Just to verify, whether it was just the Intel RST(e) RAID driver v15.5.0.1051, which gave my Z170 RAID0 system such an impressing WRITE performance, or the specific combination with the EFI RaidDriver v15.5.32794, I have done today an additional test. At first step I flashed a BIOS, which contains the Intel EFI RaidDriver v15.5.0.2875, then I attached my RAID0 array, restored my previous BIOS settings and booted into my yesterday created RAID0 array. Everything worked fine. Then I ran 2 benchmark tests. Both gave me Total Scores of more than 9.000 points.
Here are the screenshots showing the best results I got with the Intel RST(e) RAID driver v15.5.0.1051 in combination with the BIOS EFI RaidDriver v15.5.0.32794 (left picture) resp. with the BIOS EFI RaidDriver v15.5.0.2875 (right picture):
My conclusion: This is the proof, that it is the Intel RST(e) RAID driver v15.5.0.1051, which gave my RAID0 array the by far best WRITE and TOTAL scores I have ever seen with all my systems and tested Intel RAID drivers. In combination with the EFI RaidDriver v15.5.0.32794 this driver seems to be even more performant than with the newer EFI RAID driver v15.5.0.2875.
@hodaaort : Yesterday I have done a lot of benchmark tests with different systems and AHCI and RAID drivers and today I have completed these tests by benchmarking those brandnew drivers, which were not yet available yesterday. So it will take some time until I have customized all the related threads according the results I got and the conclusions I have drawn. Tomorrow hopefully you will not find any inconsistent statement from my side.
Both drivers belong to the same development branch. The v15.2.10.1048 is newer and probably error-corrected, but hasnât been available yesterday, when I made my tests. That is why I have done today some additional benchmark tests to include these newest Intel RST(e) drivers from the v15.2 branch into my comparison test results.
This was the message before I did my additional tests today. You will not find any statement regarding the AHCI driver v15.2.10.44 within the start post of this thread, because I have replaced its test results by the newer and better developed v15.2.12.1048.
@ all: Since I got access to brandnew Intel RST(e) drivers after having done my recent benchmark comparison tests, I have tested these previously not available drivers today in AHCI and RAID0 mode. That is why I had to update my recently published results and conclusions.
Update of the Start Post Changelog:
Tests with my Z170 system running a lot of different Intel RST(e) drivers in AHCI mode and on a RAID0 array:
new: Test results using the Intel RST(e) driver v14.8.14.1061 WHQL dated 03/09/2017 (latest from the v14.8 development branch)
new: Test results using the Intel RST(e) driver v15.2.12.1048 WHQL dated 02/02/2017 (latest from the v15.2 development branch)
removed: Test results using the Intel RST(e) driver v14.8.12.1059 dated 08/23/2016
removed: Test results using the Intel RST(e) driver v15.2.10.1044 WHQL dated 01/13/2017
removed: Test results using the Intel RST(e) driver v15.5.0.1047 Beta dated 02/22/2017
@ all: Since there are meanwhile a new Intel RST(e) drivers v15.2.14.1051 WHQL available, I have restored my previously running Z170 AHCI system, installed the new AHCI driver and done some additional benchmark tests. The results were no surprise, but maybe of interest for you.
Update of the Start Post Changelog:
Tests with my Z170 system running a lot of different Intel RST(e) drivers in AHCI mode and on a RAID0 array:
new: Test results using the Intel RST(e) driver v15.2.14.1051 WHQL dated 03/03/2017 (latest from the v15.2 development branch)
removed: Test results using the Intel RST(e) driver v15.2.12.1048 WHQL dated 02/02/2017
@all: As completion of my extensive benchmark comparison tests running a fresh installed Win10 x64 v1703 on my Z170 system, I have tested today the performance of the Intel RST(e) RAID driver v15.2.14.1051 on my RAID0 array.
Update of the Start Post Changelog:
Tests with my Z170 system running a lot of different Intel RST(e) drivers on a RAID0 array:
new: Test results using the Intel RST(e) RAID driver v15.2.14.1051 WHQL dated 03/03/2017 (latest from the v15.2 development branch)
@ all: Since I got access to brandnew Intel RST(e) drivers after having done my recent benchmark comparison tests, I have tested these previously not available drivers today in AHCI and RAID0 mode. That is why I had to update my recently published results and conclusions.
Update of the Start Post Changelog:
Tests with my Z170 system running a lot of different Intel RST(e) drivers in AHCI mode and on a RAID0 array:
new: Test results using the Intel RST(e) driver v14.8.16.1063 WHQL dated 04/10/2017 (latest from the v14.8 development branch)
new: Test results using the Intel RST(e) driver v15.2.16.1060 WHQL dated 03/30/2017 (latest from the v15.2 development branch)
removed: Test results using the previous Intel RST(e) drivers from the v14.8 development branch
removed: Test results using the previous Intel RST(e) drivers from the v15.2 development branch
Excellent! Thanks for the test Fernando. Iâve been a long time lurker but I usually only bother with newer drivers every 6 months when I can be bothered, so i was using 15.2.x.x. Since win creators update hit I noticed that storage drivers hadnât been updated so I checked you out to see if youâd tested and was surprised by your jump in IOPs with the latest drivers on your Z170
Pleased to report I have gotten around a 15% increase. Did a couple of runs on 2 x normal 850s in Raid 0 and was hitting the 7.4k something mark. Installed the new RST straight over (I leave it on all the time so it can email me if my raid goes screwy) and âŚ
Very nice bump even if I'm not using your fancy schmancy enterpise storage chipset ;)
I ran the benchmarks again as you requested and I have to say you were right. The reason I got such a boost in benchmark scores had nothing to do with the 14.8 satadriver efi rom. I made one big mistake in that I had two different Bios versions running from my dual bios switch. Had version 1.8 and shipping version 1.1 on the other side of the switch. Unfortunately one of three of the screen shots I took got corrupted and you canât see the score but you can infer that it was probably slightly worse than the other two in the set. What this all showed was that the 13.1 firmware and driver set performed on par as the 13.2 and better even than the 14.8 set. There was something inside the 1.8 bios release that enhanced the scores what ever it could be I donât know. Updated ME, bios tweaks, who knows. At this point though Iâm not sure if Iâm going back to 13.1 as I primarily use an NVME drive to boot and the raid array just hosts my steam folder and that the scores are 20 % better on the new bios which makes a case for benchmarking your bios releases as opposed to firmware releases.
Here the scores: 14.8 Pair 13.1 Pair 13.2 Pair
EDIT: Did some more benchmarks before bed comparing the 14.8 efi driver pair to the 13.1 efi driver pair on my old bios vs the old bios and here are my scores: BIOS 18 BIOS 11
13.1/13.1 5851 4650 6841 4635 6664 4500
Avg: 6452 4595
14.8/14.8
6258 4031 5802 4497 6319 4177
Avg: 6126 4235
So whats interesting to note is the vast difference in score of 4235 on the old bios with the 14.8 EFI Driver pair vs an average score of 6452 with the 13.1 Efi Driver pair. Thats like a 50% boost in performance!
All from a newer bios update!
Take home message everyone should be updating their bios to the latest or at every upgrade do a bench and keep track of the numbersâŚ
@davidm71 : Thank you very much for having taken the time, having done these extended benchmark comparison tests and having published your interesting results within this thread. Unfortunately the inserted pictures cannot be enlarged with the consequence, that it is very difficult for the viewers to identify the numbers you got. That is why I recommend to edit your last post and to attach the Anvil results individually by using the Forum software according >this< guide. If you use the âautoâ option regarding the pictureâs size, you will probably be able to insert 3 screenshots into 1 line (with the viewerâs option to enlarge them individually).
Thank you very much for this clarification and for the correction of your statement, which you had previously posted >here<. To avoid any confusion by the visitors of this Forum, it would be a good idea, if you would edit that post and either remove the misleading image or correct your wrong conclusions.
I did some more performance testing and unfortunately I do not have the screenshots ready to publish just yet because I have new fresh results. What I have found is the pairing of the 13.5.0.2164 Satadriver with 13.5.0.1056 WHQL driver best the 13.1/13.1 Satadriver combination pairing. Was interesting at first I tried the 13.6 whql driver that Intel posted on their site at they donât have the 13.5.0.1056 driver listed for some reason. Anyhow had poor results testing 13.5.0.2126 against the 13.6 driver but not against the 13.5.0.1056. So it would seem that the 13.X branch is the better choice in regards to performance like you had indicated. However Iâm going to urge to reconsider the 13.5/13.5 combination as the better more mature choice. Mind you this was all within the margin of error though I did run the benchmark three consecutive times for accuracy. Why the 14.8 branch is slower I wish I knew. Are they doing something extra slowing down the performance? Wish we had a man on the inside from Intel to comment. Anyhow here are my scores:
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Â
13.5/13.5 Scores: 6133 6696 6693  Average: 6507  13.1/13.1 Scores  6664 6840 5851  Average: 6452 Â
@davidm71 : When you are doing your extended RAID0 performance tests, please donât forget to include the Intel RST(e) driver v13.2.8.1002, which is the latest and according to my experience the absolutely best Intel RST(e) driver of the v13 platform, and test it in combination with an Intel RAID BIOS module from the v13.2 branch (EFI RaidDriver v13.2.0.2134 resp. RAID ROM v13.2.2.2224).
I tested the 13.2.0.2134 with the 13.2.8.1002 storage driver and according to my tests it came in last and gave the worst scores! However this is a margin of only a few percent and in my opinion is so close to each other that its possible to get different results depending on the weather! In anycase I am going to stick with the 13.5 branch combination as that did give me the best averaged out scores and each run was pretty much consistent. Though the highest score out of all the 9 runs in total goes to the 13.2 branch it wasnât able to be consistent and so the average high score went to 13.5. Stay tuned will post the screen shots soon.
Note: Noticed Anvil incorrectly reports the installed raid driver on those screen shots as I uninstalled the 14.8.16.1063 driver and installed an older driver but Anvil reports 14.8 as the driver in charge even though I just verified 13.1 installed in device manager!
@davidm71 : Thanks for having done these new benchmark comparison tests and for having published the results, but it seems, that you have done at least 2 big mistakes:
You obviously forgot to uninstall the Intel RST Software v14.8.16.1063, before you tested the different Intel RST(e) drivers. This may have had an impact on all your results. Proof: Anvilâs benchmark tool usually only shows the name of the in-use storage driver. Only if any Intel RST Software has been previously installed (and is still running in the background), it additionally shows the version number of the Software.
The extreme variations of the total benchmark scores (up to >1.000 points difference!) while running 3 times the same Intel RAID driver/EFI module indicate, that the tests have not been done under the same optimized conditions.
If you want to get meaningful comparison test results, you have to make sure, that a) all tests are done under exactly the same optimized conditions and b) no software processes, which are running in the background, interfere your test results. Please have a look into the chapter âPreparations and test conditionâ, which I have layed down within the start post of this thread. Additional tip: The installation of the Intel RST Software is only temporarily necessary and should be done before you are starting the tests to get the ability to enable the WRITE-BACK CACHING option. Once this option has been enbabled, the related Intel RST Software should be uninstalled from within the Control Panel (âAdd/Remove Softwareâ). The WRITE-BACK CACHING setting will stay enabled forever as long as you do not delete or recreate your RAID array.
Your right Fernando I did not uninstall the RST 14.8 console/kit from the system. I just went into device manager and selected the storage manager driver version I want to use and installed it and rebooted. The problem I had was that I could not install an older version of the kit as I got an error claiming my âNET Frameworkâ needed to be updated. I think this error may have to do with Windows Creatorâs Update perhaps. Anyhow I should have known better and uninstalled the console and just manually installed the driver version I wanted. As far as running under ideal optimized conditions the only way to really do that would be a fresh windows install so this was a more real world kind of test I guess. Not only that the raid array I was testing doesnât host the system files and not really used that much by the OS so I thought running the test a few times over and average them be good enough but not all that scientific I suppose. Iâll run it again when I have time for sure but heres what I was able to discern:
1. 14.8 rom branch slower that 13.x branch 2. My MSI bios latest v1.8 20 percent faster than older bios
So Iâll be sticking with the 13.X branch for right now and go with your suggestions until I do a proper better cleaner test.
@ all: The availability of some new Intel RST(e) drivers were the reason, why I have done today completely new benchmark comparison tests with my Z170 system running in AHCI mode.
Update of the Start Post Changelog:
New: Test results with a Z170 system running in AHCI mode Tested AHCI drivers:
Win10 in-box MS AHCI driver STORAHCI dated 03/18/2017 (newest available MS AHCI driver)
Intel RST(e) AHCI driver v13.2.8.1002 WHQL dated 07/09/2015 (latest from the v13 platform)
Intel RST(e) AHCI driver v14.8.16.1063 WHQL dated 04/10/2017 (currently latest from the v14 platform)
Intel RST(e) AHCI driver v15.2.16.1060 WHQL dated 03/30/2017 (currently latest from the v15.2 series)
Intel RST(e) AHCI driver v15.5.2.1054 WHQL dated 04/24/2017 (currently latest from the v15.5 series)
Intel RST(e) AHCI driver v15.7.0.1014 WHQL dated 06/06/2017 (currently latest from the v15 platform)
Thanks for all you guys do here. Great info. I had a question Fernando about the testing you listed here with the drivers. I was wondering there is no mention of which firmware you were using with these drivers you tested. Should we assume that with each branch version driver you used the same branch firmware listed in the "best" thread? Example firmware 12.9.0.2006 with driver 12.4.9.1000.
Also, you mentioned doing clean installs then doing these tests. Have you done this with the Samsung m.2 and if yes then did the unsafe shutdowns revert back to "0" or was the amount of shutdowns the same as before right after you installed the O/S?
If you should mean with âFirmwareâ the related Intel RAID ROM/EFI module of the BIOS, please have a look into the second post of this thread, where I have published the benchmark results I got with an Intel RAID0 system. There you will find all the requested informations Note: As long as your Intel SATA Controller has been set to âAHCIâ within the BIOS, the version of the Intel RAID ROM/EFI BIOS module is absolutely irrelevant, because it will not be loaded while booting.
The topic âunsafe shutdowns of M.2/PCI connected SSDsâ has nothing to do with the topic of this thread, which is only about the relationship between different AHCI and RAID drivers and the performance of the SATA connected SSD.
@ all: After having done recently completely new benchmark comparison tests with my Z170 system running in AHCI mode, I was curious to know the impact of the latest Intel RST(e) drivers on my Z170 RAID0 system. That is why I have yesterday done a lot of new tests with my Z170 system running Win10 x64 v1703 on a RAID0 array. Contrary to the recently published AHCI results the RAID0 ones were again absolutely different - depending on the development branch of the tested driver version. Since I think, that some of you may be interested in the new RAID0 test results, I have published them within the second post of this thread.
Update of the Start Post Changelog:
New: Test results with a Z170 system running Win10 x64 v1703 on an Intel RAID0 array (2x256 GB Samsung 840 Pro) Tested Intel RAID drivers:
Intel RST(e) RAID driver v13.2.8.1002 WHQL dated 07/09/2015 (latest from the v13 platform)
Intel RST(e) RAID driver v14.8.16.1063 WHQL dated 04/10/2017 (currently latest from the v14 platform)
Intel RST(e) RAID driver v15.2.16.1060 WHQL dated 03/30/2017 (currently latest from the v15.2 series)
Intel RST(e) RAID driver v15.5.2.1054 WHQL dated 04/24/2017 (currently latest from the v15.5 series)
Intel RST(e) RAID driver v15.7.0.1014 WHQL dated 06/06/2017 (currently latest from the v15 platform)
Final thoughts: Since such intensive RAID driver comparison tests under optimized and similar conditions are very time consuming, need a full concentration of the person, who is doing the tests, and stresses the used hardware, I am not sure, whether I will repeat such tests in the future. Users, who want to know the performance of the upcoming Intel RAID driver versions and to compare it with driver versions, which have been tested by me, should do the required work themselves. Any publication of such future benchmark comparison tests are welcome!
@ all: The availability of some new Intel RST(e) RAID drivers and the fact, that I own 2x250 GB Samsung 960 EVO SSDs, were the reason, why I have started yesterday completely new benchmark comparison tests with my Z170 system running Win10 x64 on 2 different RAID0 configurations:
ânormalâ Intel SATA RAID0 array consisting of 2x256 GB Samsung 840 Pro SSDs (connected to Intel SATA ports) and
special Intel NVMe RAID0 array consising of 2x250 GB Samsung 960 EVO SSDs, connected to on-bord M.2 ports and running with the NVMe protocol.
The results were very surprising for me and may be interesting for all users with an Intel 100/200-Series Chipset, who have already created an Intel RAID0 array or are going to do it.
Update of the Start Post Changelog:
New: Test results with a Z170 system running Win10 x64 on an Intel SATA and NVMe RAID0 array Tested RAID drivers:
