So with the Asrock x570 creator and a 3950x, after you use the gpu slot at 16x and only the top m.2 slot, you will be out of lanes ?
I want to add a second m.2 PCIe 4.0 drive at full speed, not disabling any sata ports.
If I understand correctly, if I used the 2nd m.2 slot on the motherboard, it will disable the last 4 SATA ports only ?
I read somewhere Asrock Creator did it differently and would disable the third PCIe slot on the motherboard instead.
In any case, will a gigabyte PCIe 4.0 m.2 ssd run at full speed in this second slot ?
Or is it better to add a “PCIe 4.0 adapter m2-card” and run the gpu slot at 8x lanes ?
Currently I have a 980GTX that doesn’t saturate 8x PCIe 3.0, the 2080ti however, does…
But since this board supports PCIe 4.0 and thus double the bandwidth, if I run a 2080ti on this board in 8x lanes, will it be running at PCIe 3.0 or PCIe 4.0 ?
Or do you need the GPU card to be PCIe 4.0 as well ?
I neglected to mention in my post that there are 4 lanes dedicated to the chipset, Ryzen 3000 has 24 PCIe lanes but only 20 are usable directly. Another thing about X570 are that those 16 downstream lanes are a set of 4x4 lane PHYs per the July deck (last pair are reconfigurable) we’ll get back to this.
M2_2 shares its PCIe lanes with the bottom PCIe x16 slot, which is PCIE6 in the English manual, so only one or the other can be used.
This is the key point (more a suspicion as I don’t own the board, but the data makes sense): as M2_2 can only be used when PCIE6 is empty, that M.2 connector basically has a set of 4 high speed lanes all to itself, or one PHY’s worth. Two of the four PHYs in the X570 as advertised can independently reconfigure their mode (which is the ‘pick 1’ option on the slide deck), so those lanes can act as either PCIe or SATA depending on what the end-point negotiates.
In this instance if you put in a SATA M.2 drive in M2_2, X570 will set its ‘pick 1’ to SATA mode at the expense of the PCIe capability on that PHY. By using a M.2 SATA drive you are not sacrificing the 4 X570 dedicated SATA ports, instead you’re throwing away 4 PCIe lanes which would be physically impossible to use anyway as PCIE6 can’t be used with M2_2, and you’ve just installed a SATA M.2 drive. In the case of a PCIe M.2 drive in M2_2, X570 sets its ‘pick 1’ to PCIe and you still have the 4 dedciated X570 SATA ports. The remaining PHYs in either case would only ever be PCIe as it would be hardwired somehow to devices per my previous post. Running an M.2 PCIe SSD in M2_2 should run at near full speed, accounting for the X570 x4 Gen 4 link etc.
With the 980GTX and 2080Ti, none of those are PCIe Gen 4 so they would be running at Gen 3 speeds regardless of where they sat on the board.
Does anyone know if this actually uses ecc including 2-bit errors? From what I understand many Ryzen boards “support” ECC but don’t actually make use of it
So if I use a second drive in m2_2, that would mean the three PCIe slots would run at 8x, 4x, 0x then ?
In case that I wouldn’t use the m2_2 slot, but instead use a “PCIe m2 adapter card” and put it in the third PCIe slot. They would need to run at 8x, 4x, 4x right ?
In any case, using a second m2 drive, would result in my GPU not being able to run at 16x lanes, eventhough these videocards don’t get near that speed yet.
So if I want to add another card to my system in the future, I would have another 4x lanes on the second PCIe slot, while running the gpu at 8x.
The top 2 PCIe x16 slots (PCIE1 and PCIE4) come directly from the CPU, whereas M2_2 and PCIE6 come from X570. In your example, running 2 cards in the top 2 PCIe x16 slots and a drive in M2_2 will give a distribution of x8, x8, x0 for slots PCIE1, PCIE4 and PCIE6 respectively.
Keeping this example but removing the drive in M2_2 then installing a card in PCIE6 gives a distribution of x8, x8, x4 (as M2_2 is disabled)
The only slot which can run at 16 lanes is the PCIE1 slot, and for that to happen two things need to be true:
PCIE4 has to be empty (otherwise x8 x8 occurs)
A CPU (not APU) needs to be installed in the AM4 socket.
What happens in any other slot doesn’t impact the interaction between PCIE1 and PCIE4.
A second drive in M2_2 will only change PCIE6 at the bottom.
Came across this thread after I set up my system with a m.2 in the second slot and a SATA 4x pcie 2.0 add-in card in PCIE6. The manual doesn’t given you any info not to do so that I noticed. I don’t have anything in PCIE4 at the moment.
One thing I can confirm is that PCIE6 is functional as is the second m.2 slot, one does not disable the other. I’ll do some experimenting with moving it up to PCIE to see if that changes write speeds to either the drive site on the add in card or the m.2 individually or being written to at the same time.
Edit: No difference in write speed or any other noted problem between the drives on the 4x sata card or the 2nd m.2 drive with the Sata card in the 3rd or 2nd slot. The sata HDD are only writing at ~100MB so it could be that there is reduced performance with the 2nd m.2 drive if you have something saturating the bandwidth of the 3rd pcie slot but for my tests, you can have both 2nd m.2 slot and the 3rd pcie slot populated at the same time, and with no no noticeable negative impact (in my very limited test).
I don’t have another PCIE device at the moment to test functionality with all three PCIE slots and the second m.2 slot populated.
Hello, this point is still unclear to me. I am interested in this board and want to be sure i wont cause bottlenecks.
Ill be running:
(2) 2080 ti (in x8 x8)
(2) pcie m.2 nvme gen 4s.
(2) Sata Hdd
(2) Sata SSd
Ill be running external TB SSDs as well.
Is there any advantage to running this setup on the x570 taichi with titan ridge card? I dont need any of the enhanced features of the creator, but It might be nice to have a third pcie slot for space for the bottom gpu on the creator motherboard where it would be partially air inhibited by the titan ridge card on the Taichi.
No problem with all those devices. No shared resources/other problems.
But note the only reason you would want to pay the premium for this board is because you want the 10gbit NIC and Thunderbolt 3 (usb 4 in the future?) built in.
Hi CBME, so you effectively have 12 active SATA ports in your configuration, with an M.2 PCIE installed in the 2nd slot? (8 onboard, plus 4 from add-in card).
Do you also have an M.2 in slot 1?
What model card (or SATA chip if you are aware) is your SATA add in card using?
What BIOS are you running, and do you also happen to know what IOMMU groupings this configuration presents?
I can not confirm that it is possible to populate the M2_2 and the x4 (PCIE6) slot at the same time. It’s seither one or the other. I recently purchased the mainboard and populated all three mechanical x16 PCIE slots with three single slot GPUs and both of the M2 slots with PCIE 3.0 NVMEs. AS soon as I remove the third GPU from the x4 (PCIE6) and reboot the system, the last M2 NVME becomes available.
What it is further disturbing is that ASRock hasn’t stated anything about this limitations in the manual.
I also watched @wendell review but he hasn’t mentioned it either.
Lastly there is a setting available to config the PCIe slots but it isn’t clear if it controls bifurcation or how the lanes are split between the three mechanical x16 slots. I tried nevertheless without success.
I’m really disappointed about the way ASRock and other Reviewers handled this behaviour.
Hi All.
I purchased this MB and have a couple of questions for the experts about the best way to configure my build. This PCIe lane stuff is a little confusing for a first time builder and now ex-Mac user.
I’ll mainly be using this machine for video editing/grading 6k BRAW footage and some Blender rendering work. 6k BRAW at 3:1 compression (worst case scenario) requires 323 MB/s.
I use DaVinci Resolve and Puget Systems recommends keeping OS and programs on one drive, video footage on a second drive, and scratch and cache files on a third drive for max performance. Maybe this is overkill with the speed of these new NVMe drives but might as well optimize the build best as possible.
I have a 2080 ti which I will put in the first PCIe x 16 slot. I have a Rocket 4.0 1TB NVMe that I’ll put in M2_1 for OS and programs. Now trying to figure out the best approach for the second and third drives for project and cache files.
Should I use the M2_2 slot + an M2 expansion card in another PCIe slot? Or maybe Raid 10 and some SATA SSDs?
I would like to reserve the SATA ports for a RAID 10 array to store large files longer term, but have a separate Synology NAS to keep inactive project files.
What would be the best way to set this up to maximize all of the lanes and get best performance out of NVMe or a combo of NVMe and SSD/HDDs?
“It’s either one or the other.” Again, as I noted, that isn’t the case. If what you are saying is that you can’t have all PCIE slots and the second m.2 populated, that could be the case as I couldn’t test it. But you absolutely can have something in PCIE6 and the second m.2 at the same time as I already noted having done. You didn’t mention trying something in PCIE 6 without something in PCIE 4 so it is possible that is the issue you had. It may also be the type of device.
Overall I agree that Asrock was sketchy in not making the possible conflicts and limitations clear.
(just replying in case anyone comes across this thread and has simliar questions - didn’t get notified that there was a reply here back in Jan).
I have 13 SATA ports (card has 5) and 11 being used. Only one that is in active use all the time, one to a SSD. The rest are my blu-ray movies (and a copy/online backup) that are used whenever system is scanning or family is watching.
Both M.2 slots are populated, the first with the OS on a pcie 4.0 drive and the second with pcie 3.0 drive.
SATA card - cheap one. JMicron JM535 https://www.amazon.com/gp/product/B07ST9CPND/ref=ppx_yo_dt_b_search_asin_title?ie=UTF8&psc=1
Bios was the latest. Only thing I’ve had trouble with is my 4 g.skill dimms which I have to overvolt to get the system to boot at 3600.
IOMMU - don’t know, haven’t considered it.
The second nvme m.2 slot (M2_2) will not be recognized by the BIOS/UEFI. It will only available after booting a OS (e.g. Linux) from another device (e.g. M2_1). It’s not possible to boot from the second nvme m.2 slot (M2_2). I even tried with different bootloaders like grub, rEFInd and OpenCore. ASRock messed up.
ah[quote=“tristank, post:35, topic:146682”]
It’s not possible to boot from the second nvme m.2 slot (M2_2). I even tried with different bootloaders like grub, rEFInd and OpenCore. ASRock messed up.
[/quote]
Ah. That I’ve not tried as I’ve only had my boot drove in slot 1 so I’d not come across that design failure.
Hi North, I just bought the Asrock Creator x570 w/ AMD 3950x. Which exact ram are u using, how many gb, 2 slots or 4. Did you overclock and if so what settings did u use? Planning to learn DaVinci Resolve. If u have 64gb ram do u think 128gb would be better or overkill? Any assistance would be appreciated.
Looks like for Memory QVL (Matisse) you are limited to 2666. Most of the QVL are Kingston.
Mayb 2 x 32GB to start.
$662.84 X570 Creator at amazon.ca
$639 AMD Ryzen 9 3900X at Canada Computers
2 x $222 Kingston 32GB DDR4 2666MHz 288pin DIMM ECC Reg Memory Module KSM26ED8/32ME at Newegg.ca
So if you want a 9211-8i in the bottom x16 at x4 slot and the 2 other x16 slots at x8 to have GPU, can you have a x1 GPU like a GT710 to boot Unraid? So in this way if you want higher LAN you are stuck picking a MB with the higher speed LAN on the MB? Or is there a way to pick say X570 Phantom Gaming X MB and in the future how would you add 10G LAN, USB3 port adapter?
It sounds like there was some confusion about whether or not M2_2 and PCIE6 could be used simultaneously. If anyone is still wondering, I can share that I was able to successfully populate PCIE6 and M2_2 with NVME SSD’s, and Windows recognized them just fine. The configuration this worked for was:
PCIE1 - GPU
PCIE4 - NVME SSD
PCIE6 - NVME SSD
M2_1 - NVME SSD
M2_2 - NVME SSD
All SSD’s are Intel. CPU is a 5950x on the ASRock Creator. Hope this helps.
Swapped out my main 4.0 SSD with a new one and for the heck of it was doing speed tests. Found that it tested at expected speeds but that there was still something strange with transfers between NVME’s.
When trying to transfer large files between the drives it maxes out at 1.8 when writing to my 4.0 drive and 2.0 when writing to my 3.0 drive. It starts and stays at that speed throughout a 30GB test transfer. The 1TB 980 Pro has the ability to write at 5GB/s for about ~150GB before the cache is full and it drops to a little under 2GB/s. The slower 2TB HP ex950 3.0 drive tests at ~3.5/2.9GB/s and is tested to sustain writes around 2.9 for 317GB so it is not the bottleneck.
2GB/s is the 4.0 x1 speed limit though I’m not sure if the system is dropping to 1x on the SSD 1 for some reason of if that is just a coincidence.
Anyone else seeing faster transfers or aware of what might be my bottleneck?
NVME controllers using windows default. All drivers current. Graphics card on PCIE 1 and SATA card on PCIE 4. On Windows 11 but saw this on 10 as well.
