PCIe 4.0 wont be supported on B450, X470 and older chipsets

[Trixanity]

While not completely related, I figured I'd just post it here instead of making a new one or bumping one of the older threads.

A diagram of the X570 chipset just leaked:

Spoiler

Looks a bit similar to X470 but I haven't had the time to compare.

[Dabombinable]

This sounds a lot like lga1155, with the change from gen 2.0 to 3.0

[porina]

5 minutes ago, Trixanity said:

A diagram of the X570 chipset just leaked:

Looks a bit similar to X470 but I haven't had the time to compare.

Not surprising but upgraded link to chipset, and upgraded lanes from chipset... no change on either of the M.2s so does that mean it is essentially still at 3.0?

[Dabombinable]

Just now, porina said:

Not surprising but upgraded link to chipset, and upgraded lanes from chipset... no change on either of the M.2s so does that mean it is essentially still at 3.0?

Well the available bandwidth is doubled overall.....so I'd say no.

[porina]

6 minutes ago, Dabombinable said:

Well the available bandwidth is doubled overall.....so I'd say no.

Where's the increased bandwidth? In case of any confusion, I was talking specifically about M.2. Both the M.2 sockets (one CPU, one chipset) are indicated in that chart as 32GB/s. That's the same as 3.0 x4 hence my question, are the M.2 running at 3.0? The alternative is 4.0 x2 but that makes no sense for most existing devices.

[mariushm]

I see the extra m.2 from chipset in the diagram, and wi-fi/bt stuff .. not sure if it's wireless integrated in chipset or just a link (pci-e/usb/whatever) to wireless chip.

 

Don't get what consumes so much power that makes chipset need active cooling.

Not sure what process they use, i suspect it's 28nm or higher... I think the previous chipsets were 45-55nm made at TSMC.

Probably could have been passively cooled if done at 12-14nm but then chipsets would cost too much

 

[Trixanity]

36 minutes ago, mariushm said:

I see the extra m.2 from chipset in the diagram, and wi-fi/bt stuff .. not sure if it's wireless integrated in chipset or just a link (pci-e/usb/whatever) to wireless chip.

 

Don't get what consumes so much power that makes chipset need active cooling.

Not sure what process they use, i suspect it's 28nm or higher... I think the previous chipsets were 45-55nm made at TSMC.

Probably could have been passively cooled if done at 12-14nm but then chipsets would cost too much

 

Multiple NVMe drives will allegedly cause the power to spike if you do simultaneous writes. At least that's what I assume from the talks on the subject. Doesn't seem too farfetched to me but it also seems like a problem 99% of people won't experience.

[porina]

43 minutes ago, mariushm said:

Don't get what consumes so much power that makes chipset need active cooling.

Assuming all else is equal apart from the bandwidth, then more data transfer = more power used. I don't know if 4.0 has any features to save power per bit compared to 3.0... if it is just same but faster, then it'll be hotter.

[Dabombinable]

7 hours ago, porina said:

Where's the increased bandwidth? In case of any confusion, I was talking specifically about M.2. Both the M.2 sockets (one CPU, one chipset) are indicated in that chart as 32GB/s. That's the same as 3.0 x4 hence my question, are the M.2 running at 3.0? The alternative is 4.0 x2 but that makes no sense for most existing devices.

When the PCIe 4x slot and x1 slots are Gen 4.0....then the m.2 will be as well.

[porina]

1 minute ago, Dabombinable said:

When the PCIe 4x slot and x1 slots are Gen 4.0....then the m.2 will be as well.

There is no indication of that. The chart shows 32GB/s which is equivalent to 3.0 x4 and that's the only scenario that makes sense. If it was 4.0 x4 why not 64GB/s?

[Dabombinable]

3 minutes ago, porina said:

There is no indication of that. The chart shows 32GB/s which is equivalent to 3.0 x4 and that's the only scenario that makes sense. If it was 4.0 x4 why not 64GB/s?

 

7 hours ago, Trixanity said:

While not completely related, I figured I'd just post it here instead of making a new one or bumping one of the older threads.

A diagram of the X570 chipset just leaked:

  Reveal hidden contents

Looks a bit similar to X470 but I haven't had the time to compare.

 

[porina]

@Dabombinable Ok, I messed up the units, should have written 32Gbps not 32GB/s which is a factor of 8 wrong. That aside, my point remains, 32Gbps is most logically 3.0 x4. It should say 64Gbps if it were 4.0 x4.

[Dabombinable]

24 minutes ago, porina said:

@Dabombinable Ok, I messed up the units, should have written 32Gbps not 32GB/s which is a factor of 8 wrong. That aside, my point remains, 32Gbps is most logically 3.0 x4. It should say 64Gbps if it were 4.0 x4.

Either way, its 4GB/sec for each of the m.2 connectors (far better than what we currently have). With 1 taking up the lanes from the x4 slot. The main thing is that the link between the CPU and PCU has double the bandwidth, so we'd potentially be seeing better NVME SSD speeds.

[Drak3]

2 minutes ago, Dabombinable said:

ts 4GB/sec for each of the m.2 connectors (far better than what we currently have).

We currently have 32GB/s on M.2 connectors. Some of us are running boards from 2016 that have it.

[Trixanity]

48 minutes ago, Drak3 said:

We currently have 32GB/s on M.2 connectors. Some of us are running boards from 2016 that have it.

You're confusing bits and bytes. You're describing x16 4.0 and I don't think you have that.

[Drak3]

1 minute ago, Trixanity said:

You're confusing bits and bytes. You're describing x16 4.0 and I don't think you have that.

No, I was typing that whilst in the middle of updating back to Corsair iCUE and didn't proofread.

[Trixanity]

17 minutes ago, Drak3 said:

No, I was typing that whilst in the middle of updating back to Corsair iCUE and didn't proofread.

X470 is quite limited in the drive configurations though which I think was what was referenced.

[porina]

1 hour ago, Trixanity said:

You're confusing bits and bytes. You're describing x16 4.0 and I don't think you have that.

It's the same mistake I made. Bits, bytes...

[Daniel Z.]

I think the amount of people that will be affected by this is basically 0. I don’t know of a single GPU that can fully saturate a PCIe 3.0 16x connection. I’m pretty sure even 4x is enough.

[Stefan Payne]

Well, was to be expected.

 

It would look totally weird if cheap or ITX Boards did support PCIe 4.0 but the highest end, 250€ Board doesn't due to the Switchers...

[PlayStation 2]

It's probably for the better, anyways: AM4 has enough PCI-E issues to begin with.

[ApolloFury]

15 minutes ago, Daniel Z. said:

I think the amount of people that will be affected by this is basically 0. I don’t know of a single GPU that can fully saturate a PCIe 3.0 16x connection. I’m pretty sure even 4x is enough.

Faster storage? Multiple drives in one PCI-E slot?

[Guest]

I thought this was pretty expected.

[XR6]

I'm not surprised this happened, it's to be expected. In fact, I would be more surprised if PCIe 4.0 wasn't disabled.

[Jurrunio]

I hope Asrock doesn't get the memo as usual