Crosshair vii hero PCIe lane distribution confusion.

[dasrichard]

I have some PCIe questions obviously, and please don't tear me apart, I'm not as versed in motherboard tech as many of you.  

(What I'm running)

Crosshair vii hero

Ryzen 2700x

msi 2080 ventus

2xSamsung evo plus 970 1tb 

4x(older)850/840 evo 256GB drives in raid 0  (these were salvaged out of my old Fx9590 pc when I built this one so I thought it to be better than discarding them)

32gb Vengeance 2666mhz

 

So obviously, I've used both m.2 slots in the board already (I do know one has it's own 4x lane to the cpu and the other is stealing 4x lane off the PCI16 slot).  I know this drops me down to 8x pcie on my graphics card.   My first question is: am I impacting my 2080 ventus running at 8x pcie vs running it at 16x pcie?    My second question will pertain if the graphics card is fine at 8x pcie: If I use the ASUS hyper pcie board on the second pci16 slot and do an NVME m.2 raid 0 with 4 additional NVME m.2 drives to upgrade from my older ssd drives, is this going to interfere with the existing m.2 that shares this second 8xpcie lane ?  (Or maybe there's an even better solution I haven't even discovered)

 

I'm already somewhat confused with the limited information I can find, and I'm not even sure if the PCIe Lane distribution images I've googled are entirely accurate, and no I won't be upgrading to new x570 and Ryzen 3000 series until at least early 2020. 

(On a side note: I'm also confused why Samsung Magician will benchmark my 970 evo plus drives at 3500mb/s read and 3300mb/s write but userbenchmark will show only 2300mb/s and 2100mb/s, or is this just more evidence of why userbenchmark is a pos?)

[Tannah]

Well I can give a bit of anecdotal information, I also have an X470 Asus ROG Crosshair VII Hero.

I did find that running MSI Duke 2070, and a M.2 NVME PCIE on the CPU M.2 slot did cause a degradation on GPU benchmarks.

But I think I had setup the perfect storm for that. I don't know if I would have noticed as much in games tho. Also if I had run a SATA m.2 NVME it may have been different too.

I know by the specs I didn't expect to see a difference but I did.

[dasrichard]

Thank you so much for the quick response.

I'm somewhat disheartened this is the case with this motherboard.  In my next build I'll pay better attention to this.   At this point I'm regretting not just buying a 2tb NVME PCIe instead of two 1tb models and leaving that 8x lane alone on CPU M.2_2 and only using M.2_1. 

I previously had never put much attention into PCIe, as with more recent builds the only card slot being used is the dedicated graphics card.  

 

 

Can you quantify approximately what you remember the performance loss was of running your 2070 Duke on PCIex8?  It's okay if you're unsure.   Maybe I need to just pull the one drive and check the difference out.

[Tannah]

Unfortunately I can't, I know the benchmarks I noticed it the most on were Firestrike and UserBenchmark. When I removed the M.2 NVME by the CPU I saw the GPU number go back to where I expected. I noticed it because with my SATA SSD I had gotten higher scores than with the PCIE NVME, which told me something was wrong.

I had sprung for 2 500gb Samsung 970 Evo SSD's, I was disappointment so I sold one to a friend and got a SATA drive.

[dasrichard]

Thank you so much for the insight!