Wednesday, June 22nd 2016

AMD "Zen" Processor Integrated Chipset Has USB 3.1 Issues, Could Escalate Costs

With its next-generation processors and APUs based on the "Zen" micro-architecture, AMD is integrating the chipset into the processor/APU die, making motherboards entirely chipset-free. This on-die chipset, however, is rumored to be facing issues with its integrated USB 3.1 controllers, according to industry sources. AMD sourced the design for the integrated USB 3.1 controllers from ASMedia. The company has a tendency of sourcing integrated controller IP from third-party manufacturers (eg: its SATA controllers and port-multipliers in the past have been sourced from Silicon Image).

Motherboard manufacturers are noticing significant drops in USB 3.1 bandwidths with increase in circuit distances (think wiring running from the AM4 socket to USB 3.1 front-panel headers on the bottom-right corner of a motherboard). Board designers are reportedly having to use additional retimer and redriver chips to get acceptable bandwidths over such ports, and in some cases even entire USB 3.1 controllers, eating into the platform's PCIe budget and escalating costs.
AMD stated that the "Zen" processor platform is on-track, and the company will not comment on "customer specific board-level solutions." ASMedia played down the reports as industry rumors, and stated that its solutions passed the necessary certifications. The DigiTimes report adds that "AMD's Zen processors are seeing stable development and have a satisfactory yield rate, and about to enter the engineering sample stage in the near future."
Source: DigiTimes
Add your own comment

83 Comments on AMD "Zen" Processor Integrated Chipset Has USB 3.1 Issues, Could Escalate Costs

#26
vega22
btarunrAMD's own XConnect external graphics card standard.

I'm sure this isn't just about desktops. The silicon is usually common between desktops and notebooks (in different packages).

Imagine you have a Zen-powered notebook and your external graphics box isn't even getting 10 Gbps.
i dont think it ever will with just usb3, i thought they needed thunderbolt?
Posted on Reply
#27
PP Mguire
vega22i dont think it ever will with just usb3, i thought they needed thunderbolt?
Yea their page says Thunderbolt 3.
Posted on Reply
#28
qubit
Overclocked quantum bit
This is a real facepalm of a problem if true. Come on AMD.
Posted on Reply
#29
bug
FrickAt this moment there are several 11 inchers for sale without USB 3.0. Chromebooks and netbooks from 2015. Low specced laptops from 2014 still in stock. And what do people buy? Cheap computers, expensive phones and tablets.
Idk about Chromebooks, nobody buys them outside US (see www.statista.com/statistics/430473/chromebook-unit-shipments-region/). A quick glance at Asus and Acer reveals exactly one Chromebook model that doesn't come with USB2.0. And that's powered by a Rockchip ARM CPU.
Posted on Reply
#30
Frick
Fishfaced Nincompoop
bugIdk about Chromebooks, nobody buys them outside US (see www.statista.com/statistics/430473/chromebook-unit-shipments-region/). A quick glance at Asus and Acer reveals exactly one Chromebook model that doesn't come with USB2.0. And that's powered by a Rockchip ARM CPU.
Not counting Rockchip systems I can find 8 models without USB 3.0 where I live, all available for purchase today at about €200. Some of them at big retailers.
Posted on Reply
#31
iO
TheLostSwedeAnyone remember this?
www.techpowerup.com/223163/amd-socket-am4-platform-demo-motherboard-pictured

Maybe it's not integrated at all, maybe it's that shiny metal AMD branded chip by the rear ports that houses the USB 3.1 controller?
Yea thats their Promontory "chipset" which should be the problem if true, as Zen's integrated SB seems to only support USB3.0.

Should be easy to fix with a new revision, like Intel's Z68 B3 chipset back then.
Edit: At least they found the problem in time unlike Intel...
Posted on Reply
#32
cdawall
where the hell are my stars
The way the responses in this article read

"USB 3.1 isn't perfect? Zen must be crap then"

Holy hell just wait for the release everyone has hardware bugs on new products Intel is no stranger nor is nvidia.
Posted on Reply
#33
bug
cdawallThe way the responses in this article read

"USB 3.1 isn't perfect? Zen must be crap then"

Holy hell just wait for the release everyone has hardware bugs on new products Intel is no stranger nor is nvidia.
Well, if there's a (potential) problem with the USB controller and the USB controller is integrated in the CPU, I think I want fair warning.
Posted on Reply
#34
PP Mguire
cdawallThe way the responses in this article read

"USB 3.1 isn't perfect? Zen must be crap then"

Holy hell just wait for the release everyone has hardware bugs on new products Intel is no stranger nor is nvidia.
I read it this way

"We may have heard this, so let's make it a story for clickbait".
Posted on Reply
#35
cdawall
where the hell are my stars
bugWell, if there's a (potential) problem with the USB controller and the USB controller is integrated in the CPU, I think I want fair warning.
Let's get this out there now. No cpu is perfect. There is nothing physically wrong with the controller it's only issue is long distances put the connections near the cpu and problem solved. Current cases don't have front panel USB 3.1 so that shits a mute point.
Posted on Reply
#36
PP Mguire
cdawallLet's get this out there now. No cpu is perfect. There is nothing physically wrong with the controller it's only issue is long distances put the connections near the cpu and problem solved. Current cases don't have front panel USB 3.1 so that shits a mute point.
Might want to refer to my previous point. Without them specifying it could mean 3.1 in general, which would be both Gen 1 and 2. That would include front panel.
Posted on Reply
#37
eidairaman1
The Exiled Airman
cdawallLet's get this out there now. No cpu is perfect. There is nothing physically wrong with the controller it's only issue is long distances put the connections near the cpu and problem solved. Current cases don't have front panel USB 3.1 so that shits a mute point.
3.1 should use the exact same port as 3.0 imho
Posted on Reply
#38
ZoneDymo
bugBusiness oriented chipsets tend to offer additional management capabilities, gaming/enthusiast chipsets come with additional PCIe lanes. All the options will be gone.
again im not an expert in this area at all, but why would this AMD cpu not offer those functions as well?
Like why would they not make it technically capable for steering on lets say 10 PCEe Lanes and then let the motherboard makers decide how many to actually physically put on the board?
And do we actually know if these "additional management capabilities are also not just part of the package but perhaps something you must activate in the bios or something?
Posted on Reply
#39
john_
HoodIf Zen actually turns out to be competitive with Intel's offerings, the last thing AMD needs is USB 3.1 problems - port speeds have long been sub-par on AMD's chipsets, Zen is their chance to achieve parity on this front, with robust support for USB 3.0 & 3.1, PCIe 3.0, NVMe, M.2, U.2, DDR4, and Thunderbolt 3. Also they should dump support for legacy ports, as Intel has - concentrate on moving forward instead of trying so hard to be "the poor man's option" - people who want those old ports can always buy sub-$50 boards that still have USB 2.0, PCIe 2.0, DDR3, Firewire, serial, parallel, eSATA, PCI, IDE, and probably floppy disk for all I know, as add-on chips. Give the other people (who aren't stuck in the '90s) a way to enjoy modern connectivity, without automatically having to choose Intel. AMD has never even been an option for me, mostly because of their lack of support for the latest port speeds (power-hungry, hot-running, and slow also lacks appeal). So get it right, AMD, or continue to be the "poor man's option", with a rapidly dwindling customer base. Zen might be your last chance to avoid bankruptcy...
10 years ago we where counting MB/sec, because they looked and they where important. A USB 2.0 controller scoring 35MB/sec, compared to one that was hitting a wall at 28-30MB/sec, it was something to take in consideration when choosing a platform. Today we are talking about hundreds and thousands of MB/sec bandwidth that hardware some times can't even take advantage. Of course there are going to be costumers out there that will be looking at USB/SATA/whatever speeds from the chipset. Of course the faster option is always better option, if price and other features are almost the same. But in reality, how many can feel the difference between a SATA III SSD at 490MB/sec and a SATA II SSD at 550MB/sec, for example?

Anyway, if someone wants to avoid "the poor man's option", he will find many reasons. On the other hand I remember reading about "the rich man's option" motherboards that where super loaded with features that you could not use because there weren't enough PCI express lanes.
Posted on Reply
#40
bug
ZoneDymoagain im not an expert in this area at all, but why would this AMD cpu not offer those functions as well?
Like why would they not make it technically capable for steering on lets say 10 PCEe Lanes and then let the motherboard makers decide how many to actually physically put on the board?
And do we actually know if these "additional management capabilities are also not just part of the package but perhaps something you must activate in the bios or something?
AMD may offer all the above, but they're taking away the option to pay for what you need since instead of several chipset, now there's only one.
As a games I don't care about remote management and as an office user I don't want to pay for 20 extra PCIe lanes that will go unused.
That's all.
Posted on Reply
#41
bug
cdawallLet's get this out there now. No cpu is perfect. There is nothing physically wrong with the controller it's only issue is long distances put the connections near the cpu and problem solved. Current cases don't have front panel USB 3.1 so that shits a mute point.
Maybe so.
The thing is, when you move something onto the CPU, the end user expects it will work pretty much the same, regardless of the motherboard choice. If in the end everyone decides to fix USB by adding those "additional retimers and redrivers", I still need to know to look after the quality of the implementation.

Personally, I don't care much as Zen will not be good enough for me to ditch my 6600k. It's just that I don't agree with posters saying this is business as usual and shouldn't be given any attention.

Also, for what we know, there is something physically wrong with the controller, because no other controller has been reported to have such issues. What ASMedia may be saying is that they delivered within the parameters specified by AMD. But we really don't know, so we should just watch for further developments.

Edit: Also, while my ancient case does not have a USB 3.1 panel, my brand new motherboard came with one. One Type-A and one Type-C no less. Though it did cost an arm a leg.
Posted on Reply
#42
Steevo
My .02 cents is the issue is with On Die Termination and the node size, and or crosstalk and the ability to keep signal integrity high enough with low voltage to the core itself without potentially killing something. Same issues people faced when they realized a few years ago that memory voltages were killing Intel chips. Lower process sizes means issues that they have never faced before are now a reality, and how many other controllers are built on 14nm Finfet that currently work and are NOT part of the primary silicon itself.
Posted on Reply
#43
AsRock
TPU addict
cdawallLet's get this out there now. No cpu is perfect. There is nothing physically wrong with the controller it's only issue is long distances put the connections near the cpu and problem solved. Current cases don't have front panel USB 3.1 so that shits a mute point.
Not really, kinda depends on what long is, and i use all my USB onboard headers and would hate to see more cables having to go over every thing else.

Lets face it ASMedia never been perfect anyways, like the SATA 3 running a max of 350Mb\s is pretty common.
Posted on Reply
#44
spacejammer
Great job, btarunr! Not only did you copy the original rumor nearly word-for-word, you also failed to mention the fact that it would cost motherboard manufacturers $5 at most to get around the issue, as well as the fact that this has nothing to do with the integrated chipset on the CPU itself, which doesn't even support USB 3.1 by itself. I guess this is what happens when you let people report on things they have no clue about.

This is disappointing to say the least, even for a rumor with no actual source.
Posted on Reply
#45
BiggieShady
SteevoMy .02 cents is the issue is with On Die Termination and the node size, and or crosstalk and the ability to keep signal integrity high enough with low voltage to the core itself without potentially killing something. Same issues people faced when they realized a few years ago that memory voltages were killing Intel chips. Lower process sizes means issues that they have never faced before are now a reality, and how many other controllers are built on 14nm Finfet that currently work and are NOT part of the primary silicon itself.
Good point. The specs AMD gave to ASMedia are made according to their 14nm Finfet zen silicon voltage range, and if that's not enough to ensure singnal strength for long routes on the motherboard, then certainly additional retimer and redriver chips are way to go, rather than reducing the lifetime of the cpu in order to produce a mobo more cheaply ... and I bet Asus is crying about expensive R&D they have to do in short time rather than few bucks increased production cost per board. It's much harder to return R&D investment.
Posted on Reply
#46
looncraz
This is simply a USB 3.1 problem - NOT LIMITED TO AMD.

HP just placed the chipset behind the USB ports and called it a day, others may follow suit.

Intel's USB3.1 solutions **HAVE THE SAME PROBLEM**

You just can't carry an 850MHz signal on motherboard traces for very far without picking up noise, so you have to boost (repeat) them to make longer distances. If the chipset is located too far away from the USB ports, then the performance will suffer or you need to use repeaters (which add a $2~5 to the cost).

Simple.
Posted on Reply
#47
TheGuruStud
At least it will be fixed...thanks for nothing Intel, assholes. Intel's issues go back for years and years.
Posted on Reply
#48
looncraz
AsRockNot really, kinda depends on what long is, and i use all my USB onboard headers and would hate to see more cables having to go over every thing else.

Lets face it ASMedia never been perfect anyways, like the SATA 3 running a max of 350Mb\s is pretty common.
ASMedia has often been handicapped by their products being second class citizens on the platforms. This will be first time we see ASMedia IP as first class citizens, with full access to all the bandwidth it wants. Kind of hard to reliably push even 400MB/s+ over an interface which maxes out at 500MB/s (PCI-e 2.0 x1). In the very least, the new connections should have PCI-e 3.0 x1, with a peak of 985MB/s.
Posted on Reply
#49
Yukikaze
Pardon my snarky remark, but:
In this thread, what I am mostly seeing is people who have no idea how chip design is done, bashing on a company for reasons they do not understand.

That said, this is a problem, but it isn't a horrible one.
Posted on Reply
#50
AsRock
TPU addict
looncrazASMedia has often been handicapped by their products being second class citizens on the platforms. This will be first time we see ASMedia IP as first class citizens, with full access to all the bandwidth it wants. Kind of hard to reliably push even 400MB/s+ over an interface which maxes out at 500MB/s (PCI-e 2.0 x1). In the very least, the new connections should have PCI-e 3.0 x1, with a peak of 985MB/s.
All my SATA 3 connected drives hit way over 400MB\s, tbh it's a none problem really ( for me fine for my needs ) but hopefully it will perform more like the Intel SATA3 at least this time.
Posted on Reply
Add your own comment
Apr 25th, 2024 04:50 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts