Tyan S8030GM2NE

jpmomo

Active Member
Aug 12, 2018
349
118
43
The Tyan FRU-CS-1230/Slim SAS 8i to 2* U.2 (NVMe) cable is now available from retail (e.g. Acme Micro), the price is a steep $75. With shipping and taxes, it's an eye watering $95 :eek:

Update: Purchased one and here's the pic. It comes with a Molex connector for power and not the usual SATA power connector (Enterprise NVMe drives will take 3.3v and 12v). Seems high quality and right now I'm looking for some refurb/second hand NVMe drives to test this out with.


View attachment 15318
I am assuming that this cable will support the new pcie 4.0 nvme 2.5" drives like the samsung pm1733 and kioxia cm6? I think the slimsas side is 8 lanes and splits into 2x4 lanes. Do you know if there is any difference with this cable that uses a molex for power vs other cables that use sata power? Have you had any issues with this cable? Which type of drives are you using it with?
thanks for any tips!
 

whbeers

Member
Jul 11, 2020
31
33
18
Tried out the new V2 bios, and it resolved the "BAAD" post code issues described above. Both systems up and running.

Note that it does reset bios settings as part of the update (dangerous because it set fans back to 30% PWM, decreasing the effectiveness of my watercooling setup), and in my case resulted in renumbered NICs such that I needed to reconfigure proxmox interfaces via ikvm before the machines were back up and running.
 

hmw

Active Member
Apr 29, 2019
331
104
43
in my case resulted in renumbered NICs such that I needed to reconfigure proxmox interfaces via ikvm before the machines were back up and running.
Yep - the v3 -> v3.01 BIOS did the same for ESXi and worse. I'm now debating whether to upgrade from v3.01 to v4.00, but I see that there's Fixed SOC P-state and DF C-state options. Might be useful for making the board more energy efficient ...
 

hmw

Active Member
Apr 29, 2019
331
104
43
I am assuming that this cable will support the new pcie 4.0 nvme 2.5" drives like the samsung pm1733 and kioxia cm6? I think the slimsas side is 8 lanes and splits into 2x4 lanes. Do you know if there is any difference with this cable that uses a molex for power vs other cables that use sata power? Have you had any issues with this cable? Which type of drives are you using it with?
thanks for any tips!
I am using the cable on a Samsung PM983 U.2 drive and it works fine. I briefly tested on an Intel NVME drive and it worked fine too. The cable is indeed SlimSAS x8 on one end and splits to 2 x SFF-8639 - I don't have a PCIe 4.0 drive to test.

Isn't the CM6 U.3 though? (although U.3 is backwards compatible with U.2)

Regarding power - I believe the U.2 drives can do without 3.3v - at least I have not noticed any issue with using Molex for power vs SATA power. This cable WON'T supply 3.3v, but haven't had any issue with Intel and Samsung drives yet

@diogin got the cable for cheaper from taobao - perhaps he can share the link? It's still available from Acme Micro for $75. They charge a hefty fee for shipping so worth it to buy two cables. It's funny, they're just down the road from me but due to Covid there's no pickup from any local place :)
 

whbeers

Member
Jul 11, 2020
31
33
18
I also just realized the version numbers are different for equivalent bios releases across the S8030GM2NE vs S8030GM2NE-2T (<- which I'm using). Apologies for any confusion around the version numbers in my posts!
 

diogin

Member
Mar 28, 2018
55
14
8
Beijing, China
I had sold out my S8030GM4NE-2T because it has flaw on the physical design of PCIE 4.0, which always reports PERR on BMC and WHEA error on Windows for PCIE 4.0 SSDs. This was confirmed by Tyan support team. The board with above flaw is revision 1.00, and I don't know whether Tyan has newer revisions for this board which have fixed the PCIE 4.0 flaw. What is your revisions of this board?

Regarding to the SlimSAS to U.2 cable - I had only tested it with a PCIE 3.0 U.2 SSD - Samsung PM963 960GB. It works well. I'm not sure if it work in PCIE 4.0 (because I have not got a PCIE 4.0 U.2 SSD), perhaps not, since PCIE 4.0 has much strict electrical requirement than PCIE 3.0. For PCIE 4.0, the length of the cable should not be too long (while the cable I bought is 65cm in length), and if it is, it should install a PCIE 4.0 retimer to calibrate the completeness of the signal. I had once searched the topic, and IIRC supermicro has released a SlimSAS 8i to 2x U.2 cable which supports PCIE 4.0, but I forgot the product number of the cable... Maybe you can search it in supermicro's website.

The cable I bought is SATA powered, it supports three different voltages - 3.3V, 5V, and 12V. The cable @hmw bought is 12V only. I am not an expert on U.2 SSD so I cannot figure which power is the standard for U.2 SSD, but at least SATA powered cable supports all of the three voltages and should be more universal. If there is a module of "voltage reducing" circuit in your U.2 SSD, then 12V-only cable should be usable.
 

diogin

Member
Mar 28, 2018
55
14
8
Beijing, China
Yep - the v3 -> v3.01 BIOS did the same for ESXi and worse. I'm now debating whether to upgrade from v3.01 to v4.00, but I see that there's Fixed SOC P-state and DF C-state options. Might be useful for making the board more energy efficient ...
These two BIOS options are requested by me :)
I had tested these options, but it didn't work as I expected it to be. Under Supermicro H11SSL/H12SSL, if I set APBDIS=1, Fixed SOC Pstate=P0, DF-Cstates=disabled, then I got a fixed FCLK=1466MHz and MCLK=1466MHz for my Epyc 7502P, which is the coupled state. But under Tyan it isn't, the FCLK and MCLK are both in weird frequency. After that I sold out my S8030GM4NE-2T and didn't know if they had solved this issue. Hope they do!

Further, these two options are for higher performance, not energy efficient. In shorts:

* APBDIS means "disable algorithm based frequency boost and throttling of data fabric". Setting it to 0 uses algorithm to control the DF frequency, and 1 to disable algorithm control and follow a P-state in Fixed SOC Pstate.
* Fixed SOC Pstate: If APBDIS=1, this option is used to control which P-state is used for data fabric. P0 means highest frequency.
* DF Cstates controls if data fabric can go to energy saving mode while the cores in CCD/CCX are in C-state >= C1.

For more details see AMD's tuning guide in: Performance Tuning Guides - AMD
 

hmw

Active Member
Apr 29, 2019
331
104
43
If anyone is interested in the notes for BIOS v4.00

Things to note:

- There's two steppings for Milan (B0 and B1)
- New microcode for Rome
- New AGESA for Rome (RomePI 1.0.0.A)


Code:
V4.00      - Date       : 03/11/2021
           - Filename   : 8030V400.ROM
           - Banner     : TYAN S8030GM2NE BIOS V4.00
           - CheckSum   : 6BA3
           - MD5        : 6f367c0aef6d360a11415ce933401376
          
           - Recovery   : 8030M001.CAP (Milan)
                          8030_001.CAP (Rome)
           - AGESA Ver. : MilanPI-SP3 1.0.0.1
                          RomePI-SP3 1.0.0.A
           - Microcode  : B1 0x0A001119 (Milan)
                          B0 0x0A00103F (Milan)
                          B0 0x0830104D (Rome)
           - VGA OpROM  : V1.10
           - LAN OpROM  : I210: PXE: V1.5.85 EFI: V8.3.04 ISCSI: V3.1.30
           - Description: Based on V3.01
                          1. Support Milan CPU Family 19h.
                          2. BIOS setup changed.
                              - Modify “APBDIS” item default to [1].
                              - Added “Fixed SOC PState” item and default is [P0].
                              - Added “DF CStates” item and default is [AUTO].
                              - Change HTTP string to HTTPS.
                          3. Enable CPU Group D platform feature.
                          4. SMBIOS Type1,2,3 retrieve from BMC FRU. (Modify SMBIOS directly is not allowed.)
 

hmw

Active Member
Apr 29, 2019
331
104
43
Updated my motherboard to V4.00, uneventful and no changes in any LAN assignment under ESXi ...
 

jpmomo

Active Member
Aug 12, 2018
349
118
43
I ordered and received the tyan recommended slimsas cable. it works properly with the pm1733 but not the kioxia cm6. both are pcie 4.0 ssd drives but the kioxia is a u3 drive. the u3 drives are supposed to be backwards compatible with u2 connections. I am able to properly connect the kioxia drive to an asrock rack romed8 mb using an occulink to u2 cable. that cable works properly with both the u2 and u3 ssd drives. That cable differs from the tyan spec'd cable in that it uses a sata power connector instead of the molex. I am not sure if there are other differences in the cables. It could also be a difference in the mb, tyan vs asrock rack. both of these drives are 12v and the tyan cable has a molex connector that apparently works with the 12v drives. I just wanted to add some additional info jus in case someone has similar questions. I also updated to the latest bios. I am planning on upgrading to the milan cpus with this board and have been testing with some early releases.
 

hmw

Active Member
Apr 29, 2019
331
104
43
I ordered and received the tyan recommended slimsas cable. it works properly with the pm1733 but not the kioxia cm6. both are pcie 4.0 ssd drives but the kioxia is a u3 drive. the u3 drives are supposed to be backwards compatible with u2 connections. I am able to properly connect the kioxia drive to an asrock rack romed8 mb using an occulink to u2 cable. that cable works properly with both the u2 and u3 ssd drives. That cable differs from the tyan spec'd cable in that it uses a sata power connector instead of the molex. I am not sure if there are other differences in the cables. It could also be a difference in the mb, tyan vs asrock rack. both of these drives are 12v and the tyan cable has a molex connector that apparently works with the 12v drives. I just wanted to add some additional info jus in case someone has similar questions. I also updated to the latest bios. I am planning on upgrading to the milan cpus with this board and have been testing with some early releases.
[/QUOTE
perhaps the Kioxia needs 3.3V? Along with 5v and 12v?
 

jpmomo

Active Member
Aug 12, 2018
349
118
43
it shows +12v on both the pm1733 and the kioxia cd6. the pm1733 is 2.3A and the kioxia is 2.6A. both drives work ok with the oculink cable for the romed8 mb. I used the following for the romed8:

DiLinKer OCuLink(SFF-8611) to PCIE SFF-8639 U.2 NVMe with Power Cable(1M)

and used the tyan recommended one for the s8030:

Tyan FRU-CS-1230 Slim SAS 8i to 2* U.2 (NVMe) Cable

pcie 4.0 can be pretty sensitive but not sure why it works with the samsung but not the kioxia (other than the main difference being the samsung is u.2 and the kioxia is u.3.
 

jpmomo

Active Member
Aug 12, 2018
349
118
43
If anyone is interested in the notes for BIOS v4.00

Things to note:

- There's two steppings for Milan (B0 and B1)
- New microcode for Rome
- New AGESA for Rome (RomePI 1.0.0.A)


Code:
V4.00      - Date       : 03/11/2021
           - Filename   : 8030V400.ROM
           - Banner     : TYAN S8030GM2NE BIOS V4.00
           - CheckSum   : 6BA3
           - MD5        : 6f367c0aef6d360a11415ce933401376
         
           - Recovery   : 8030M001.CAP (Milan)
                          8030_001.CAP (Rome)
           - AGESA Ver. : MilanPI-SP3 1.0.0.1
                          RomePI-SP3 1.0.0.A
           - Microcode  : B1 0x0A001119 (Milan)
                          B0 0x0A00103F (Milan)
                          B0 0x0830104D (Rome)
           - VGA OpROM  : V1.10
           - LAN OpROM  : I210: PXE: V1.5.85 EFI: V8.3.04 ISCSI: V3.1.30
           - Description: Based on V3.01
                          1. Support Milan CPU Family 19h.
                          2. BIOS setup changed.
                              - Modify “APBDIS” item default to [1].
                              - Added “Fixed SOC PState” item and default is [P0].
                              - Added “DF CStates” item and default is [AUTO].
                              - Change HTTP string to HTTPS.
                          3. Enable CPU Group D platform feature.
                          4. SMBIOS Type1,2,3 retrieve from BMC FRU. (Modify SMBIOS directly is not allowed.)
do you know what the significance of the 2 different steppings is? where did you find that info? Do you know if this is the same for the -2T variant? Tyan mentioned that the difference between the 2 different bios (2.0 vs 4.0) is mainly the 10GE nics.
 

jpmomo

Active Member
Aug 12, 2018
349
118
43
it shows +12v on both the pm1733 and the kioxia cd6. the pm1733 is 2.3A and the kioxia is 2.6A. both drives work ok with the oculink cable for the romed8 mb. I used the following for the romed8:

DiLinKer OCuLink(SFF-8611) to PCIE SFF-8639 U.2 NVMe with Power Cable(1M)

and used the tyan recommended one for the s8030:

Tyan FRU-CS-1230 Slim SAS 8i to 2* U.2 (NVMe) Cable

pcie 4.0 can be pretty sensitive but not sure why it works with the samsung but not the kioxia (other than the main difference being the samsung is u.2 and the kioxia is u.3.
replying to my own post however I still don't have an answer as to why the tyan recommended cable does not work with the kioxia. It seems that both drives require 12v and 3.3v:

for the samsung:

The Samsung SSD PM1733 is implemented in standardized 2.5-inch form factor and gets primary 12V power as well as auxiliary 3.3V (3.3Vaux) power through the indicated pins (#P13~15 for 12V and #E3 for 3.3Vaux in SFF-8639 connector plug) from the host system. For 12V and 3.3Vaux, the allowable voltage tolerance and noise level in SSD are described in chapter 2.5.1, the power consumption in 2.5.2 and the inrush current in 2.7.2. 2.6.1 Maximum Voltage Ratings (12V and 3.3Vaux)

for the kioxia:


Power Requirements
Supply Voltage12 V ± 10%, 3.3 Vaux ± 15%
Power Consumption (Active)19 W Typ.15 W Typ.13 W Typ.
Power Consumption (Ready)5.0 W Typ.

I guess I could try and find a cable like one of the previous posters mention with the sata pwr vs the molex.
 

hmw

Active Member
Apr 29, 2019
331
104
43
do you know what the significance of the 2 different steppings is? where did you find that info? Do you know if this is the same for the -2T variant? Tyan mentioned that the difference between the 2 different bios (2.0 vs 4.0) is mainly the 10GE nics.
If you look at v 4.00 README - it lists the three steppings (two - B0 and B1 - for Milan and one - B0 - for Rome) for the microcode

I'm surprised that there's two steppings, wonder which CPUs are B0. I asked on Anandtech as to the stepping of the CPU they tested but no reply yet