Soldato
- Joined
- 6 Jan 2013
- Posts
- 22,178
- Location
- Rollergirl
Agreed. I have dual M.2 but to have the second installed I would need to drop the GPU down to x8 - so I have only installed the one.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***
- Thread starter Jamin280672
- Start date
Agreed. I have dual M.2 but to have the second installed I would need to drop the GPU down to x8 - so I have only installed the one.
The MSI Z390 MEG Godlike has 4 M.2 slots, while the Gigabyte Z390 Aorus Master has 3 M.2 slots.
Caporegime
- Joined
- 8 Jan 2004
- Posts
- 32,579
- Location
- Rutland
I’m not saying it’s not possible they just take up a lot of surface space, usually wedged between the PCI-E slots or even on the back of the board - which is a mess usability wise and cooling wise. The format can’t replace SATA entirely because of this, or beat it in terms of drives supported on one motherboard.
- Joined
- 12 Feb 2014
- Posts
- 2,997
- Location
- Somewhere Only We Know
Its only ASUS who do it like this, every other board manufacturer has the 2nd M.2 slot running at PCI-e x4 2.0, unfortunately though, this only allows that 2nd nvme drive to run at half speed, that's what I was saying in my previous post, AMD would only have to add 4 more PCI-e lanes to their CPU's to stop this from happening, you could then have 2 nvme's running at full speed, without sacrificing anything else like reduced graphics performance or loosing the use of a PCI-e slot, SATA ports etc.
There is no problem. I hope you don't try to overargue the engineers of the motherboards' manufacturers
You don't use the surface space anyways, it is mostly empty, or filled with elements which shouldn't be there.
Caporegime
- Joined
- 8 Jan 2004
- Posts
- 32,579
- Location
- Rutland
There is no problem. I hope you don't try to overargue the engineers of the motherboards' manufacturers
You don't use the surface space anyways, it is mostly empty, or filled with elements which shouldn't be there.
There clearly is a problem when manufacturers are putting drive slots on the back of motherboards meaning you have to remove your entire system to change the drive.
The issue stems from the format originating from laptops, where the format makes sense as you have a max one or two drives and very limited vertical space. It got bodged onto desktop systems as people adopted it for the speed benefit. The form factors doesn’t work as well meaning we’re left with a random mix of solutions (vertical mounting/add in cards/rear mounting) and holding onto SATA because m2 can’t rival SATA in terms of drives per motherboard space.
The motherboard manufacturers have to make the best of what they’ve got, doesn’t mean the M2 format isn’t stupid. U2 isn’t any better either.
NVMe is a great format and there are many ways to put them on the motherboard, they don't have to go flat on board, that is just the cheapest way, they also go in slots perpendicular to the board like a DIMM and be cooled with a small fan, mobo makers are usind the dad area on the motherboard because it is cheap and easy to do so.
What is the point of 4 nvme slots on z390, supported chips don't have the lanes and bus from cpu to chipset would get saturated if you actually used them all.
What is the point of 4 nvme slots on z390, supported chips don't have the lanes and bus from cpu to chipset would get saturated if you actually used them all.
My guess is that the point is to actually put 4 NVMe drives and actually enjoy using them lol
There aren't enough lanes on a chip that fits z390 that'll run 4 NVMe or enough bandwidth on the motherboard to cope, seems redundant unless you have to disable everything else on the motherboard, what am I missing here? its 4 lanes per NVMe, a GPU needs at least 8, the chip has 16, the rest come from chipset that talsk to CPU through a wet piece of string, you'll bring your system to its knees, pointless slots for styling in this case.
Having more than one M.2 (NVMe capable) slot on a board without the full bandwidth available to the slot isn't ideal, and only makes sense if the drives are not being used at the same time, but given the ideal scenario is transferring data between those drives (for most people) that is when you actually need the bandwidth.
Getting back to the PCI-E 4.0 argument, you could effectively make an 8x PCI-E 4.0 slot into a 16x PCI-E 3.0 slot using a converter card, thus removing the need to add more lanes if the GPU is running at 8x PCI-E 4.0 which would be more than enough bandwidth for current cards. I've been using the 4-way m.2 cards for a couple of years which take up an entire 16x PCI-E slot, and means that you need HEDT to take real advantage of them, but with PCI-E 4.0 the normal desktop segment can now take advantage of this setup, thus not wasting motherboard space either.
Obviously native PCI-E 4.0 drives will come out, that are even faster due to controller improvements and NAND improvements, so we'll be back to the same scenario as now, but hitting 7000MB/s per drive not 3500MB/s, so maybe not as much of an issue since even running at half speed wouldn't be the end of the world.
Getting back to the PCI-E 4.0 argument, you could effectively make an 8x PCI-E 4.0 slot into a 16x PCI-E 3.0 slot using a converter card, thus removing the need to add more lanes if the GPU is running at 8x PCI-E 4.0 which would be more than enough bandwidth for current cards. I've been using the 4-way m.2 cards for a couple of years which take up an entire 16x PCI-E slot, and means that you need HEDT to take real advantage of them, but with PCI-E 4.0 the normal desktop segment can now take advantage of this setup, thus not wasting motherboard space either.
Obviously native PCI-E 4.0 drives will come out, that are even faster due to controller improvements and NAND improvements, so we'll be back to the same scenario as now, but hitting 7000MB/s per drive not 3500MB/s, so maybe not as much of an issue since even running at half speed wouldn't be the end of the world.
Yes, my proposal is to not use them at the same time - when using simultaneously, they begin to share the available bandwidth.
Native will be too much, at that moment I will begin to agree with the people claiming about the diminishing returns. Let's first have 3-4 complete NVMe 3.0 drives on a single board, and then move to the higher speeds..
Native will be too much, at that moment I will begin to agree with the people claiming about the diminishing returns. Let's first have 3-4 complete NVMe 3.0 drives on a single board, and then move to the higher speeds..
Last edited:
That's not true though, since the uses I have for the drives means that I could effectively use drives that operate at 10,000MB/s or beyond and still be constrained by the drives. There is more than one use case for a drive, and consumers don't drive this sort of technology, just benefit from it trickling down.
God, I hope you do realise that the vast majority of systems nowadays are still being sold with 100MB/s 5400rpm HDD....
I am not a huge fan of nvme right now, I understand why it exists, its purposely designed for nand storage. But I guess its the implementation I dont like and that SATA ports get reduced for it.
Installing a NVME drive whilst the board is not in the case is not too bad, but when you have board designers sticking a m.2 slot in between the gpu and cpu and have a installed board, with GPU, CPU cooler etc. its a ridiculous design, compare the awkwardness of that vs sliding a drive in out a drive bay. Also for my use case NVME offers no performance advantage, I feel right now its good for ego boosts, and niche usage patterns that can take advantage of the i/o throughput they offer. But I am not asking for NVME to be removed off boards, thats going backwards, just I feel there needs to be a better system where a NVME drive doesnt need to be placed directly on the board, I actually prefer pci-e based NVME drives, simply because they clearly easier to install, but they seem to be out of favour compared to the chewing gum style drives that go directly in m.2 slots. So its m.2 I dont like not NVME itself.
Installing a NVME drive whilst the board is not in the case is not too bad, but when you have board designers sticking a m.2 slot in between the gpu and cpu and have a installed board, with GPU, CPU cooler etc. its a ridiculous design, compare the awkwardness of that vs sliding a drive in out a drive bay. Also for my use case NVME offers no performance advantage, I feel right now its good for ego boosts, and niche usage patterns that can take advantage of the i/o throughput they offer. But I am not asking for NVME to be removed off boards, thats going backwards, just I feel there needs to be a better system where a NVME drive doesnt need to be placed directly on the board, I actually prefer pci-e based NVME drives, simply because they clearly easier to install, but they seem to be out of favour compared to the chewing gum style drives that go directly in m.2 slots. So its m.2 I dont like not NVME itself.
I have thought about that and to be honest, the only explanation I have is that if they move them further, the signal will have issues because of the distance. Hence, they are not being installed as normal SATA drives in their respective bays.
Caporegime
- Joined
- 8 Jan 2004
- Posts
- 32,579
- Location
- Rutland
That’s not the case, U2 is M2 with a cable and works fine. Just like you can extend PCI-E slots with risers.
M2 has been adopted simply because it was already being widely used in laptops, so drives were easily available. The M2 format needs a good chunk of PCB space (unless mounted vertically) so they’re bunged anywhere they can possibly fit on already crammed mobos, hence the use of the back of the board, between the PCI-E slots etc.
You could bung a whole load of U2 ports on a mobo but nobody does as mainstream U2 SSDs are rare and the cables are expensive and relatively bulky. Plus mainstream CPUs are already burning through their PCI-E lanes as it is.
The results from the poll are very nice for AMD
TBH, I don't wait DDR5, its timings are so loose, that at least a couple more years after its first market introduction, they won't offer anything substantial over the fastest DDR4 that we have today.
Only that a new memory controller has always been a pain in the ass for AMD...
As for AM4, yes, 4 generations of CPUs on the same socket is great!
DDR5 will useful for APU's and should brings the cost of memory down along with increased capacity.
All the moves so far to different DDR's have only increased the cost, not one of them brought the cost down. They have all increased the capacity though.
So will DDR4-4000 be, no? Capacity is related to manufacturing process nanometres, not to DDR version, right?