PCIe 5.0 Going to 32GT/s - Spec Available to Members
Click here to post a comment for PCIe 5.0 Going to 32GT/s - Spec Available to Members on our message forum
PrMinisterGR
Any idea when we'll start seeing PCIe 4.0 motherboards? I don't believe that Intel is introducing them with CoffeeLake and I haven't heard anything from AMD about Zen 2 and them either.
schmidtbag
Ricepudding
schmidtbag
http://amfeltec.com/products/flexible-x4-pci-express-4-way-splitter/
where if you don't need all the extra bandwidth for a single card, you can take advantage of more expansion cards instead. I'd be really interested to see if there's a splitter out there that could, for example, convert a single gen 3.0 16x slot into two 2.0 16x slots.
From what I recall, I think I remember seeing a benchmark of the 1080Ti on various PCIe slots, and you lose a few FPS on gen 2 on some tests. It's only maybe a 2% performance loss though, so I'd say gen 2 has held on real strong.
I know there are some SSDs out there that can saturate the bandwidth much easier.
To me, the real interest of newer PCIe generations is the performance for the 1x and M.2 slots. We're approaching times where it should be possible to comfortably do 1080p gaming on a 1x slot. There are also products like this:
Fox2232
Ricepudding
Aura89
http://www.guru3d.com/articles-pages/pci-express-scaling-game-performance-analysis-review,2.html
http://www.guru3d.com/index.php?ct=articles&action=file&id=14856&admin=0a8fcaad6b03da6a6895d1ada2e171002a287bc1
http://www.guru3d.com/index.php?ct=articles&action=file&id=14861&admin=0a8fcaad6b03da6a6895d1ada2e171002a287bc1
I understand this isn't using a 1080ti, and the difference could be larger (though i don't think it would be) but if someone was gaming on a PCI-Express 1 16x with a 1080ti, they would likely get less performance then on 2.0, and 3.0 to a lesser degree, but it would be completely and totally playable.
In reality, Gen 1 holds up fairly well as well. Bandwidth just isn't so much of an issue when it comes to graphics cards it seems. Not saying more isn't welcomed, just isn't the most important thing, by far.
nick0323
You recently done an article reviewing PCI-E bandwidth performance. My motherboard is PCI-E 3.0 ready but my CPU is holding me back on PCI-E 2.0. I see no reason to upgrade my CPU to unlock the PCI-E bandwidth thus far.
I'm looking forward to an updated review when PCI-E 4 debuts.
Exascale
Venix
we do not have pci-x 4 yet and they are talking about 5 ? i would have been more excited to see news like that for sata 4 ! sata 3 we saturated that one since 2012 or something no ? now though throw 1 pci-x 5 lanes on sata name it sata 5 with 3.9gb max throughput and then let's watch all the ssd companies battling it out to max it first!
although and 1xpci-x 4 lane on sata 4 with 2gb throughput will be also great :P
Emille
The above benchmark shows around a 10% performance drop at max speed. Imagine then an overclocked 1080ti...what would it be, 25% as you are greatly exceeding the bandwidth and then if you have a typical z170x motherboard and ypu are running sli, half bandwidth per card clearly faster pcid lanes are needed right now yet we haven't heard anything about when pci 4.0 will be implemented.
M.2 drives are already fast as hell and I imagine by 2 time they will be twice as fast as now
The samsung 950pro was 25ppmv/s. The 960 pro is 3500mb/s.
If 4.0 is twice as fast and the drives are at the point of saturation of bandwidth. You could have a board with 6 m.2 slots all running at pci-e 4.0 2x speed which would be the same as 3.0 4x.
For my bext build. Not only do I want to have no mechanical drives. I want to have no sata drives at all.
It is an archaic interface and there is no point to it this day in age. Hopefully faster pci-e will get us there.
schmidtbag
Emille
schmidtbag
Emille
Exascale
Both NVMe over PCI-e and SATA NAND are not fast enough for some use cases, which is why people are eagerly awaiting nonvolatile memory that is significantly faster than NAND like XPoint PCM, which is roughly 10x the performance of NVMe NAND when its used in an NVMe PCI-e drive itself. It should be able to perform roughly 100x faster than an NVMe NAND SSD when its used in its upcoming DIMM format.
The whole "we dont need anything faster" has never been true with computers. This is especially true of memory and storage, which are two of the biggest bottlenecks in most systems. That's also why youre seeing things like Nantero's NRAM and Everspin's STT-MRAM being developed.
Also, arguing that SATA is fast enough based on some hypothetical user not being able to notice a difference isn't helpful. Start using the drives in a way that is really taxing on the I/O subsystem and you'll probably see that the sustained performance and latency of the drive with the faster interface is significantly better.
Windows 8 and newer is also pretty good at SSD maintenence actually. The real issue would be with some low end consumer level drives having little or no overprovisioning, so they slow down as they fill.
schmidtbag
Aura89
D3M1G0D