PCIe 4.0 and Intel?

[trparky]

With the launch of several PCIe 4.0 devices including SSDs that are showing faster sequential read and write speeds (which really only benefits certain workloads), do you think that as more PCIe 4.0 devices come to market Intel will be forced to release PCIe 4.0 compatible hardware?

 

[bug]

With the launch of several PCIe 4.0 devices including SSDs that are showing faster sequential read and write speeds (which really only benefits certain workloads), do you think that as more PCIe 4.0 devices come to market Intel will be forced to release PCIe 4.0 compatible hardware?

I've heard Intel may be going straight for PCIe 5.0 (whenever they'll mange to release something), but "forced to release"? As you have noted, the benefits are way to limited atm to force Intel into anything.

 

[trparky]

I've heard Intel may be going straight for PCIe 5.0 (whenever they'll manage to release something), but "forced to release"? As you have noted, the benefits are way too limited atm to force Intel into anything.

Yes, the benefits right now are limited but I can't help but to say... Don't underestimate market forces. As more PCIe 4.0 devices come out even though those devices may not actually fully take advantage of PCIe 4.0's additional bandwidth, the hype of PCIe 4.0 will, of course, ramp up and well... you get the idea.

 

[bug]

Yes, the benefits right now are limited but I can't help but to say... Don't underestimate market forces. As more PCIe 4.0 devices come out even though those devices may not actually fully take advantage of PCIe 4.0's additional bandwidth, the hype of PCIe 4.0 will, of course, ramp up and well... you get the idea.

You're making it sound as if Intel was resisting releasing PCIe 4.0. They simply don't have anything new to release, so they don't. Imagine how the public would react to yet another Skylake rehash, but with PCIe 4.0 on top and requiring $200+ mobos with a fan on the southbridge. Next year, if they get their crap together, they will most likely add support for something faster than PCIe 3.0, at least on their HEDT motherboards (if they keep HEDT around).

 

[trparky]

You're making it sound as if Intel was resisting releasing PCIe 4.0.

I don't think I'm saying that at all, what I am asking is if Intel, for the first time in a long time, will be forced to follow someone else instead of being the leader?

 

[bug]

I don't think I'm saying that at all, what I am asking is if Intel, for the first time in a long time, will be forced to follow someone else instead of being the leader?

That's a petty statement to make...

 

[trparky]

That's a petty statement to make...

Now that I think about it, yeah... it probably is. But it being petty doesn't at all take away from the question of whether or not market forces will force Intel into doing something that they initially didn't plan to do. I can't help but think how people who don't know any better will think "Well, AMD has this 4.0 and Intel still has 3.0. I don't know what it means but 4.0 has to be better than 3.0. Right?"

 

[TheLostSwede]

With the launch of several PCIe 4.0 devices including SSDs that are showing faster sequential read and write speeds (which really only benefits certain workloads), do you think that as more PCIe 4.0 devices come to market Intel will be forced to release PCIe 4.0 compatible hardware?

Forced? Well, that's a strange way of putting it. Intel will move to PCIe 4.0 for consumer platforms, as PCIe 5.0 will simply be too complex and too costly initially, to implement in consumer devices. I mean, people are complaining about the cost of PCIe 4.0 motherboards, what do you think they'll say if Intel comes out with new boards that require even more costly PCB designs and components? According to leaked roadmaps, Intel will apparently have PCIe 5.0 in the server space by next year.
As far as leaked information goes, Intel should have PCIe 4.0 by 2021 when it comes to their consumer platforms.

 

[Bill_Bright]

Forced? No. Enticed or encouraged? Maybe but I doubt it. Not if they can successfully move to PCIe 5.0 while ensuring backwards compatibility.

I do not agree it will be "too" complex and "too" costly. Yes, it will be expensive and difficult but I believe migrating to PCIe 4.0 then migrating again to PCIe 5.0 will cost a lot more strategically (in the long) run than it will cost to migrate directly to PCIe 5.0. And that's where Intel has the advantage - they have the deeper pockets to survive the long run expenses.

There are two wild cards yet to be totally revealed and/or tossed in the mix. The first is the general consumer. How long are we willing to wait for a stable platform and how much are we willing to spend? The other wild card is the 3rd party component makers. Are they willing to jump in the fray, and when? Nobody, not us, nor Intel, or any in the IT press know those answers - yet.

 

[trparky]

How long are we willing to wait for a stable platform

How far away is PCIe 5.0? For some people who have already maxed out PCIe 3.0 SSDs (people who do 4K video editing, for instance), that may be too long.

 

[TheLostSwede]

PCIe 5.0 is next year, on server boards. There are currently no plans to bring it to consumer tech.

Forced? No. Enticed or encouraged? Maybe but I doubt it. Not if they can successfully move to PCIe 5.0 while ensuring backwards compatibility.

I do not agree it will be "too" complex and "too" costly. Yes, it will be expensive and difficult but I believe migrating to PCIe 4.0 then migrating again to PCIe 5.0 will cost a lot more strategically (in the long) run than it will cost to migrate directly to PCIe 5.0. And that's where Intel has the advantage - they have the deeper pockets to survive the long run expenses.

There are two wild cards yet to be totally revealed and/or tossed in the mix. The first is the general consumer. How long are we willing to wait for a stable platform and how much are we willing to spend? The other wild card is the 3rd party component makers. Are they willing to jump in the fray, and when? Nobody, not us, nor Intel, or any in the IT press know those answers - yet.

You forget the fact that the PCIe trance length will be shortened again, to about half that of PCIe 4.0, unless a much more costly board design and more re-drivers are used. There's a point where it's not feasible to move to a new technology at a reasonable price point and PCIe 5.0 is that at the moment. As such, you won't see consumer PCIe 5.0 boards, regardless what you agree to or not.

On top of that, there is no sensible way for a consumer platform to take advantage of the benefits PCIe 5.0 brings over PCIe 4.0 for the next few years. So it's better to let the next "jump" mature a bit, let the costs come down and then move to PCIe 5.0 for consumer platforms in 3-5 years.

 

[TheoneandonlyMrK]

I don't think I'm saying that at all, what I am asking is if Intel, for the first time in a long time, will be forced to follow someone else instead of being the leader?

They have had pciex 4 compatible server parts and boards for a while, consumers don't get because Intel said so.

 

[Bill_Bright]

How far away is PCIe 5.0?

As TheLostSwede already noted, we should be seeing it as soon as 2020, at least in the server market. High-end consumer market will surely soon follow.

For some people who have already maxed out PCIe 3.0 SSDs (people who do 4K video editing, for instance), that may be too long.

Very few people have actually maxed out PCIe 3.0. Until the general population does consistently and they start demanding en masse, implementation will be slow, at best. When our operating systems, networking, office productivity and gaming software demands the greater bandwidth AND when the PCIe 3.0 bottleneck becomes the greatest bottleneck, then we will see a much greater, industry wide acceptance of 4.0/5.0.

You forget the fact that the PCIe trance length will be shortened again,

No, I didn't forget. I just think migrating entire industries (and corporate and individual consumers) through two generations of advancements will cost more in the long run than a single jump straight to 5.0. I might be wrong, but I don't think so. Like I said, nobody really knows at this point. We are all just speculating.

For sure, if we were talking another 8+ years between 4.0 and 5.0 (3.0 came out in 2010 - 2011), then no doubts we would move to 4.0 first. But as you noted, 5.0 will be here next year. And for that matter, 6.0 is not far behind that!

 

[trparky]

Very few people have actually maxed out PCIe 3.0.

Even with 4K video editing?

 

[TheLostSwede]

Like I said, nobody really knows at this point. We are all just speculating.

Not really. I know for a fact that Intel won't launch PCIe 5.0 for consumers in 2020 or 2021.
In fact, Intel will have PCIe 4.0 support in their CPUs in 2021 for the LGA1200 socket.
Q1 next year is Z490, the first LGA1200 socket and as we already know, Intel will launch a 10 core "consumer" part for that.
But I'm sorry to say, anyone hoping for PCIe 5.0 for consumer parts, is at least going to have to wait three years, most likely longer.

For sure, if we were talking another 8+ years between 4.0 and 5.0 (3.0 came out in 2010 - 2011), then no doubts we would move to 4.0 first. But as you noted, 5.0 will be here next year. And for that matter, 6.0 is not far behind that!

PCIe 4.0 has been available a couple of years in the server space already. This things launched at the end of July 2017 https://www.ibm.com/us-en/marketplace/power-systems-ac922/details This was roughly two months after PCIe 4.0 was announced as ratified. https://en.wikipedia.org/wiki/PCI_Express#PCI_Express_4.0
Using your logic, it took two years from there for PCIe 4.0 to end up in a consumer product. As such, the earliest we should expect PCIe 5.0 in consumer products is 2022, but my guess it'll be later than that and this is a guess.
PCIe 6.0 is not happening that quickly, as the standard is only expected be ratified in 2021, which means we might see some devices then, but so far, there has not been a single PCIe 5.0 device announcement, despite it now being two and a half months since PCIe 5.0 was ratified. I have at least not heard about anything and I can't seem to find a single press release announcing anything either.
IBM is likely to be the first company with PCIe 5.0, in its new Power10 server CPUs https://www.nextplatform.com/2018/03/24/openpower-at-the-inflection-point/

Even with 4K video editing?

4K is so 2014...
All the cool kids want 8K editing on the new Mac Pro now...

 

[Bill_Bright]

Even with 4K video editing?

Yes. In the grand scheme of things, very few people do such intensive 4K editing that they max out the PCIe 3.0 bus. If they are that serious - professionals for example - they would being using a different platform altogether.

I know for a fact that Intel...

Yeah right.

Have a good day.

 

[aQi]

With the launch of several PCIe 4.0 devices including SSDs that are showing faster sequential read and write speeds (which really only benefits certain workloads), do you think that as more PCIe 4.0 devices come to market Intel will be forced to release PCIe 4.0 compatible hardware?

They are cooking PCIe 5.0
That will enable all PCIe 4.0 supported devices to work on equal speeds on PCIe 5.0 as its backward compatible.

 

[EarthDog]

With the launch of several PCIe 4.0 devices including SSDs that are showing faster sequential read and write speeds (which really only benefits certain workloads), do you think that as more PCIe 4.0 devices come to market Intel will be forced to release PCIe 4.0 compatible hardware?

Yes. Why would we think otherwise?

I would assume the next chipset will have 4.0 support.

 

[TheLostSwede]

Yeah right.

Have a good day.

Give me one reason why I can't know this?
Or does that somehow intimidate you, so you just come up with a silly reply?

 

[aQi]

Yes. Why would we think otherwise?

I would assume the next chipset will have 4.0 support.

I double doubt on that. Intel might fast forward to PCIe 5.0

 

[EarthDog]

I double doubt on that. Intel might fast forward to PCIe 5.0

Nope. Not any time soon (on the consumer platform)...

 

[aQi]

Nope. Not any time soon (on the consumer platform)...

But i do doubt the adaption of PCIe 4.0 onto coming 10th gen. You see Intel is planning on putting the PCH under the IHS along with the processor die. But ending up with the same 14nm fabrication. Not to forget PCIe 4.0/5.0 has its own complications.
Im not saying its impossible but on 14nm with PCH along side. I mean how?

P.S. yeh man PCIe 5.0 is 2022 lol

 

[vega22]

Such a small % have nvme ssd near the pcie3 limit. Why would they bother about that when way more people are after the cores amd are offering, not the pcie bus?

 

[aQi]

Such a small % have nvme ssd near the pcie3 limit. Why would they bother about that when way more people are after the cores amd are offering, not the pcie bus?

What do you mean by that ?

 

[EarthDog]

Such a small % have nvme ssd near the pcie3 limit. Why would they bother about that when way more people are after the cores amd are offering, not the pcie bus?

It is odd when people say 'near the limit'. Many drives are 'at the limit' with the amount of bandwidth they can provide and are topped out in sequential reads. In working with large 4K and 8K files, aren't these large and mostly sequential?

I mean I get that most OS and other things, including drive snappiness have to do with random read 4k, but many of the performance PCIe 3.0 NVMe drives are banging on that upper limit of sequential reads so using 4.0 and drives which can perform faster, can be of benefit, in particular when able to utilize the bandwidth on large files. I also believe that 4k rando performance increases on the new drives. Not because of the bandwidth, but due to general improvements overall in the controller, NAND, etc.