So Pci 4.0 is just a phase / gimmick

PAM. One can ask if it is still digital or it is actually digital with analog modulation. Technically it was that way always. But we had logical 0 an 1 based on voltage thresholds. And then we had uncertainty area for error. Now PCIe 6.0 has logical 0, 1, 2, 3 per pulse and that means there is uncertainty state in between 0 and 1. 1 and 2. 2 and 3. I wonder how sensitive this will be for UP/DOWN voltage spiking. And changes to different voltages CPU uses. MB's will definitely become quite expensive even for low end chipsets. And then there is question: "How are we going to use 256GB/s bandwidth between CPU and GPU on PCIe 6.0 x16?" While GPU does not have practical use for RAM that reads 50GB/s for per frame use, maybe CPU can have use for VRAM that can actually deliver those data. Or are we going for GPU to GPU communication for purpose of 2 GPUs acting as one?

PCIe 3.0 stuck around for a long time, I wonder where the next sticking point will be. On the regular consumer side, we aren't even making much use of PCIe 4.0's bandwidth, much less what 5 and 6 offer. I don't know a ton about what the PCIe lanes actually do so I'm curious to see how this shakes out. It reminds me a bit of USB, where 2.0 was around forever and then we got too many versions to keep up with.

IchimA:

So Pci 4.0 is just a phase / gimmick

How ? Just because faster PCI-E arriving in the future ?

Considering all the technologies being mentioned are backwards compatible, I'm not seeing the pressing issue with advancing too fast. Hard drive tech is advancing fast too. A few extra variations of spec is all.

0blivious:

Considering all the technologies being mentioned are backwards compatible, I'm not seeing the pressing issue with advancing too fast. Hard drive tech is advancing fast too. A few extra variations of spec is all.

Actually SSDs are not that magical in terms of IOPs. Yes, SSDs moved IOPs from hundreds/thousands HDDs had to around 100k. And NVMe moved it even close to 500k. But there is still space to grow.

IchimA:

So Pci 4.0 is just a phase / gimmick

We know that before AMD put it on his product... On the diagram from this period it was shown as a short living standard... now it's confirmed 🙂 But even if you have PCIe 3.0 or even 2.* you can still use it, and so the 4.0

Does it makes sense? Todays CPUs are not able to fully saturate PCI 4 except is some specific situations. And today, there are very few PCI 4 products, mainly SSDs strictly for AMD. Intel plans, talks, about PCI 5 and uses it as excuse for skipping PCI 4. Why PCI 6 already? PCI 5 will be, probably, out in 2021 and as with PCI 4 some time later we will have SSD. Before PCI 6 ends up in hardware we will have PCI 8. Makes no sense. My bigger problem with PCI is too few lanes in consumer CPUs. Threadripper and Xeon have a lot but not consumer CPU, even expensive ones.

Specs finalize, production and then from server to regular desktop and average consumer so I would expect 4.0 and 5.0 to both get a bit of time before 6.0 is common and then pricing and availability plus CPU and motherboard implementation and how these traces and lanes will be set up and utilized. Will be interesting to see if the newer chipsets and motherboards plus upcoming processors tweak the current 4.0 support and then how Intel and AMD utilizes this. Probably seeing 5.0 first though I have no idea if that's late 2021 or even 2022 for desktop hardware let alone 6.0 here. Assuming it's still up and down compatible with earlier and current standards and hardware anyway so not really too critical and far as desktop / consumer hardware goes I don't see it being too critical although higher-end M.2 / NVME implementations might see some fun stuff depending on what happens here and on the OS in terms of actually utilizing this fully. 🙂 GPU's well that's where that backwards compatibility might be a bit problematic as it would be 3.0 or 4.0 based something like that but aside from AMD's 5500 I think it was (x8 based.) it shouldn't be too critical? EDIT: Eh too early to say I suppose but I don't think 3.0 x16 is struggling that much with current graphics cards still so probably a lesser issue. (Probably other possible uses as speeds improve and maybe other benefits.) EDIT: Need to read up on this compatibility too, guessing sending more power through it and having GPU's fed above the current 75w something could be difficult but it'd be neat with lower-end cable-less as a possibility or less 8x-pin tripple connectors at least. 😛 Eh as if, assuming the card can take it someones going to go it for ha ha. (Well less extreme editions could still see some possible use of it too of course.) EDIT: Nope PAM sounds more like sending more data nothing on voltage or wattage. (Still nifty though along with some of these other features.)

Don't get too excited. It has historically taken about 2 years for any major new tech standard to begin reaching consumers, so that 2021 date is actually 2023 at the earliest. We haven't even seen any solid indication that PCIe 5.x that was cleared early last year is coming, and the snail's pace that Intel pushes new tech out at means we will probably not see it until 2022.

All I hope to see from PCIe 6.0 is to drop x16 slots. We still don't have anything that saturates 3.0 @ 16. What we need is more bandwidth for smaller slots. Motherboards can be made simpler and cheaper if we just drop x16.

schmidtbag:

All I hope to see from PCIe 6.0 is to drop x16 slots. We still don't have anything that saturates 3.0 @ 16. What we need is more bandwidth for smaller slots. Motherboards can be made simpler and cheaper if we just drop x16.

So, where would you to put a 1.5-2Kg GPU? In the air with the power of magic? PCI-e 4.0, 5.0 and 6.0 are already a big deal for smaller slots.

Alessio1989:

So, where would you to put a 1.5-2Kg GPU? In the air with the power of magic? PCI-e 4.0, 5.0 and 6.0 are already a big deal for smaller slots.

Have you tried to break off a PCIe slot? Even the non-reinforced ones are on there pretty good. EDIT: Also, there could still be physical x16 slots but they're all electrically x8.

The epic sag is hard to avoid without a GPU bracket these days. I succumbed last year, and bought a bracket. Recent GPUs are the size and weight of a small Buick.

Lol...at the end of the day..... All these version 4..5...6.. .....meant for what objective(s)? IMO, only marketing strategies.....for *@#$#@:) ----------------- Made me remember of USB gen x...y..zzzzz;)

0blivious:

The epic sag is hard to avoid without a GPU bracket these days. I succumbed last year, and bought a bracket. Recent GPUs are the size and weight of a small Buick.

True. I think the impetus is going fall to case manufacturers to support the GPU weight more than the mobo slot. I'd love to see the format shift (though I don't know exactly how) to better support a beefy, cpu-like cooler on the GPU. Maybe liquid coolers shifting the cooler weight from the pcb to a radiator is ultimately the easiest (if more costly) solution.

For me the future is here. Yesterday just bought 2 M.2 EVO Plus. +3 860 I've had and I'm more than fine.

considering all pcie 4.0 motherboards are more expensive and require better quality to handle 4.0 i expect to see non watercooled 1300$ motherboards "2080ti style" also pcie 4.0 this far for me has been a gimmick I get the same speeds in real use with 4.0 drives or 3.0, file copy is around 2.1Gb/s on both, don't bother with it

schmidtbag:

Have you tried to break off a PCIe slot? Even the non-reinforced ones are on there pretty good. EDIT: Also, there could still be physical x16 slots but they're all electrically x8.

So what are you asking? Most of the current motherboards of the market only have 16 lines on one slot.

I'm excited from a Server point of view, I'm already running dual port 100gb Mellanox cards and raid cards (MegaRAID 9560-16i) that are PCIE 4.0 attached to NVMEs that saturate the PCIE 4.0 quite abit, extra headroom is always great and much needed in the Server space.