How are people getting along with setting up your NVME SSDs? If your system isn’t a Threadripper, Xeon or Series X processor what are you getting away with in using multiple M.2s? Ryzen 5000, Intel 10th / 11th gen are budgeted in there own motherboard chipset / cpu layouts. Some mbo’s have more than 2x m.2s on them and they disable other IO depending on how many M.2s you scale. Enabling 3x Gen 4 PCIe m.2s on a 5800x reduces your x16 graphics down to 8x. Or perhaps your 10th gen Intel can run 4 m.2 leaning on the iGPU. The 11th gen Intel may lend you 4 more lanes towards an extra NVME AIC or JBOD/usb/sata/U.2 ports.
Can you share some of your hardware combos? I’ve been dance back and forth between Intel and AMD setups in my mind but picking a motherboard match for either brand has been somewhat of a challenge factoring in the PCIe lanes. Thanks.
Most of the tests I have seen indicate reducing from x16 to x8 isn’t a big deal performance wise on a GPU – I wouldn’t sweat it too much. A bit of Google should confirm that.
Free ur mind from those mundane frustrations. Get a TR and enjoy 64 lanes on std TR, 128 on Pro. No worries, lots of nvme accepted PCI-e 4.0 full speed.
What are you using? Those Gen 4 Hyper M.2 x16 riser cards are pretty rare.
Some boards mix in PCIe 4.0/3.0 in rather odd combos. Also it seems only a few offer ram overclock speeds 5000+MHz.
My previous searches found ECC, dimm.2 and U.2 features in AI workstation type motherboards looking pretty interesting for desktop based cpus. But yeah the cost goes $360+.
Getting RAM running at 5000 MT/s is usually a lot harder than picking out the most expensive RAM rated for it and buying that. More so if you’re filling all the RAM slots. Might be easier with 11th Gen Intel, but I haven’t looked into overclocking memory on that platform.
Edit: x8 PCIE 3.0 is basically as fast as x16 as to not matter in any real world situations. If you’re a competitive, high FPS Pro-gamer you probable won’t want to plot while playing. x8 PCIE 4.0 is as fast as PCIE 3.0 x16, and really shouldn’t matter (if you’re using a 4.0 GPU).
Nice setup! The X570 chipset offers 16 more lanes besides the CPU. I guess each X570 MoBo distributes those 16 with 16-20 cpu lanes in there own way. Any reason why you had used a 2nd riser before filling the 2nd M.2? Is it because the 1st M.2 is for the OS and raid 0 on the 2 riser M.2s?
Which graphics card did you use for the X1 slot? That’s pretty slick .
I had some heating issues with the m.2 and with heatsinks they would be very tight against either the graphics card or the other riser. The risers were only 25 bucks and come with good heatsinks.
My os is on a sata ssd.
I put a request out for a x1 graphics card on a hardware marketplace website and some guy had an old radeon hd4350 lying around he sold me for 12,50
Be carefull with just adding PCI-e lanes from CPU and motherboard.
Ryzen has 24 PCI-e lanes.
16 go to the 2 x16 PCI-e slots. If you populate both each runs at x8, if you only populate 1 it runs on x16.
4 go to 1 of the m.2 slots.
The last 4 go to the chipset. Everything connected to the chipset shares those 4 lanes.
So the USB ports connected to the chipset, any PCI-e slots other than the 2 x16 slots, the 2nd m.2 slot, 3rd m.2 slot, network and SATA all share 4 lanes to the CPU.
Usually that sharing is not a problem, 4 lanes is a lot of bandwidth. But with PCI-e 4.0 SSDs actually using the available bandwidth of 4 lanes, you might at least theoretically run into a bottleneck if you connect them to the chipset.
Yes that’s a good point actually. I didn’t really understand how that worked but I just read up a bit on that.
It’s quite odd that x570 can be marketed as offering an “extra” 16 lanes of pcie 4.0, when all those connections have to share the same 4.0x4 connection to the CPU…
The only difference with this motherboard seems to be that there is only one x16 slot connected to the CPU
The other x16 slot is “fake” and only runs on x4 via the chipset.
But again, with luck on your side, everything works out
So the chipset is running atm:
10 x sata
one x4 nvme ssd
one x1 graphics card
As the nvme’s are actually not running anywhere close to 4.0 speed when plotting, this should be fine for the CPU-Chipset connection.
The other two x4 nvme’s are directly on the cpu , one in the top m.2 slot, one in the top x16 slot.
.