Question Questions about a MB I'm looking at (ASRock Z790 PG Riptide)

I like that this board has a total of 6 m.2 slots (5 SSD, 1 Wi Fi) with low profile heat spreaders that can nest 3 NVMe drives under them and even fit under a massive 4 slot GPU.

My concern is about the top two m.2 slots, which occupy the space of one NVMe drive, both facing inward. One of those two top slots is a Gen 5 that can operate at twice the bandwidth of Gen 4, for when Gen 5 NVMe drives come out. So why I wondered did they put two slots in one physical space when pretty much all m.2 slots are backwards compatible with previous gen drives?

The answer not only seemed a HUGE disappointment, but a bit baffling. The pdf manual says if you occupy that 5th m.2 slot, which is the Gen 5 one, the Pci-E 1 slot is automatically downgraded to 8x. This I thought would be unacceptable if running a behemoth like the RTX 4090 I eventually plan to get, as it requires a lot of power and bandwidth.

I guess though what it comes down to is how much of an FPS drop would you realistically get on a Pci-E Gen 4 slot if running at 8x vs 16x, since the 4090 can't even use Gen 5. This would also pretty much be the equivalent of running it at 16x Pci-E 3, vs 16x Pci-E 4. I've seen bandwidth tests way back when Pci-E Gen 3 came out, showing very little frame loss even on high end GPUs, about 2 to 3 FPS. But of course the 4090 is a beast like we've never seen before.

So my questions are, does anyone know why a Gen 5 m.2 slot would rob Pci-E bandwidth on a MB, vs just SATA 3 bandwidth like Gen 4 m.2 slots do? Also does anyone know of a site that's doing Pci-E bandwidth testing on the 4090?

[Pre Edited]
Well damned if Steve at GamersNexus didn't seem to answer both questions before I could even ask. He HAS indeed run a Pci-E Gen 3 vs 4 test on a RTX 4090 FE, and aside from some synthetic benches, saw the usual 1-3% difference at the res the card was made for, 4K (3 FPS at most). The Total War Hammer game he tested was the only exception, but only at 1440p and especially 1080p, where the difference in FPS was as much as 15%. I can only assume this might be due to the game being more CPU dependent (not sure it is).

He also showed a Pci-E gen chart with specs up through Gen 7, and it appeared to make it obvious why a Gen 5 m.2 slot requires so much bandwidth. From what I've read a Gen 5 m.2 slot runs at the same 128 speed a Pci-E Gen 7 will run at, when THOSE beasts come.


[Edit 2]

I just found out W1zzard, of TechPowerUp, the guy that wrote CPU-Z, GPU-Z and is spot on with his reviews, also did a Pci-Ex Gen 3 vs 4 bandwidth test on the RTX 4090 FE. His results showed even less of a difference. He also pointed out at the start of the review that the use of a Gen 5 NVMe drive will only rob Pci-Ex bandwidth on Intel based MBs that have such a slot. He said the AMD based MBs that have a Gen 5 NVMe slot don't rob ANY Pci-Ex bandwidth when that slot is used.

I'm not sure why that would be the case, but I have a feeling more will be revealed when we see tests pitting the Intel and AMD Gen 5 m.2 MBs against one another.

 
Last edited:

Lutfij

Titan
Moderator
Intel and AMD don't make progression with tech in the same method, in spite of operating within a similar ecosystem/environment. Either of them can introduce technology that might be adopted by the other, taking Thunderbolt for example being exclusive to Intel for a time.

As for your bandwidth question, why it would rob, might fall into the same discussion of lane allocation by a platform/chipset/controller. I've come to learn that when you have a board that's got all possible connectivity options, that they're often more than not, too good to be true. It's why I look into a board/components that a step or two down with ports/slots that are not sharing lanes or even using a third party controller to allocate more lanes than the CPU or chipset can assign. One other thing that can and will affect lane allocation is how the board maker(not the chip maker) decides to implement the slots on their board for either reduction in production/componentry sourcing cost or reducing manufacturing time(both will affect profit skimming by board maker).
 
Intel and AMD don't make progression with tech in the same method, in spite of operating within a similar ecosystem/environment. Either of them can introduce technology that might be adopted by the other, taking Thunderbolt for example being exclusive to Intel for a time.

As for your bandwidth question, why it would rob, might fall into the same discussion of lane allocation by a platform/chipset/controller. I've come to learn that when you have a board that's got all possible connectivity options, that they're often more than not, too good to be true. It's why I look into a board/components that a step or two down with ports/slots that are not sharing lanes or even using a third party controller to allocate more lanes than the CPU or chipset can assign. One other thing that can and will affect lane allocation is how the board maker(not the chip maker) decides to implement the slots on their board for either reduction in production/componentry sourcing cost or reducing manufacturing time(both will affect profit skimming by board maker).
I'm mostly just concerned about how AMD can be running Gen 5 NVMe drives without taking away any Pci-Ex lanes, when Intel is taking 8 of the Pci-Ex lanes when using a Gen 5 NVMe. I have started a conversation with W1zzard on TechPowerUp, and I expect to hear from him soon. I'll let you know what he says, although he may not know yet.
 
Last edited:
Well, so far his answer only mentioned the obvious I had clearly stated to him I already knew, but avoided saying how AMD is managing to do this without stealing Pci-Ex lanes. Worse yet, he directed me to ask it on their often caustic public forum vs PM. This is exactly why I stopped hanging out on TPU years ago. I used to think W1zzard was an exception to that problem, now I'm not sure.
 
Dec 2, 2022
4
0
10
The latest AMD AM5 CPUs have 24 PCIe 5.0 lanes and the latest Intel CPUs (Raptor Lake) only have 16 PCIe lanes, so there will always be more of a trade-off with Intel systems. AMD has always had more available PCIe lanes available for peripherals.
 
The latest AMD AM5 CPUs have 24 PCIe 5.0 lanes and the latest Intel CPUs (Raptor Lake) only have 16 PCIe lanes, so there will always be more of a trade-off with Intel systems. AMD has always had more available PCIe lanes available for peripherals.
Yes, that's what I found out, but when I look at i7-13700K vs R7 7700X reviews, Intel not only has the gaming performance edge, the difference between running games at 16x Gen 4 vs 16x Gen 3 (equivalent to 8x Gen 4) is miniscule.

Besides that I'm reading that Gen 5 drives will have 12000 to 13000 Mb/s read/write speeds, and I've only been using my current NVMe at the lower speed to avoid it taking up a SATA 3 port. So it runs at less than 2500 Mb/s, and any Gen 4 drive I use will be about 3 times that. I think that's plenty good enough for OS and programs, and I don't play MP anymore so I don't worry about MP maps loading in the blink of an eye anyway.

I am now looking at the bigger picture, and it still looks more like Blue team than Red team for me.
 
Dec 2, 2022
4
0
10
Yes, that's what I found out, but when I look at i7-13700K vs R7 7700X reviews, Intel not only has the gaming performance edge, the difference between running games at 16x Gen 4 vs 16x Gen 3 (equivalent to 8x Gen 4) is miniscule.

Besides that I'm reading that Gen 5 drives will have 12000 to 13000 Mb/s read/write speeds, and I've only been using my current NVMe at the lower speed to avoid it taking up a SATA 3 port. So it runs at less than 2500 Mb/s, and any Gen 4 drive I use will be about 3 times that. I think that's plenty good enough for OS and programs, and I don't play MP anymore so I don't worry about MP maps loading in the blink of an eye anyway.

I am now looking at the bigger picture, and it still looks more like Blue team than Red team for me.

No need to tell me - I just bought an ASRock PG Riptide board and i7-13700k CPU and am extremely excited o get it running. I have a new PCIe 4.0 NVMe SSD (which I won't install in the M.2_5 slot to take away half the PCie bandwidth.) I have no idea whether a PCIe 4 graphics card at x8 will actually show any difference than a PCIe 5.0 card at x16 - and I really don't care - so the ASRock Z790 PG Riptide seems to be the perfect choice for me. But I am a lowly Linux-only user, so hey - Linux rules!
 
No need to tell me - I just bought an ASRock PG Riptide board and i7-13700k CPU and am extremely excited o get it running. I have a new PCIe 4.0 NVMe SSD (which I won't install in the M.2_5 slot to take away half the PCie bandwidth.) I have no idea whether a PCIe 4 graphics card at x8 will actually show any difference than a PCIe 5.0 card at x16 - and I really don't care - so the ASRock Z790 PG Riptide seems to be the perfect choice for me. But I am a lowly Linux-only user, so hey - Linux rules!
Cool, are you planning to use the heatsink ASRock supplies for the NVMe drive, or are you using one you got with your drive or some 3rd party one? If you use the ASRock one, please let me know what the temps are at idle and load. BTW, what case are you using and what CPU cooler? I plan to get the Phanteks Enthoo Pro II with glass panel, and a Arctic Liquid Freezer II 360 and mount it at the top blowing out. And also what NVMe drive do you have? I'll probably just use my EVO 970 500 GB at first, but eventually I plan to get a WD N850X Gen 4, maybe even a 4 TB, in hopes for a price drop after the Gen 5 ones come out.

I'm ready to get the CPU, MB, RAM, case, and cooler, but am waiting for the RAM I want to get in stock.

And actually sites are testing for Pci-Ex 4 at 16x vs Pci-Ex 3 at 16x, because the gen 3 at 16x is equivalent to gen 4 at 8x, as there are no GPUs as far as I know that use gen 5 Pci-Ex yet, at least not Nvidia ones.
 
Last edited:
Dec 2, 2022
4
0
10
Cool, are you planning to use the heatsink ASRock supplies for the NVMe drive, or are you using one you got with your drive or some 3rd party one? If you use the ASRock one, please let me know what the temps are at idle and load. BTW, what case are you using and what CPU cooler? I plan to get the Phanteks Enthoo Pro II with glass panel, and a Arctic Liquid Freezer II 360 and mount it at the top blowing out. And also what NVMe drive do you have? I'll probably just use my EVO 970 500 GB at first, but eventually I plan to get a WD N850X Gen 4, maybe even a 4 TB, in hopes for a price drop after the Gen 5 ones come out.

I'm ready to get the CPU, MB, RAM, case, and cooler, but am waiting for the RAM I want to get in stock.

And actually sites are testing for Pci-Ex 4 at 16x vs Pci-Ex 3 at 16x, because the gen 3 at 16x is equivalent to gen 4 at 8x, as there are no GPUs as far as I know that use gen 5 Pci-Ex yet, at least not Nvidia ones.

My setup is a little more modest, but I have 64GB of memory (I run lots of VMs). I plan to use either my PNY RTX 3060 card or just the embedded Intel for graphics. I doubt I'll use the MB heatsinks for M.2 as mine currently don't seem to fluctuate much once at operating temp (mostly Samsung Evo 980s). I just ordered a pretty basic case - NZXT 710 - mainly for the two USB3 type A and one USB C ports on top.