Discussion in 'Article Discussion' started by bit-tech, 19 Jun 2019.
5x2 = 8...? Really...?
My bad: it's 'cos I'd originally started writing it in megabytes per second (which was a very-almost-doubling from 500MB/s to 985MB/s between PCIe 2.0 and 3.0) but switched to gigatransfers per second. Fixed now!
Don't pick on @Gareth Halfacree he's obviously just been doing his accounts for the taxman.
Joking aside, I read the article 'literally' and it didn't make sense
USB-IF could learn some lessons from this...
So something i don't get and I'm probably going to sound stupid for asking, but from what i gather implementing PCIe 4.0 on X570 has been more involved due to difficulties in maintaining signal integrity between the CPU and first PCIe slot, isn't that only going to get worse with PCIe 5.0/6.0?
Or is that what the four-level pulse amplitude modulation, that i don't have the first clue about, intended to address.
If what AsRock told GamersNexus is accurate, then yes, implementing PCIe 4 is a pain in the lower back (and also a major factor in the much higher price of X570 boards), so yep logic dicateates that implementing PCIe 5 & 6 will be bloody hard too.
No clue though how much stuff like four-level pulse amplitude modulation can offset that difficulty.
PCIe 4 doubled signal rate compared to PCIe 3.0 (implementation headaches), and PCIe 5.0 doubles signal rate again (more implementation headaches). PCIe 6.0 does not, it avoids signal rate increases by changing signal levelling. That's somewhat less of a headache as it does not change the traces, 'just' the PHY layers in the chips.
Sounds like a good reason to just skip PCIe 5 then
Depends on who you are: if you're making a motherboard, implementing PCIe 5.0 means you're already ready for PCIe 6.0. If you're a chipmaker (at either end of the bus), implementing PCIe 6.0 means you have to deal with both the PCI2 5.0 signalling rate increase and the switch to 4-level PAM.
Is it just my imagination or are we going backwards now by adding faster links that draw more power and are harder to implement, yet almost nobody uses.
I was about to mock the pointlessness of USB 3.2 10/20gbit when it occurred to me that M.2 NVMe drive prices have tanked and external enclosures are a thing. Dammit, there goes another $97 (58 quid) on Amazon; up to 900MB/s on a thumb(ish) drive should be fun though.
Separate names with a comma.