News AMD Puts Hopes on Packaging, Memory on Logic, Optical Comms for Decade Ahead

[Kamen Rider Blade]

I used to like the idea of a front-accessible hot-swap backplane. But, then I thought about how often I ever swap drives and realized all I really care about is a mounting bay that's more accessible than the legacy kind that requires you to remove both side-panels and unfasten 2 screws on each side.

So, my server case is something like this:

Lian Li Announces the PC-M25 Chassis with a Plethora of Drive Mounts

Lian-Li Industrial Co. Ltd announces the PC-M25 chassis. This small form factor (SFF) chassis offers ample storage options and a hot-swap HDD cage for advanced storage applications. The PC-M25 includes a hot-swap HDD rack where users can conveniently install and remove up to five 3.5" drives...

www.techpowerup.com

It offers good airflow, a little bit of vibration-resistance, with the rubber gromets, and it's easy enough to slide drives in and out or change around their cabling.

I have something similar for a NAS.

Lian Li PC-Q26

But I'd rather have a Tower full of 5¼" drive bays so I can determine which Drive BackPlane to use.
Personally, I'm partial to ICY DOCK.
I've bought and used ALOT of their Drive BackPlanes and they work great.
Customer Support is pretty decent.

Apple doesn't make it easy for you to swap drives - that's for sure!

Then we can beat Apple at their game and make it stupid easy!

Vibration-resistance isn't something I've seen in a conventional, front-accessible hotswap setup. Also, my airflow is better, and I'll bet my case is cheaper than it would've been with such a backplane.

It would be cheaper, but I'm willing to pay more to get a better BackPlane.
That's why I use ICY DOCK.
Every single one of their BackPlanes meets your requirements.

Uh, no it's not.

Ok, you don't want to be a advocate for a easier experience?
I'll do it. Swapping drives should be as brain-dead easy as possible.
Even Tool-less.
Adding/Removing/Swapping Data Drives should be the easiest task possible.
That's made ALOT easier with Tool-less Drive BackPlanes.

 

[Ogotai]

bit user :

id still wouldnt want to take the chance with ebay and some used comp hardware


KRB :

looks like the u.2 cables are still as much as the one a local store has, not worth it IMO, never used the U.2 port on the board i have as i didnt need to, between the onboard sata ports, and the highpoint 16 port raid card i have already, i didnt think i would need it ;-)
considering this is on an X99 board, its quite possible a sas to 4 sata cable may not work if it is a pcie based U.2 port ?

 

[Kamen Rider Blade]

i have something similar to This but its made buy Vantec in 2 of my comps, easy-ish access to sata drives with out having to open my case, although the vantec ones i have, require 4 screws to attach the hdd to the tray....

I own several ICY DOCK FlexCage MB975SP-B R1
It's worth every penny.
3x 5¼" Drive Bays for 5x 3½" HDD slots.
All Tool-less

 

[Kamen Rider Blade]

bit user :

id still wouldnt want to take the chance with ebay and some used comp hardware


KRB :

looks like the u.2 cables are still as much as the one a local store has, not worth it IMO, never used the U.2 port on the board i have as i didnt need to, between the onboard sata ports, and the highpoint 16 port raid card i have already, i didnt think i would need it ;-)
considering this is on an X99 board, its quite possible a sas to 4 sata cable may not work if it is a pcie based U.2 port ?

If Intel & the MoBo manufacturers were following the proper specs, it should work.
Pray that Intel & your MoBo manufacturers did their "Due Diligience" to make the U.2 port work as intended.
U.2 is very common, it's pretty damn old by now too!

 

[Ogotai]

very similar to the vantec ones i have, 5 hdds in it as well

from what i have read, u.2 is pcie based, so connecting a min sas to sata cable may not work at all. as there is nothing there to convert the pcie signals to sata. the board is an asus x99-e 10g workstation board

 

[Kamen Rider Blade]

very similar to the vantec ones i have, 5 hdds in it as well

from what i have read, u.2 is pcie based, so connecting a min sas to sata cable may not work at all. as there is nothing there to convert the pcie signals to sata. the board is an asus x99-e 10g workstation board

If they don't have the basic HBA (Host Bus Adapter) chip there to convert the logic, then that's on either Intel or the MoBo maker.

That should be a basic feature for any end user since that is a VERY common scenario.

We're not asking for RAID support, just basic HBA functionality.

 

[Kamen Rider Blade]

It offers good airflow, a little bit of vibration-resistance, with the rubber gromets, and it's easy enough to slide drives in and out or change around their cabling.

Speaking as a customer of many Older Lian Li cases.

Those Lian Li Rubber Vibration dampener Gromets that they put around the screw to attach to your drive, my older sets are starting to deteriorate after all these years.
They leave a bit of black residue when I handle them and they're harder than their original state.

I wish they spent a bit more $ and used Quality Silicone Grommets instead of Rubber.

Silicone is supposed to last 4x longer than Rubber.

 

[InvalidError]

USB Type-E, is that what you want to call it?
Personally I don't see value in a fatter USB, I'd rather push PCI-SIG's reversible external OCuLink plug

I called it "PCIe type-E", not USB.

I am a big fan of integrated power+data. PCIe has power+data on a single connector, as do USB up to 240W, though almost nothing supports that at the moment. I'd be happy with 75W at 12V - standard PCIe slot power, enough to run 4xHDDs if necessary. I'd like to avoid having to run power and data on two separate cables to wherever internal devices get stuffed, so any internal consumer connector without power is a no-go in my book. All standards will need a 12VO refresh sooner or later to raise power distribution efficiency and meet future efficiency standards, may as well get it done sooner than later.

The drive-side end of U.2 is the same as SATA, with a whole bunch of extra pins on the connector's backside. That is the fat clunky part I was referring to. And the board-side connector doesn't provide power, so you need to have a clunky power cable splice for that or a backplane to run power to, which is problematic when practically no consumer cases are setup for that and most consumers won't pay $50 for it either - I like the idea, wouldn't pay for it though since I can go 5+ years without swapping drives in my PCs.

 

[Kamen Rider Blade]

I called it "PCIe type-E", not USB.

I see, well PCI-SIG already devised a External Plug, the way you wanted it, more or less.

I am a big fan of integrated power+data. PCIe has power+data on a single connector, as do USB up to 240W, though almost nothing supports that at the moment. I'd be happy with 75W at 12V - standard PCIe slot power, enough to run 4xHDDs if necessary. I'd like to avoid having to run power and data on two separate cables to wherever internal devices get stuffed, so any internal consumer connector without power is a no-go in my book. All standards will need a 12VO refresh sooner or later to raise power distribution efficiency and meet future efficiency standards, may as well get it done sooner than later.

External OCuLink is basically what you're asking for.

Spoiler: PCIe OCULink cable

Spoiler: 4x PCIe lanes in a Reversible connector ~ size of USB Type-A connector

We need them to get External OCuLink out ASAP and for the industry to start using it for external PCIe attached devices.

OCuLink connectors and cables support new PCIe standard

By Ed Cady, Contributing Editor OCuLink-2—Optical Copper (Cu) Link 2nd generation—is an interconnect system for inside and outside the box that supports

www.connectortips.com

If you think that PCIe x4 lane plug is too many lanes, should they follow the PCIe spec and create a smaller PCIe x2 lane & PCIe x1 lane plug that is even smaller than the External OCuLink plug?

The drive-side end of U.2 is the same as SATA, with a whole bunch of extra pins on the connector's backside. That is the fat clunky part I was referring to. And the board-side connector doesn't provide power, so you need to have a clunky power cable splice for that or a backplane to run power to, which is problematic when practically no consumer cases are setup for that and most consumers won't pay $50 for it either - I like the idea, wouldn't pay for it though since I can go 5+ years without swapping drives in my PCs.

That's just the nature of internal plugs and how they were devised since back in the day.
I'm used to it, so I just deal with it.

I guess I swap drives more often than most.

 

[InvalidError]

If you think that PCIe x4 lane plug is too many lanes, should they follow the PCIe spec and create a smaller PCIe x2 lane & PCIe x1 lane plug that is even smaller than the External OCuLink plug?

No need for different motherboard connectors, shouldn't be too hard to design a plug to accommodate 1/2/4-ways splitter cables.

We need them to get External OCuLink out ASAP and for the industry to start using it for external PCIe attached devices.

Not really: most bandwidth usage is highly asymmetrical and USB4.1/TB4-mode gives up to 120Gbps in one direction, which is almost on par with 4.0x8 and should be enough to run current GPUs. Since all but the lowest-end chipsets will have USB4.1 built-in within the next three years, practically all but the lowest-budget motherboards are going to have at least one of those ports within 3-5 years.

The only problem with converting all x4-or-slower interfaces to USB4.1 is the stupid amount of per-port overhead to support all of the alt-mode crap and USB-PD. That works for 1-2 ports, beyond that requires something leaner. OCu-Link may have made some sense 10 years ago but now that 3.3V and 5V are about to be deprecated, it makes no sense to push it into the consumer space now.

 

[InvalidError]

BTW, HWUB just posted a video on memory performance scaling. Both Ryzen 7k and Intel 13th-gen get upwards of 20% more performance going from the cheapest DDR5-4800 to decently fast 6000-30 for Ryzen and 7200 for Intel. So there would be at least this much to gain from having extremely fast on-package DRAM guaranteeing peak baseline performance no matter what external memory is used.

View: https://www.youtube.com/watch?v=XW2rubC5oCY

 

[InvalidError]

from what i have read, u.2 is pcie based, so connecting a min sas to sata cable may not work at all. as there is nothing there to convert the pcie signals to sata. the board is an asus x99-e 10g workstation board

Modern chipsets have multiple FlexIO/HSIO lanes that support some combination of PCIe, SATA and USB3.x/4.x. No additional logic would be required to 'convert' those from anything to anything else, just need to tell the chipset what protocol each lane is meant to speak if it cannot auto-detect. On many modern motherboards, especially on B-series one where fewer chipset PCIe lanes are available, the 3rd NVMe slot shares IO with SATA and you lose SATA ports (usually two) when using that NVMe slot. Same often goes for the PCIe x4 slot.

 

[Ogotai]

Invalid, keep in mind this is on an X99 based board, with an 8 core i7 or i9 cpu, cant remember which one at the moment, and i think due to the PLX switch chips, the chances of something being disabled cause i used something else IE nvme slot vs sata, might not happen that easily

 

[InvalidError]

Invalid, keep in mind this is on an X99 based board, with an 8 core i7 or i9 cpu, cant remember which one at the moment, and i think due to the PLX switch chips, the chances of something being disabled cause i used something else IE nvme slot vs sata, might not happen that easily

LGA2011 CPUs have up to 40 PCIe lanes directly from the CPU so most x99 boards don't have a PLX chip on them. At a glance, the x99-e 10G doesn't appear to have one either. From the "up to 10 SATA, up to 8 PCIe, up to 6 USB3" verbiage on the chipset marketing slides, it looks like x99 may have been one of the earlier chipsets with flexible HSIO lanes.

The one NVMe slot on that board is 3.0x4, which would have to come from the CPU and wouldn't support SATA or USB though.

 

[Ogotai]

actually i think it does, hwinfo i think detects them and shows the pcie lanes being split to other things, and there is this :

Parts

Puget Systems provides part information, advice, and tech specs for all of the computer hardware we use to configure our workstation PCs

www.pugetsystems.com

https://www.anandtech.com/show/8781...otherboard-review-dual-10gbaset-for-prosumers <--- this is for the Asrock version with the same name as the Asus i have.
https://www.asus.com/ca-en/motherboards-components/motherboards/workstation/x99-e-10g-ws/ asus' site for the board, kinda surprised its still there.
https://www.tweaktown.com/reviews/7925/asus-x99-10g-ws-workstation-motherboard-review/index.html
https://www.anandtech.com/Show/Inde...al-10gbaset-ethernet-via-an-integrated-x550t2 <--- found the anandtech review of the board

but i still have no idea about the u.2 port i dont think the cables i have will work with it, as they are mini sas to sata

 

[InvalidError]

actually i think it does, hwinfo i think detects them and shows the pcie lanes being split to other things, and there is this :

Hm, can't find any mention of it in Asus' marketing fluff. What few times I remember looking at a board using a PLX PCIe switch, the manufacturer made sure you knew it was in there.

 

[Ogotai]

for some reason asus, doesnt mention it * shrug *

too bad there isnt such a board, while using plx chips, in X5/670 or Z790 , i could use a board with more pcie lanes that isnt a TR board.