
HP Recommended

BIOS v. 3.92 Disaster = ! and ?


@BambiBoomZwrote:

Fellow HPians,

 

In March, when preparing  z620_2 (Xeon E5-1680 v2 8C@4.3GHz/  z420 liquid cooling / 64GB / Quadro P2000 / HP Z Turbo Drive 256GB + Intel 730) to add music production to it's 3D CAD and graphic design duties, I updated the BIOS from v.3.91 to v.3.92, which was only a few weeks old at the time.  This was done using the .EXE in Windows as I've done  many times before on the five HP z-series I've had. {cut]

BambiBoomZ


 

It is a fact that performance using the updated Spectre microcodes takes a hit.

I have first hand experience of this on both the P9X79 Pro + E5-1650v2 and a recent Maximus VIII Impact Z170 + i5-8600k (modded bios + CPU hardmod - this one was fun) system I put together. On both I have reverted to the pre-Spectre fix codes. SInce you are running higher than stock clocks, there is one aspect of my findings that might be relevant for your case:

- at stock clocks, the spectre fix microcodes deliver inferior performance but overall its close enough to pre spectre codes - nothing to make a big fuss about

- the situation however deteriorates rapidly when overclocking - higher clocks do not scale with increased performance as well as on the pre spectre microcodes

BOttom line - the spectre fix is an afterthought on the part of Intel and the CPU is not running as was originally envisaged by the manufacturer. You gain something but also loose something. For my usage scenario, I am more than happy to forego the 'security' of the Spectre patch and just have better performance, but in a corporate or more security concious scenario people's milleage may differ.

 

On the subject of overclocking, I have something you might find useful - ability to change Core ratio multiplier without the hassle of opening XTU each time, and having it running (and crashing) in the background.

This method uses a powershell script to change core ratio on all cores at startup and relies on I) the fact that XTU has access to the MSR registers and ii) that it accepts command line input, and then kills the XTU process.

 

It goes as follows:

Files: https://files.fm/u/ut43xqzk

1) download and install XTU (this you have already 😉 )
2) open Powershell and type Set-ExecutionPolicy RemoteSigned - choose All (a) - one time thing
3) create a folder on your c: drive - OCXTU (i.e. C:\OCXTU)
4) copy the two attached files from inside the .rar archive to that directory
5) create a link in your windows startup folder to oc.vbs

Now every time you start up your PC, core ratio multiplier on all cores will be set to 43x and XTU will not be running (and crashing) in the background. The script kills the XTU process right after the multiplier has been changed. If you want to change the multiplier, edit the oc.ps1 file and change the 43 to whatever core ratio you want.
You don't have to add it to startup if you do not want to - you can keep the os.ps1 file on your desktop and run it manually if you wish - or have several .ps1 files with different oc settings. The purpose of the oc.vbs file is to make it a seamless / automatic process.

Give it a go and let me know in case of any problems.

 

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.

Master's Graduate
Master's Graduate
480 455 25 122
Message 9 of 26
1,649
Flag Post
HP Recommended

BIOS v. 3.92 Disaster = ! and ?

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

 

 

 

Reply
0 Kudos
Distinguished Professor
Distinguished Professor
2,768 2,729 240 538
Message 10 of 26
1,637
Flag Post
HP Recommended

BIOS v. 3.92 Disaster = ! and ?

your comments about my  and your system's passmark score needs some clarifications

 

you state that your passmark score increased on the older bios vs the newer one ok fine, but just what does that mean in the real world, does a increase from the  2d score of 579 to 794 actually do anything  overall in the software you run?

i'm not nickpicking, just actually currious if running your day to day applications actually show a diffrence that can be seen in day to day usage 

 

this is why i have said numerous times that benchmarking the software you actually run before and after a change is a much better way to see if the change is actually usefull to you i personally find that some overclocks/changes increase the power consumption/stability of a system so much so that it negates the  increase in the applications overall speed resulting in a net decrease in overall profits or crashes/locks up requiring constant reboots and lost time

 

 

your comment that you moved your applications from one z620 model workstation to another z420 model and got a faster 2d score so the bios update is at fault is flawed, you have to benchmark between the same hardware for it to be valid i mean if i did 0-60 in a VW Golf, and then did the same in a Golf Turbo GTI faster, so there must be a problem with the non turbo golf would you think that was a valid comparison?

 

 

I never said (HP will never, "change their stance) rather i said "i don't think hp is going to change their stance)

please do not misquote me giving the impression i was stating a fact rather than my  personal observation

you are correct that hp does use some i 7 cpu's now  in their low end workstations and i was not aware of this,  i should have said that hp does not use consumer cpu's currently in their mid to high end workstations  that i have seen and is unlikely to in my opinion, but as you have noted hp  does so now in the low end so they might in the other workstations systems,.... i stand corrected

 

i  received the z820 and 1080 cards from work, where i had a choice between the 1080 or slower cards, which would you pick? as i noted in my previous post i am looking into upgrading the z820 to a x99  based chipset or better (z840)

and as such the 1080 card will most likely have a longer useable lifespan compared to the 1070 i could have also picked

 

in reguards to my cpu speed, the software i run is mostly more receptive to more cores than Mhz, and as such the 12 cores + hyperthreading e5-2640 2.50GHz  is quite fast in compleating it's tasks although a faster multicore would be nice it has to be cost efective and be a good ROI (return on investment) i'm not going to spend 400/500 to buy the 2.80mhz  e5-2667 as the speed increse in the apps i use will not increase to the point where it is cost efective

 

last, you observations on my z820 memory speed is slow compared to the passmark listings is not really correct, as the z820 is able to run a faster memory bus with limitations on the number of ram modules that can be used,.. also the type of ram used ECC/ECC Registered/Unbuffered ram will also determine the memory bus speed so comparing my ECC   ram to systems using faster unbuffered ram  is again not a valid comparision test

 

• 2 GB and 4 GB PC3-14900E 1866MHz DDR3 Unbuffered ECC DIMMs

• 4 GB, 8 GB and 16 GB PC3-14900R 1866MHz DDR3 Registered DIMMs

• 32 GB PC3-14900L 1866MHz DDR3 LR DIMMs

• 1.35V and 1.5V DIMMs are supported, but the system will operate the DIMMs, safely, at 1.5V only

 

finally,..... this topic has gotten away from it's original posting which was that the hp spectre bios update was a disaster

and my reply that in my opinion it was not a diisaster unless i see something directly related to this sublect on this thread i will skip posting as it's becoming to long and off topic

Reply
0 Kudos
† The opinions expressed above are the personal opinions of the authors, not of HP. By using this site, you accept the Terms of Use and Rules of Participation