WxChaser
[H]ard|Gawd
- Joined
- Dec 23, 2000
- Messages
- 1,328
Thanks Paul - just awaiting the arrival of the Samsung 950 Pro M.2 512 MB cards, they are delayed shipping to me till next Tuesday November 2nd (the final parts I need to complete the build).
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Bootable PCIe NVMe RAID 0 on ASUS Z170-A
- Thread starter amp88
- Start date
Well ASUS could chime in but apparently their support staff is under educated in the areas of UEFI and NVMe RAID0 on their own boards. I exhausted ASUS support before figuring this out on my own. Thankfully I am a certified computer technician and a network administrator with over 20 years experience and I was able to work through this myself.
ASUS should be reading this thread and documenting the instructions for their own support staff to learn from rather than incorrectly informing customers that this configuration is not supported and giving up.
~Paul
.
WxChaser
[H]ard|Gawd
- Joined
- Dec 23, 2000
- Messages
- 1,328
I never was able to see the 2 Samsung 950 Pro M.2's to create a Raid array. I tried the 1101 and 1203 Asus Bios versions and that made no difference. It would let me create a Raid array between one Samsung 950 Pro and my Intel 750 - but I wanted a bootable Raid array with the M.2's.
I think the support for the NVME technology is still a bit premature especially for Raid array setups.
Has anyone else had luck in creating a Raid 0 with the Z170 Deluxe? I tried multiple times in many different ways.
I'm up and running with the 2 Samsung 950 Pro's, an Intel 750 SSD and an older Intel 730 SSD (SATA) drive.,
D
Deleted member 82943
Guest
Op post benches?
WxChaser
[H]ard|Gawd
- Joined
- Dec 23, 2000
- Messages
- 1,328
Paul is a frikkin genius! I finally got to see both Samsung 950 Pro's and created RAID 0 array in BIOS -when prompted to load drivers had the USB drive with X64A drivers on them - once they were loaded proceeding to install Windows 10 Pro on the RAID! I had an Intel 750 SSD drive in the lowest PCIe slot and moved it up on Paul's suggestion. Put the PCIe adapter card with the other 950 Pro in it in that slot and boom - finally! Will update this post and follow up after Window's installation is completed.
So happy it worked!
So happy it worked!
WxChaser
[H]ard|Gawd
- Joined
- Dec 23, 2000
- Messages
- 1,328
The only question I have now that the RAID 0 is set up, is whether or not I should install the Samsung NVMExpress version 1.0 drivers that are out.
Here are some info snapshots and brief benchmarks:
Here are some info snapshots and brief benchmarks:
D
Deleted member 82943
Guest
Wow
WxChaser
[H]ard|Gawd
- Joined
- Dec 23, 2000
- Messages
- 1,328
Quick AS SSD screen shot
WxChaser
[H]ard|Gawd
- Joined
- Dec 23, 2000
- Messages
- 1,328
Yes it is screamingly fast and the way I envisioned setting it up before I built the system.
I can imagine boards like the Asrock with 3 or 4 NVMe slots!
Thanks.
WxChaser
[H]ard|Gawd
- Joined
- Dec 23, 2000
- Messages
- 1,328
This is what it looks like now after a little reconfiguration - still have some tidying up to do inside the case.
WxChaser
[H]ard|Gawd
- Joined
- Dec 23, 2000
- Messages
- 1,328
One more lunch time benchmark: 
D
Deleted member 82943
Guest
Do your nvme drives in raid 0 have trim enabled? First gen ssd raid didn't pass trim commands.
SomeGuy133
2[H]4U
- Joined
- Apr 12, 2015
- Messages
- 3,447
hey chaser. I was messing aroudn with my RAMdisk and I noticed that crystaldiskmark is single thread limited at a certain point and your RAID 0 is getting close to that single thread limit. I just wanted you too know that when you bench that check task manager and keep an eye on single thread to make sure your not single thread limited in the benchmark. If you got low results compared to what your expecting that could be a cause.
http://hardforum.com/showthread.php?t=1886840
Crystal may need an update or be useless by time next gen NVMe drives come out lawl.
http://hardforum.com/showthread.php?t=1886840
Crystal may need an update or be useless by time next gen NVMe drives come out lawl.
SomeGuy133
2[H]4U
- Joined
- Apr 12, 2015
- Messages
- 3,447
you obviosuly didn't look at the limits of Crystal. Some of the speeds of an NVMe drive are near the limits of what crystal can report. Crystal requires CPU to run benchmark and RAID0 of the 950 PRO is near that limit...wow way to post without actually reading turd.
Also His 4K Q32 seems low. That should scale with RAID right?
EDIT: actually all his tests for being RAID 0 are extremely low. Sequential should scale much better than it is. Either the RAID 0 is not working or he is already capping Crystaldiskmarks upper benchmarking limit.
Yea what CPU do you have? What usage are you getting because those benchmarks are way way too low unless something else is bottlenecking your results.
RAID 0 should give more then a mere 26% boost in Seq
Last edited:
My experience (as well as others I've used/seen in person) are with Intel Xeon's in a 2P setup in the 12-16 Core range (EACH CPU) with NVME Arrays of 8-12 drives.
The point I'm trying to make is that a RAM drive barely utilizes the CPU to get those #s and when you're using NVME to get up there it eats away at the cores to get there.
If you want IOPs you need cores.
2x Intel 750 400GB SSD are hardly high performance relative to 4, 8, 12 P3700s.
If what you said is true than what I've experienced and seen on 2.x Ghz Xeons would not be possible, yet it is... thus it's not single core limited the NVME requires much more CPU to get out the IOPs, and you're going to run out of CPU doing so long before you hit "your limit" that you are seeing with a ram disk.
What i'm trying to get across to you is to stop making these blanket statements when you yourself have 0 experience, you're theorizing and guessing... so you should really stop going around to everyone's posts and saying it like it's a fact. CDM can be changed Q and Thread to use more cores....
The point I'm trying to make is that a RAM drive barely utilizes the CPU to get those #s and when you're using NVME to get up there it eats away at the cores to get there.
If you want IOPs you need cores.
2x Intel 750 400GB SSD are hardly high performance relative to 4, 8, 12 P3700s.
If what you said is true than what I've experienced and seen on 2.x Ghz Xeons would not be possible, yet it is... thus it's not single core limited the NVME requires much more CPU to get out the IOPs, and you're going to run out of CPU doing so long before you hit "your limit" that you are seeing with a ram disk.
What i'm trying to get across to you is to stop making these blanket statements when you yourself have 0 experience, you're theorizing and guessing... so you should really stop going around to everyone's posts and saying it like it's a fact. CDM can be changed Q and Thread to use more cores....
Last edited:
SomeGuy133
2[H]4U
- Joined
- Apr 12, 2015
- Messages
- 3,447
I am talking about crystal benchmark capping due to how the program is written with only using a single core. As I showed I saw 100% scaling in the benchmark as I increased CPU clock speed. It was not an issue with CPU being maxed as a whole but Crystal maxing out single thread. Read my fucking post before posting ignorant responses. I showed crystal to be single thread limited.
EDIT:
4K particularly I noticed in taskmanager that single thread was maxed. 8% constant for that test was the reading for crystaldiskmark
8% is single thread max for 12 logical cores.
4KQ32 2044/1626=25.7%
CPU GHz 4.5/3.6=25%
EDIT: so either he is maxing CDM in single thread or his total CPU is capping the speeds or something else is wrong like his PCIe links.
Either way that RAID 0 speed is terrible considering the base speed of 1 drive. It is not scaling right according to that benchmark (if the benchmark itself isn't reporting wrong).
Last edited:
I'm not saying CDM isn't limited at some point based on how it's coded. I'm saying you will use up your CPU/cores with NVME drives before CDM does when you use NVME drives and NOT a ramdisk.
If that was NOT the case then the 2.xGhz Xeons we use in 2P setups would never come close to his score with 4Ghz or yours, yet I've seen it there
at the MUCH lower frequency.He's also benching his "C" drive...
And yes, we see 2x his 4KQ32 with LOWER FREQUENCY CPus.
I would urge an IOMETER test and not 1GB only, record every second, import the data and graph it against the CPU usage.
SomeGuy133
2[H]4U
- Joined
- Apr 12, 2015
- Messages
- 3,447
I'm not saying CDM isn't limited at some point based on how it's coded. I'm saying you will use up your CPU/cores with NVME drives before CDM does when you use NVME drives and NOT a ramdisk.
If that was NOT the case then the 2.xGhz Xeons we use in 2P setups would never come close to his score with 4Ghz or yours, yet I've seen it there
again let me see that. My point is his RAID 0 speeds are terrible and something is bottlenecking if not the benchmark program itself. His numbers are straight up bad for some reason.
Also a 2.6GHz HW from what I found should not report higher then 1180 MBps for 4KQ32 in crystaldiskmark
Testing....nm I have to reboot to reset base clocks on my 1650v3 to test 2.6GHz. I'll try that later so see if it really does scale 100% like I am saying (I expect it will since it clearly is getting single thread limited in my test)
Please then show me those results you claim to show greater then 2GBps (or even 1.6GBps) for 4KQ32 (in crystaldiskmark) because if you can do that on a 2.6GHz HW then how the hell is the program single thread limited on my server?
Last edited:
Did you miss where I pointed out he's doing hte tests on his C: drive?
AKA his boot drive?
It's not going to show performance of an unused array at all.
[image below this -- if not seen maybe hotlink not working PM me]
(Not my system but someone local I've seen in person, and attempted to >25GB/s with)
(Those are Intel 750s (1.2TB) & 2.7GHZ)
CDM is not what i'd suggest though as I said I would use IOMETER and graph it.
My point was to say you're wrong
and point out why... your 1650 V3 will choke on IOPs with more than a few NVME drives, so it's really no comparison to a 24+ Core system.SomeGuy133
2[H]4U
- Joined
- Apr 12, 2015
- Messages
- 3,447
Did you miss where I pointed out he's doing hte tests on his C: drive?
AKA his boot drive?
It's not going to show performance of an unused array at all.
CDM is not what i'd suggest though as I said I would use IOMETER and graph it.
image shows nothing and i dont care about IOMETER. my point is that NVMe drives are near the limit of what crystaldiskmark can show becuase of the single thread limit. Programs like AIDA64 are clearly threaded and are not single thread bottlenecked.
So your saying you wont show me these Crystaldiskmark results that you claim to have showing a 2GHz HW getting 1.6GBps 4KQ32?
You are wrong.
I've seen near 10GB/s in CDM and I PMed you the image url.
You're not understanding what i"m saying, NVME isn't maxing CDM, the CPU is maxed by NVME and CDM in general is not accurate, but it's not maxing 1 core and causing it to flop.
Read what I say.
The image clearly shows >2GB/s 4KQ32 at 2.6ghz.
SomeGuy133
2[H]4U
- Joined
- Apr 12, 2015
- Messages
- 3,447
I saw the picture then explain to me how my CPU not even loaded and shows single thread scaling?
RAM DISK != NVME
I already told you that.
And CDM = bad and inconsistent in GENERAL. (At this level)
If you want to accurately see IOPs then use IOMETER as I suggested, record @ 1S interval, and then graph it against the CPU usage of all cores. This will show you what's going on much better.
(I'm not sure how IOM does with a RAMDISK -- suggestion was for NVME, but if you try it with ram disk please share)
Last edited:
SomeGuy133
2[H]4U
- Joined
- Apr 12, 2015
- Messages
- 3,447
As I said I am not wrong. Look at the video. My point still stands. That CDM has a single thread limit and has issues. I honestly don't have a clue how that person has higher. If you don't believe me watch the video. CDM is single thread limited. I am done wasting my time because I already proved it. If CDM is not single thread limited than explain the video showing my CPU is not even remotely close to being maxed and CDM clearly acting as if its single thread limited. Explain the video....
And claiming NVME is not RAM actually supports my point that CDM is badly limited by single thread. If anything RAM drive will show the programs limits before NVMe drive so your further proving my point that the program should not be used for NVMe or any fast drives.
So my point still stands that:
1. His RAID 0 has crap scaling and there is a bottleneck
2. CDM has single thread limits and NVMe and certain RAIDs are near it or surpass it.
It has near 100% scaling in results so it clearly is single thread limited.
https://www.amazon.com/clouddrive/s...ZknrQ9bL2RQ?v=grid&ref_=cd_ph_share_link_copy
EDIT:IOM and AIDA do not see logical disks so I can't test :/ I really wanted to test AIDA yesterday
Last edited:
As an Amazon Associate, HardForum may earn from qualifying purchases.
Your QD1/T1 4K = VERY HIGH = HIGH CPU USAGE on CORE 1
NVME drives don't do insanely high IOPs on 1Q/1T like your RAM DRIVE
Like I told you, your RAM DRIVE != NVME
CDM NOT good tool for testing a RAM drive, inconsistent with NVME, but it's not limited by 1 CORE on NVME like you keep saying, so you need to STOP SAYING THAT. 1 CORE is NOT LIMITING NVME how you think in CDM.
I don't know how to explain to you anymore.
The other guy has poor RAID0 perf. in NVME because he's bench marking his OS drive.
I'm not going to hold your hand and keep explaining this.
If CDM only ran on 1 core then you couldn't change the threads
and have it USE ALL YOUR CORES....BUT....CDM does let you do this!!!I don't think you're understanding this is the actual problem:
4K is always 1 Core
4KQ32T# in CDM = higher # = more cores used.
That's how it's SUPPOSED to work, but what I said at the start NVME != RAMDISK still holds true
Last edited:
As an Amazon Associate, HardForum may earn from qualifying purchases.
SomeGuy133
2[H]4U
- Joined
- Apr 12, 2015
- Messages
- 3,447
Your QD1/T1 4K = VERY HIGH = HIGH CPU USAGE on CORE 1
NVME drives don't do insanely high IOPs on 1Q/1T like your RAM DRIVE
Like I told you, your RAM DRIVE != NVME
CDM NOT good tool for testing a RAM drive, inconsistent with NVME, but it's not limited by 1 CORE on NVME like you keep saying, so you need to STOP SAYING THAT. 1 CORE is NOT LIMITING NVME how you think in CDM.
I don't know how to explain to you anymore.
The other guy has poor RAID0 perf. in NVME because he's bench marking his OS drive.
I'm not going to hold your hand and keep explaining this.
If CDM only ran on 1 core then you couldn't change the threads
I don't think you're understanding this is the actual problem:
4K is always 1 Core
4KQ32T# in CDM = higher # = more cores used.
That's how it's SUPPOSED to work, but what I said at the start NVME != RAMDISK still holds true
lawl. I just showed you the program is single thread while testing Q32 rofl. If it can use all cores...why doesn't it have that enabled by default? How do you make it use all cores?
EDIT: Ah I found it but it doesn't exactly work. It is at best 2 or 4 threads.
I looked around the settings and I even tried 4KQ64T64, 4KQ32T(1,2,4,6,12,32,and MAX), and the max Q and T it had and the program still runs are 8% for nearly all of them except for T2-12...the program is basically single thread except with a few select settings. Is that really hard to understand? I still can't get it to use more than 50% of my CPU no matter what setting I use so my points still stand.
His RAID 0 is scaling like crap and CDM is a terrible program to use because it is coded like shit.
I can provide you with the other video if you like showing its single thread in almost all the settings and at best 2-4 threads. I get the same results for T2-12 and anything above T12 defaults to T1 or worse. (Actually T6-12 is actually worse than T1 in Seq)
Again keep saying NVMe is not a RAM drive...no shit and it further proves my point that the program is shit and can't keep up in certain cases.
Once Optane is released the program will be utterly worthless and thats the point....It really shouldn't be used because it can't gauge fast drives.
Last edited:
Look, you've been proven wrong and now you're arguing the program is shit. I already told you it was shit and not the best option. I already told you you were wrong.
I already explained to you that a a RAM DRIVE != NVME because a NVME is not going to max 1 Core @ 1Q1T that's why NVME doesn't matter for 4K 1T1Q.
I already told you the other person was shit due to his OS drive being on the same drive he's bench-marking.
What else do you need to understand?
I told you to use IOMETER for real #s for NVME IOPs is what to use.
I told you that you need cores for IOPs...
You obviously still don't understand that 4K = 1 THREAD 1 QUE it's MADE FOR THAT.
If you want to use more CORES THEN INCREASE THE THREADS.
What else do I need to tell you so that you understand all this?
I could care less what your tests show, and that other guy show because I use IOMETER for real data, and you have no clue what you're doing so anything you do is invalid in my opinion.
HOWEVER ---- CDM is not single core limited only --- LIKE YOU WENT AROUND SAYING IT WAS!!
You've been proven wrong that CDM is not SINGLE CORE limited.
If you still fail to grasp how CDM works, and what each row represents with Q/T in them then you are a lost cause, but please stop trying to educate me when you have 0 NVME drives to even use... LOL!!!!
CDM is Single Core in numerous tests that's the damn point when are you going to grasp how it works? I'm thinking never, you keep ignorantly arguing this point making yourself look like a fool. However, CDM is NOT single core limited, you can change this, and the other rows are what this is for. This is to COMPARE SINGLE THREAD vs >1... Do you understand this now? I spelled it out and held your hand.
I wasn't aware when I started replying to you that you actually had 0 clue how to use CDM and what each row represents but now that you've replied numerous times it's very clear you had no idea how to use the program, why it was using 1 core, how to add more threads, etc...
How do you know when Optane comes out it will be useless? They'll likely upgrade it, but neither of us know if it will be useless or work just fine...
Stop making ignorant claims about the future that you once AGAIN know NOTHING ABOUT.
Now that I've educated you have a good night, and don't go around spreading bull shit.
Oh, and PS: The reason why they have 4K @ 1CPU is because that's how most desktop systems utilize the file system... you probably don't even understand that though and the implications of it, and why the 400,000 IOPS rating of 1 NVME drive in a usual desktop environment will NEVER get utilized. (unless you're running VMs, or something made to take advantage)
I already explained to you that a a RAM DRIVE != NVME because a NVME is not going to max 1 Core @ 1Q1T that's why NVME doesn't matter for 4K 1T1Q.
I already told you the other person was shit due to his OS drive being on the same drive he's bench-marking.
What else do you need to understand?
I told you to use IOMETER for real #s for NVME IOPs is what to use.
I told you that you need cores for IOPs...
You obviously still don't understand that 4K = 1 THREAD 1 QUE it's MADE FOR THAT.
If you want to use more CORES THEN INCREASE THE THREADS.
What else do I need to tell you so that you understand all this?
I could care less what your tests show, and that other guy show because I use IOMETER for real data, and you have no clue what you're doing so anything you do is invalid in my opinion.
HOWEVER ---- CDM is not single core limited only --- LIKE YOU WENT AROUND SAYING IT WAS!!
You've been proven wrong that CDM is not SINGLE CORE limited.
If you still fail to grasp how CDM works, and what each row represents with Q/T in them then you are a lost cause, but please stop trying to educate me when you have 0 NVME drives to even use... LOL!!!!
CDM is Single Core in numerous tests that's the damn point when are you going to grasp how it works? I'm thinking never, you keep ignorantly arguing this point making yourself look like a fool. However, CDM is NOT single core limited, you can change this, and the other rows are what this is for. This is to COMPARE SINGLE THREAD vs >1... Do you understand this now? I spelled it out and held your hand.
I wasn't aware when I started replying to you that you actually had 0 clue how to use CDM and what each row represents but now that you've replied numerous times it's very clear you had no idea how to use the program, why it was using 1 core, how to add more threads, etc...
How do you know when Optane comes out it will be useless? They'll likely upgrade it, but neither of us know if it will be useless or work just fine...
Stop making ignorant claims about the future that you once AGAIN know NOTHING ABOUT.
Now that I've educated you have a good night, and don't go around spreading bull shit.
Oh, and PS: The reason why they have 4K @ 1CPU is because that's how most desktop systems utilize the file system... you probably don't even understand that though and the implications of it, and why the 400,000 IOPS rating of 1 NVME drive in a usual desktop environment will NEVER get utilized. (unless you're running VMs, or something made to take advantage)
Last edited:
SomeGuy133
2[H]4U
- Joined
- Apr 12, 2015
- Messages
- 3,447
lawl. It has nothing to do with CPU cores or my CPU. I have told you that and PROVED you WRONG again and again. here is another video showing you my CPU is not even remotely maxed at various Q depths and threads so my point still stands. The program is terrible for fast drives because there is a very easy to reach limit. -_-
yes please keep educating me with ignorance man. I provided you with the fucking objective data proving you wrong. So fucking ignore facts. Jesus.
https://www.amazon.com/clouddrive/s...z0ZGbstavZ0?v=grid&ref_=cd_ph_share_link_copy
Seriosuly keep rambling and looking like a stoog. I gave you actual examples of the program clearly have upper limits that NVMe drives can even reach. But keep thinking your right and ignore the actual proof I have posted of running the test and showing its limits in real time what the fuck ever. Ignorance is bliss.
Cores man and the threads!!! It dont fucking matter watch the video!!! I can set 512 threads or 2 threads and I still don't max the CPU!!!! Oh and 512 threads actually defaults to 1 thread for whatever reason!!! and 6 threads is equal to 2 threads...jesus christ your ignorant.
yes please keep educating me with ignorance man. I provided you with the fucking objective data proving you wrong. So fucking ignore facts. Jesus.
https://www.amazon.com/clouddrive/s...z0ZGbstavZ0?v=grid&ref_=cd_ph_share_link_copy
Seriosuly keep rambling and looking like a stoog. I gave you actual examples of the program clearly have upper limits that NVMe drives can even reach. But keep thinking your right and ignore the actual proof I have posted of running the test and showing its limits in real time what the fuck ever. Ignorance is bliss.
Cores man and the threads!!! It dont fucking matter watch the video!!! I can set 512 threads or 2 threads and I still don't max the CPU!!!! Oh and 512 threads actually defaults to 1 thread for whatever reason!!! and 6 threads is equal to 2 threads...jesus christ your ignorant.
Last edited:
As an Amazon Associate, HardForum may earn from qualifying purchases.
What the hell are you talking about?
You're the one who was arguing it was limited to 1 CPU.
I told you posts ago CDM wasn't reliable.
I also told you it wasn't limited to 1 CPU like you kept saying.
You're just going in circles now.
I also told you post #1 NVME != RAM.
As an Amazon Associate, HardForum may earn from qualifying purchases.
LOL !!!
Says the guy with 0 NVME drives testing a RAM disk thinking it's going to tax all the cores like NVME which don't connect via RAM.
Your videos of a ram disk don't = NVME so you need to stop ASSuming they do.
I flat out told you it was not reliable, stop trying to argue CDM is bad I TOLD YOU THAT.
You told me CDM had 1 core limit I told you it didn't, and I proved to you it didn't.
You're forgetting where you started arguing, ha ha. and now you're arguing my point of CDM not being reliable...
I suggest you get NVME if you want to argue what they do to a CPU when you run high Q and high T.
SomeGuy133
2[H]4U
- Joined
- Apr 12, 2015
- Messages
- 3,447
i have a 950 PRO derp derp genius. Using the RAM drive to see what the upper limits of the program is derp derp or is that too hard to comprehend?
Your right RAMdisk is faster and proves there is an upper limit so now we know what the upper limit of the program is no matter what.
I showed you a video of it running at every possible thread and it doesn't max the CPU so there is clearly a limit to what the program can test...wow complicated aye?
And that upper limit is in the range of current NVMe RAIDs and next gen NVMe drives. So yea...how the fuck dont you understand that or do you enjoy trolling hard?
I'm going to break this down for you since you're not comprehending what you yourself have said through this entire thread.
YOU ARE FLAT WRONG.
You now know it's not 1 core limited.
Did you read what I said here in my previous post? I'm not arguing it's not limited at some point, I also said from the start that NVME in CDM != RAMDISK. Your point of 1 core or 20 threads and a RAMDISK don't hold true when you use a NVME ARRAY or SATA array, you're flat wrong in regards to taxing the CPU cores. RAM doesn't or might blip but NVME and SATA arrays do. I already said this NVME!=RAMDISK you're too dense to comprehend this though.
No, you're using a ramdisk not NVME or SATA array so you've proven nothing with your "upper limit" you've simply shown it doesn't work for RAM, and I never said it did.. in fact I said it wasn't good for ram, and isn't reliable in general, in fact I said use IOMETER but you kept arguing it was 1 core limited until you finally learned how to use CDM. You're also only testing 16GB in a system (RAM) that can do way more than that in 1 second, and you wonder why it doesn't tax more of the CPU?? like I said you don't understand what you're testing and how it all works electronically. There's a reason SANs have high cores to get high IOPs, they actually use the CPU unlike your ignorant belief it's not needed because your ridiculous ramdisk test that you believe mimics NVME.
All your tests in the world on RAMDISKS mean nothing when it comes to NVME performance because NVME affects the CPU completely different than ram, you fail to realize this.
YOU ARE FLAT WRONG.
You now know it's not 1 core limited.
Did you read what I said here in my previous post? I'm not arguing it's not limited at some point, I also said from the start that NVME in CDM != RAMDISK. Your point of 1 core or 20 threads and a RAMDISK don't hold true when you use a NVME ARRAY or SATA array, you're flat wrong in regards to taxing the CPU cores. RAM doesn't or might blip but NVME and SATA arrays do. I already said this NVME!=RAMDISK you're too dense to comprehend this though.
No, you're using a ramdisk not NVME or SATA array so you've proven nothing with your "upper limit" you've simply shown it doesn't work for RAM, and I never said it did.. in fact I said it wasn't good for ram, and isn't reliable in general, in fact I said use IOMETER but you kept arguing it was 1 core limited until you finally learned how to use CDM. You're also only testing 16GB in a system (RAM) that can do way more than that in 1 second, and you wonder why it doesn't tax more of the CPU?? like I said you don't understand what you're testing and how it all works electronically. There's a reason SANs have high cores to get high IOPs, they actually use the CPU unlike your ignorant belief it's not needed because your ridiculous ramdisk test that you believe mimics NVME.
All your tests in the world on RAMDISKS mean nothing when it comes to NVME performance because NVME affects the CPU completely different than ram, you fail to realize this.
Last edited:
Ugh. I'm trying to get this working on my Z170-WS and the settings you posted in picture form aren't accessible anymore. What ere the magical UEFI settings needed for NVME RAID boot?!Ok here are the UEFI settings needed for the ASUS Z170-A and NVMe RAID0 for boot. I am using the latest release from Asus Which is UEFI/BIOS Version 1101.
Once these settings were configured I used the WINDOWS 7 USB/DVD DOWNLOAD TOOL () to create a Windows 10 x64 bootable USB flash drive from my Windows 10 x64 ISO file. (Any UEFI boot disk should work).
At this point I rebooted the machine and entered the UEFI/BIOS. In the boot section I selected the UEFI version of my flash drive as the boot override option.
On reboot Windows begins the installation; however, the newly created RAID0 volume is not detected. Here you must feed the installation (add) the Intel Rapid Storage Drivers (Browse to the inf file in the X64 folder (iaStorAC.inf)). You can download these at the link below.....
After the RST driver is loaded the RAID0 volume was detected and installation proceeded as normal from there.
I hope this is helpful.
~Paul
If I follow ASUS' instructions I can create a RAID volume in the BIOS, but when I try to install windows it doesn't see the drive, even with the RST drivers loaded (it does detect that they're compatible with a device on my system) but the volume just doesn't show up in the Windows installer.
Excuse me,
I would to ask ... My motherboard is ASUS Z10PE-D16 WS
have 6 PCI-E Gen3 x16
1th and 3th is the bandwidth of PCI-e 3.0x8
may i directly on the BIOS,
2 Samsung 950 Pro PCI-e 3.0x4 do RAID0??
Intel (r) C600 Series Chipset does not support RAID ??
But I have found a related driver.. ... Do not know OK??
Intel® Rapid Storage Technology Enterprise NVMe (Intel® RSTe NVMe) RAID Driver
Version: 4.5.0.2125 (Latest) Date: 8/29/2016
I would to ask ... My motherboard is ASUS Z10PE-D16 WS
have 6 PCI-E Gen3 x16
1th and 3th is the bandwidth of PCI-e 3.0x8
may i directly on the BIOS,
2 Samsung 950 Pro PCI-e 3.0x4 do RAID0??
Intel (r) C600 Series Chipset does not support RAID ??
But I have found a related driver.. ... Do not know OK??
Intel® Rapid Storage Technology Enterprise NVMe (Intel® RSTe NVMe) RAID Driver
Version: 4.5.0.2125 (Latest) Date: 8/29/2016
Last edited: