Issue with Samsung 9100 Pro speed issues on Z890 / Arrow Lake

[Tim21]

I recently installed a Samsung 9100 Pro in an MSI PRO Z890-A Wifi motherboard along with a core ultra 285K processor. I am not getting the advertised sequential read/write speeds of the 9100 Pro.  The SSD is installed in an m.2 slot that is directly connected to the CPU using 4 PCIe 5.0 lanes.

There are several other users that are experiencing the same issue as me. The common denominator is Z890 motherboard (regardless of make or model) and Arrow Lake processor, when the SSD is inserted in a PCIe 5.0 m.2 slot connected directly to the processor.

This thread is one of them: Core Ultra 285K Bandwidth limitation of PCIe 5.0 dedicated SSD port - Intel Community

I have contacted both MSI and Samsung, but neither know the cause and none of their suggestions have worked.

Since the SSD itself can reach advertised speeds except for when in the configuration described above, and since multiple users are experiencing the issue on other Z890 motherboards, I think it is time Intel escalate the issue.

[RobbieR_Intel]

Hello Tim21,

Thank you for your response and for the feedback. I appreciate your efforts in troubleshooting the issue. I understand that the solutions provided have not been effective, and I will need to further investigate this matter internally to quickly find an appropriate resolution/ information. Your patience and understanding are greatly appreciated as we work towards addressing the issue.

Best Regards,

Robbie R.

Intel Customer Support Technician

[Tim21]

This might shed a little light on this issue. Here is a link from thessdreview.com with a review of what they have experienced related to this issue. It appears to be nearly identical to what I am experiencing. The review includes a response from Intel stating that there is an latency issue with the four m.2 PCIe lines on 200s series CPU:

 Intel Core Ultra 200 Series Motherboards Are Not Achieving M.2 Slot Gen 5 14GB/s SSD Performance | The SSD Review

[RobbieR_Intel]

Hello Tim21,

Thank you for patiently waiting and for providing the feedback and we appreciate all the test you have done.

The upstream memory read performance of the ARL S processor SOC tile root ports 13 and 14 (PCIe Lanes 1 to 16), configured as 1px16 or 2px8, may be reduced due to increased roundtrip latencies during random upstream memory read operations. These latency variations depend on the specific workload and the capabilities of the PCIe endpoint device.

Your results could be caused by the system configuration and/or SoC architecture.

Please continue contact us if you have any other questions.

Best Regards,

Robbie R.

Intel Customer Support Technician

[RobbieR_Intel]

Hello @Enid73703,

Thank you for posting in our communities, however for us to further assist you, I highly suggest for you to post a a new thread as your system configuration is different from Tim21. Once you have posted, we will assist you accordingly.

Best Regards,

Robbie R.

Intel Customer Support Technician

[RobertEnger]

Dear Intel:

I too purchased a new (very expensive) Z890 mobo and Ultra 9 285k CPU.  And I too see Gen5 M.2 SSD performance limited to 12GB and change.  Please add me to your list of disgruntled customers.

WILL THERE BE A REISSUE of the cpu in a form compatible with the z890 and 1851 socket  that will give us full throughput?

(If so, will we have to buy it, or will it be supplied as compensation for the bottleneck in the existing 285k?)

OR, will we have to wait for the next generation of Intel CPUs (using the new socket) and trash our motherboard too?  

(and if we have to wait for the next generation, will there be any attempt by Intel to release something SOONER rather than later?)

Bob Enger

[Enid73703]

My speeds are bouncing between 2,400 and 14,000. Thought it might be temp. But have same temp on Samsung 990 Evo 4tb.

Intel has responded that this is a Samsung issue and Samsung is reporting that it is an intel issue.

[Enid73703]

Thank you for patiently waiting and for providing the feedback and we appreciate all the test you have done.



The upstream memory read performance of the ARL S processor SOC tile root ports 13 and 14 (PCIe Lanes 1 to 16), configured as 1px16 or 2px8, may be reduced due to increased roundtrip latencies during random upstream memory read operations. These latency variations depend on the specific workload and the capabilities of the PCIe endpoint device.



Your results could be caused by the system configuration and/or SoC architecture.



Please continue contact us if you have any other questions.



Best Regards,



Robbie R.

Intel Customer Support Technician

[PatrickC9]

I also have issues with the combination of a Samsung 9100 pro 4TB SSD, an ASUS ProArt Z890 creator wifi motherboard and an Intel Core Ultra 9 285K CPU. The SSD is connected to the first M2 slot, which supports PCIE Gen 5 x4 lanes and is connected straight to the CPU. The Samsung magician app correctly identifies the M2 slot as being gen 5 x4 capable, yet the SSD is instead running at gen 1 speed.

My motherboard is updated to the latest BIOS and I tried reseating the SSD. I also tried forcing the PCIE speed for the M2.1 slot to gen 5 in the BIOS settings, but still it was running at gen 1 speed. Read/write speeds are abysmal (<450MB/s), while the WD SN850X 8TB drives in slots M2.2 and M2.3 are running at PCIE gen 4 x4 speed and achieving the expected performance.

[pressed_for_time]

This looks more like a motherboard fault with M2_1 that anything else. It might be worth running it past ASUS support first to see if there is any solution they can offer. If not a replacement motherboard is probably required.

[PatrickC9]

Hi pressed_for_time, thanks for thinking along!

Note that although I can't capture it in the screenshot because it only appears when hovering, the Max slot speed for M2.1 is reported correctly as gen 5 x4 by Samsung Magician.

In the meantime I tried some more things:

• I noticed that in the device manager there was an error indication for the Intel rapid storage device. So I downloaded the latest driver for it from the ASUS website, installed it and then that error was gone. Unfortunately, it didn't make a difference for the SSD speed issue. (BTW: weird that neither ASUS drive hub nor Windows update detected that this driver was missing)
• I removed the 2 other SSDs (WD SN850X gen 4 x4) to make sure that the issue is not caused by compatibility issues between the Samsung and the WD SSDs. This didn't make a difference.
• I then moved the Samsung SSD to slot M2.2, in which the WD SN850X worked at its max speed of gen 4 x4. In this slot the Samsung also ran at only gen 1 x4 "speed".

So it doesn't appear to be a defective motherboard or M2.1 slot, but a defective Samsung SSD (although I can't test it on anything else then this ASUS motherboard). Or a very serious compatibility issue between the chipset (driver) and this SSD, which is unlikely given that others are successfully running it at gen 5 speed on this chipset.

[Tim21]

I agree with pressed_for_time, it is probably a motherboard issue, or a motherboard setting. Mine says that it is using 5x4 in Magician.

[PatrickC9]

Hi Tim21, thanks for your response!

Given that in your case it's reported correctly as gen 5 x4 I apparently have a different problem. Given my findings described above in my reply to pressed_for_time, it appears that I got a bad sample of the 9100 pro.

[RobertEnger]

Presumably you can test the Sammy 9100 on one of your Gen4 slots, and verify that the 9100 does or does not connect at Gen4x4.

(That might help to decide if the 9100 or the Gen5-socket on the Z890-mobo are at fault?)

I suppose you can try one of your Gen4 SSDs in the Gen5 slot and see if it connects at Gen4x4?

(I have not tried this.)

I DO have three Crucial T705  gen5  SSDs  installed in the Gen4 slots on my proart-z890.

They work just fine (but only yield gen4 speeds, of course).

The Crucial 705 SSD adapts down to the bus it is plugged into.  I assume the Sammy 9100 will do that as well?

[I am using the T705 because they have higher bandwidth into the bulk NAND, after the high-speed SLC buffer is full.  The 9100, and all the Sammy stuff, drops down to about 1.6GBytes/sec after the internal SLC buffer gets full.  For many uses that is JUST FINE.  I move a lot of large video files, so writing continuously for a long time is important for me.  The Sabrent and Crucial seem to have better back-end NAND bandwidth than the Sammy designs.  Supposedly, 16-channel NAND SSD controllers will be used in consumer gear soon.  That should speed things up even more.  AND, really put a fire under Intel to fix the Gen5-SSD bottleneck on on the Arrow Lake chips.

If you are interested, you can see the Sustained Sequential Write test results for the 9100 at Tom's Hardware

https://www.tomshardware.com/pc-components/ssds/samsung-9100-pro-ssd-review/2 

(click to make the picture full sized and then click-to-scroll-right to the last picture showing the 900 seconds graph.  If you look closely, there is a partially hidden button to the lower-right that allows you to view the original.  That picture is somewhat easier to make out, especially the colors in the legend...) ]

[PatrickC9]

Hi Robert,

I already tried testing the Samsung 9100 pro in the M2.2 slot instead, which is gen 4 x4 capable and the WD SN850X actually runs at that speed in this slot. However, the Samsung was still running only at gen 1 x4 speed (<450MB/s), so it appears to be an issue with the SSD. Any SSD and M2 slot should be forwards and backwards compatible, and the SSD should run at the lowest speed supported by both the SSD and the slot. So in M2.1 the gen 5 x4 capable 9100 pro should run at gen 5 x4 speed, and in M2.2 it should run at gen 4 x4 speed. In both cases it's only running at gen 1 x4 speed.