Is anybody getting slow speeds with the Intel 750 SSD after updating to the Intel NVMe Drivers? I'm not positive that the speeds I'm getting are, in fact, slow but I have posted some pics below along with a comparison pic of a https://www.youtube.com/watch?v=JMSq8nmDTO0 video on YouTube by Asus and Intel where their speeds are much greater for some of the tests. I believe I set the test correctly. I set it to 3 (not sure what that is) and 8GB (not really sure what that is either but they say to do that in the video). Just looking for any input or ideas... And maybe I'm just being overly neurotic about the numbers matching exactly but in some cases their numbers are literally double what mine are, and I have the Intel NVMe drivers installed...soooooo....not sure what to do, or if there even is anything I can do at this point...????
CPU: i7-6700k Skylake
MoBo: GA-Z170X Gaming 7 Gigabyte Motherboard (BIOS: F6b)
RAM: 16GB G.Skill DDR4
Primary Drive: Intel 750 PCIe SSD 1.2TB (Intel Driver 126.96.36.1997 (Latest))
Secondary Drive: 8TB HGST UltraStar HDD 7200RPM
This was the test prior to installing Intel NVMe drivers.... (my apologies for the sideways pics....)
This is after installing Intel NVMe drivers......
HOWEVER, if you look at the benchmarks from thehttps://www.youtube.com/watch?v=hxpwmrmw6jI Intel 750 Interview by Intel and Asus on YouTube....at 20:15 their speeds are significantly higher for Seq, 4k, and 4kQ32T1.....
As you can see, their 4K Reads/Writes are double mine. And the "Seq" numbers are significantly greater as well.
I believe I have installed the proper Intel NVMe drivers but I took some pics just in case... from reading other posts here at the Intel Community Forum I realized I was initially looking in the wrong place b/c I was thinking I couldn't get the Intel drivers to install b/c I was looking at the disk drive details..like below...
So I looked at the Storage Controller in Device Manager...
So I guess my question is why am I getting such slow speeds if I've done everything I can to get this working right.....??????????? Any ideas..??
Intel® SSD performance is tested withhttp://www.iometer.org/ IOmeter*, since it allows the configuration of the test to take advantage of the drive technology. The expected performance can be located in page 8 of the http://www.intel.com/content/dam/www/public/us/en/documents/product-specifications/ssd-750-spec.pdf Intel® SSD 750 Series Product Specifications.
You can find information about benchmarking for your drive in the document: http://download.intel.com/support/ssdc/hpssd/ssd-750/sb/intel_ssd_750_series_evaluation_guide_332075... Intel® SSD 750 Series Evaluation Guide.
There are many tools used to measure drive performance and CrystalDiskMark is one of the different tools available. This is not a tool we normally use, so we cannot refer to the methods and results of those tests.
I noticed a difference in the pictures from your test, and the ones from the benchmarking video. Your test says "4K Q32T1" and their test shows "4K Q32T8". Doing a quick check online this indicates your test is made with 1 data thread, and their test uses 8 threads. The 750 benefits from multiple threads in some cases (for example, the advertised performance is obtained with Queue Depth 32 and 4 threads), so you might want to check in the software documentation if it is possible for you to change this.
Intel® SSD performance is tested withhttp://www.iometer.org/ IOmeter*, since it allows the configuration of the test to take advantage of the drive technology.
You can't say that when there is a Newegg Tv youtube (4-15-15) with Intel employees showing off the drive performance using CrystalDiskMark.
I have a 400GB Intel 750 connected to a Asus Maximus VIII Impact via U.2. I'm guessing our results are different from the youtube video's due to our data being routed through the Z170 PCH. Those youtube videos were pre-Z170 release and the Intel 750 was being served by 4-Gen 3.0 CPU lanes. Their test setup probably reduced their graphics card to 8-Gen 3.0 CPU lanes to give 4 to the drive. Our chipsets keep the graphics card at 16-Gen 3 CPU lanes but route the intel 750 over the pch adding latency. My test results were worse than yours but I'm not near my desktop to share right now. This is an interesting topic and some input from intel would be of value.
That's some in depth analysis there brother. Honestly I had to do a little research just to understand what you're talking about lol.. I'm a bit of a noob when it comes to PC Builds and hardware. I'm a web design/shopping cart guy so I focus on software more. But yes I basically get the gist of what you're saying and it sounds like you're on to something there. Lord knows I had a helluva time building this thing b/c there was no documentation for the Z170 chipset. Everything was based off the previous CPU to Skylake and the X99 I believe..?? But that's why I waited for the Skylake to come out b/c I knew that would be the fastest possible combination and best build I could possibly make....meaning that instead of a mere 6 months to obsolescence I now have a year.
Hopefully Intel will do some more research and testing with this configuration now. Hint Hint Nudge Nudge Wink Wink Intel.... I'm sure many of us would like to see what kind of results you guys get with the Intel 750 and the i7-6700k CPU and Z170 chipset.
Thanks for the help Nicholas1,
Nicholas, I've done some research on the PCIe lanes available to us with our current configuration and it seems to me that the Intel 750 SSD should have at least 4 Gen 3.0 PCIe lanes available to it if all we have connected is one graphics card consuming 16 lanes because from what I read the i7-6700k has 20 lanes of gen 3.0 available. So I'm just deducing that if the GPU gets 16 of those and I have the Intel 750 connected to the bottom slot (PCIe x4) then it is using those 4 lanes. I don't have anything in the middle PCIe x8 slot.
Is that not right? Why would it be routed over the PCH?
I should preface that I'm a noob when it comes to hardware and I barely understand PCIe as it is. I mainly got this info from some articles I read...
"Z170 expands support to 20 lanes of PCI Express 3.0 with 8Gbp/s of bandwidth each."
Read more: http://www.digitaltrends.com/processor-reviews/intel-core-i7-6700k-review/# ixzz3rjw70drD http://www.digitaltrends.com/processor-reviews/intel-core-i7-6700k-review/# ixzz3rjw70drD
Is that not right? Why would it be routed over the PCH?
Yes, it's true the processor has 20 lanes available. However, if your discrete GPU is occupying 16 lanes, the remaining 4 are for the DMI connection to the PCH. Prior to Skylake, the DMI was running at 4 lanes Gen 2.0 creating a massive bottleneck (~1.5GB/s max realized xfer speed) so the drive was a worthless purchase if not linked directly to the CPU. With z170, there is an additional 20 Gen 3.0 lanes available on the PCH (that's why the article says with Z170, not with intel 6th gen cpus), however, all data is still routed over the DMI.
I don't know your motherboard..it may not be possible, but you definitely would need to see your GPU reduced to 8x speed or slower in GPUz. At that point 4 of your CPU lanes would be unused. No big deal. I can't do this because I have the 2.5" and a limited ITX board.
This is my system:
Asus Maximus 8 Impact
16GB TridentZ 2800MHz Ram
2.5" Intel 750 ssd 400GB firmware 8EV10171 w/ driver 188.8.131.527
No other storage devices
With your initial post you had decent numbers except for the Q32T1. You just needed to change it to T4 or 8 for that one 4k test. I however have a problem. My numbers are quite worse than yours. See attached. I've tried everything including suggestions from Jonathan. I may need to open a service request with intel on mine.
Yup. You were right. Once I figured out where to go to change it to T8 the numbers are looking much better. I have attached results below. Everything looks basically ok at this point. The only thing that confuses me is the bottom 4k reading. It doesn't seem to change no matter what. But maybe that doesn't matter....?? Are the only ones that are important the top 2 numbers? Meaning the Q32... numbers?
The numbers from the last test are low due to the configuration it uses, it is Random Read test using 4K, no queuing and a single worker (data thread):
Random Read 4KiB (Q= 1,T= 1) : 35.219 MB/s [ 8598.4 IOPS]
As we mentioned before, the Intel® 750 SSD is designed to work better with queuing and multiple threads as it would normally be under normal system operation. For example, the advertised performance is obtained with Queue depth 32 and 4 threads.
Queue Depth allows the SSD to have multiple requests from the host at the same time by enabling the concurrent processing of commands that the host can queue up.
Higher queue depths typically allow the SSD to generate higher IOPS through concurrent processing of commands. However, just keep in mind that if the queue size goes beyond the drive's specifications, latency may be increased.
Right, I understand that I need to use queuing and multiple threading etc.. for the top numbers... BUT, what is puzzling to me is how and why are they getting higher numbers (the bottom numbers that are T=1, Q=1)? Just check their https://www.youtube.com/watch?v=JMSq8nmDTO0 numbers at 25:13..(different video but the same results that I guess Intel was told to show everybody b/c they are the exact same numbers shown on every YouTube video out there). Their bottom 2 numbers are T1 and Q1. So how is it that their Q1 numbers are double mine?
Now, my Q32 T8 numbers are looking pretty good and my system is noticeably faster than my last SSD so I'm willing to just drop all this if you tell me I don't need to be concerned with the bottom numbers. I just want to make sure I'm maximizing the performance of my drive. And when I see someone else with the exact same drive get any numbers that are double mine under the exact same configuration, which in fact, they are, then I start to worry. But if you tell me I don't need to worry about those then I will be good.
The test parameters will have a direct impact in the results.
When the drive is under normal operation the computer will use different data threads and queue depth. A test using a single Worker and no command queuing will always have low results. Those results should not be relevant if you are measuring the actual performance of the SSD.
You guys hit the nail on the head with the Pcie lanes. Thank you.
ASUS Z170-Deluxe Bios 1702
Intel SSD Toolbox 3.3.3
SSD Driver 184.108.40.2067
Pcie x16 slot_1 = GTX980Ti @ x16
Pcie x 16 slot_2 = empty
Pcie x16 Slot_3 = SSD @ ?
I assumed (oh that word) with bios on auto Slot_3 would have 4-lanes available but... this was incorrect as this slot is shared IRQ with amongst other things, the Intel SATA ports. So with Pcie bios setting = Auto the SSD was only getting 2-lanes.
I manually set the pcie slot_3 in the bios to x4 this incidentally has disabled the Intel SATA ports which sucks; however the SSD meets specs when tested with Iometer.
Seq write = 790- close enough. I was using the pc while the test was occurring.
Seq read = 2182
Ran read = 430212
Ran write = 196231 - close enough.
Not sure what the deal is with Crystal:
Now I just need figure out how to get my SATA ports back without some elaborate reassignment of IRQ's. I have 2-SATA ports on the ASMEDIA driver but need one more.
Alternatively I could move SSD to slot_2 and run the GPU @ 8x which is fine (Pcie 3.0) but not what I signed up for (what happens when an enterprise level device meets gaming enthusiast).
Edit: Only SATA 5 and 6 ports are disabled, we are good to go as I have plenty of SATA ports now.
Yeah I think this is probably an issue of me not understanding how to use CrystalDiskMark properly. I'll do some more research and see if I can figure out how to change those settings. I thought, upon watching that YouTube video, that all I had to do was change the second numeric value to 8GB and it would change it to 4KQ32T8, but it's still saying 4KQ32T1....just realized that...sorry. But thank you for the prompt response and assistance mate! It's why I love Intel and have been a Intel SSD fan since the Cherryville 520.
So I finally installed IOMeter and ran the tests (definitely not as easy as CrystalDiskMark btw...) but, ultimately, I am still getting what appears to be rather low benchmarks for the Intel 750, though I just couldn't say why at this point. I feel like I've done everything I can to make it run right: latest BIOS, Intel NVMe driver, Intel SSD Toolbox, it's in the right slot PCIe x4 gen 3.0, etc... is there something else that I'm not thinking of anyone...? I just find it hard to believe that with the newest CPU on the market I'm getting lower numbers than expected.
My results from IOmeter are:
128K sequential WritesMiBps (Binary)1080.877421
(I guess this is close enough to 1200 so not too concerned with this one...)
4K Random WritesWrite IOps120952.5005
(This is less than half what it should be..... 290,000...????)
4K Random ReadsRead IOps359170.6106
128K Seq Reads (I think I messed something up at this point b/c this is all it spit out...)Test TypeTest Description0Version1.1.0Time Stamp2015-11-12 05:03:35:357Access specificationsAccess specification namedefault assignment128k sequential read0size% of size% reads% randomdelayburstalignreply131072100100001131072<td data-sheets-value="[null,3,null,0]" style="padding-right...
When setting the random tests, please make sure to set the different workers (data threads) as mentioned in the http://download.intel.com/support/ssdc/hpssd/ssd-750/sb/intel_ssd_750_series_evaluation_guide_332075... Intel® SSD 750 Series Evaluation Guide. If you use a single worker, the CPU will not be able to keep up with the drive and the test result will be lower. This is similar to the 4K Q32T1 vs. 4K Q32T8 value from CrystalDiskMark in the advertisement video.
- If you would like, we can take a look at the SSD logs, they can be obtained with https://downloadcenter.intel.com/download/18455/Intel-Solid-State-Drive-Toolbox Intel® SSD Toolbox, by selecting the drive and clicking the "Export" button.
- Please confirm that the test is configured as shown in the evaluation guide and upload the complete file from the test results for further analysis.
I'm also in this same problem of slow speeds even slower then a SATA 3 SSD
I'm very interested if someone from Intel can explain3096Current PCIe Link Speed0x33097Negotiated PCIe Link Width0x4
I at first had Negotiated at 0x2 but after I looked at how many lanes my moatherboard can support with the amount of drives I had I had to move some drives from one port to another and now ended with the same as Tyson on this side.
Can anyone comment on possible reasons for the Current PCIe Link Speed only being 3 instead of 4
There was some information from your tests that we were not able to confirm since the file was incomplete, however, I would like to provide you with additional recommendations that have worked for other users before performing any further testing.
- Make backup of any important data you may have (as preparation for further troubleshooting).
- Connect your Intel SSD to the PCI Express 3.0 x16 slot, running at x4 (PCIEX4).
If the issue persists, try the following:
1. Reset the PC BIOS back to factory default values.
2. Contact http://www.gigabyte.com/products/product-page.aspx?pid=5481# dl Gigabyte Support and obtain the most recent BIOS for your Motherboard, also to confirm their recommendations about BIOS configuration for NVME* drives.
3. Configure your Motherboard again for use with your Intel® SSD 750 Series and other PC components.
Try the test again. If the results are not satisfactory, you can attach the files to your reply by using the Advanced Editor, then clicking on the "Attach" function that will be available at the bottom of the reply box.