Early Haswell processors will reportedly have a USB 3.0 bug

[Matthew DeCarlo]

If you're planning on upgrading to Intel's Haswell platform later this year, you may want to wait for the first round of chips to clear shelves before taking the plunge. Hardware.info claims to have spoken with a "reliable source" who says the next-generation processors have a small issue with USB 3.0.

The site viewed a document that Intel reportedly sent out to inform system builders about a bug with Haswell when waking from sleep and using a USB 3.0 device. Apparently, if a system exits S3 level sleep while a user is accessing data on a USB 3.0 drive, their session with that data will essentially end.

For instance, Hardware.info suggests that if you left a PDF document open before putting your PC to sleep, you might be left with blank pages when it wakes up, forcing you to restart the application. Likewise, resuming a video where you left off might require reloading the file or restarting the player.

Although it wouldn't comment on the rumor, Intel is reportedly aware of the problem and it's making partners accept the flaw before purchasing affected processors. Intel is said to be working on a fix that should appear in a later CPU stepping, but it sees the bug as more of an annoyance than anything.

For whatever it's worth, there's supposedly no risk of data loss and it doesn't seem like this will turn out to be repeat of Intel's early 6 series "Cougar Point" chipsets, which launched alongside Sandy Bridge in 2011 and were recalled due to a flaw that caused 3Gb/s SATA support to degrade and fail over time.

Permalink to story.

https://www.techspot.com/news/51826-early-haswell-processors-will-reportedly-have-a-usb-30-bug.html

 

[St1ckM4n]

An annoyance? I think it's more than that, Intel...

 

[misor]

*conspiracy mode:
intel "leaked" the rumors so that more current sandy/ivy products still on store shelves will be sold before the arrival of next gen processors.

I don't really think intel will **** on their milking cows, early intel adapters/fanboys, for the second time.

 

[Darth Shiv]

There's probably a laundry list of issues they de-prioritised for hitting the launch date. Interesting to see what else they don't think is important...

 

[hahahanoobs]

Interesting to see what else they don't think is important...

Every CPU is released with errata. Just be happy that because of your love of tech, you're in the know on day one. No biggie, just wait for the fix in the next revision. Assuming this issue is real/will hit retail. It's certainly nothing worth whining about.

 

[Darth Shiv]

Every CPU is released with errata. Just be happy that because of your love of tech, you're in the know on day one. No biggie, just wait for the fix in the next revision. Assuming this issue is real/will hit retail. It's certainly nothing worth whining about.

I'm not whining about the one they tell us about!

 

[dividebyzero]

I'm not whining about the one they tell us about!

Intel actually tell everyone about every CPU they make (as comparison, try to find the errata for an AMD CPU). Whether or not people can be arsed to actually read the specification update is more the question.
Case in point for example:
Ivy Bridge CPU family (3rd gen Core)> click "specification update" > Page 5 revision updates, Pages 9-12 erratum per stepping, Pages 17-44 erratum in detail, workaround if available, implication of errata.

 

[Darth Shiv]

Intel actually tell everyone about every CPU they make (as comparison, try to find the errata for an AMD CPU). Whether or not people can be arsed to actually read the specification update is more the question.
Case in point for example:
Ivy Bridge CPU family (3rd gen Core)> click "specification update" > Page 5 revision updates, Pages 9-12 erratum per stepping, Pages 17-44 erratum in detail, workaround if available, implication of errata.

They didn't document things like "we put in out of spec SATA power saving features in our mobile chipsets that destabilises SATA 3 SSDs and makes SandForce 2 and any other fast SATA 3 SSD device look faulty" though, or "our graphics component of the chipset, whilst we claim supports hardware decoding, doesn't - we didn't test it" for HD2000... who knows what else is in the current generation. Pretty easy to setup unit tests for the capacities of the SATA and USB buses when you can just plug in RAM drives or specialised COMs devices from programmable ROMs.

Plenty of "I want a refund" features not in the errata!

 

[dividebyzero]

They didn't document things like "we put in out of spec SATA power saving features in our mobile chipsets that destabilises SATA 3 SSDs and makes SandForce 2 and any other fast SATA 3 SSD device look faulty" though, or "our graphics component of the chipset, whilst we claim supports hardware decoding, doesn't - we didn't test it" for HD2000... who knows what else is in the current generation. Pretty easy to setup unit tests for the capacities of the SATA and USB buses when you can just plug in RAM drives or specialised COMs devices from programmable ROMs.
Plenty of "I want a refund" features not in the errata!

Wow....that's a lot of words ! I think you answered your own rant by pointing to chipset erratum rather than the CPU erratum you were originally talking about.
Here's the chipset erratum >> pdf<< (the SATA power saving issue is No.4 for example). I'm sure you can find the graphics, Gb LAN, SSD, and IO erratum as well under their respective actual categories.

 

[Darth Shiv]

Wow....that's a lot of words ! I think you answered your own rant by pointing to chipset erratum rather than the CPU erratum you were originally talking about.
Here's the chipset erratum >> pdf<< (the SATA power saving issue is No.4 for example). I'm sure you can find the graphics, Gb LAN, SSD, and IO erratum as well under their respective actual categories.

You are correct... they have listed the errata *now* but for the issues I mentioned it was months after release and after the horse had bolted so to speak. We'd already bought into the platform.

Edit: Just to add to that, every time a new chipset or processor comes out, early adopters are burnt on two fronts. Firstly, the errata is raw and people haven't discovered what new stuffups the platform contains. Secondly, they pay early adopter tax so they pay extra to work this stuff out. Call it a rant or whatever but it is the cold hard truth.

 

[dividebyzero]

Ah, okay, I see your point. You expect Intel to catch every erratum prior to shipping ?
Just as a point of comparison I'd take AMD's latest southbridge (I.e. chipset for chipset)- the SB950 as an example. Launched June 2011, yet it wasn't until June 2012 that AMD seemed to point out some of the erratic USB and SATA behaviour of the chipset. Off the top of my head I can think of a couple trio of USB issues - USB2.0 not being detected after resuming from S3/S4, system resets causing BSOD's because the USB controllers aren't reset also, and data cache errors with consecutive DMA requests. The SB950 also suffers from a loss of SATA performance due to IDE timeout errors.

So while I would agree that new adopters tend to take the brunt of the troubleshooting, that is almost entirely the case with any hardware/software- even the enterprise market. While some of us realize that the issue is vendor agnostic...and generally don't make an issue out of it, you seem to think that the problem is somehow Intel's sine qua non...unless of course you've also railed -for instance- against the s****y SATA and oddball USB performance of the 990FX/X/970 series of boards...if you have (although I can find no evidence thus far) I would at least consider you an equal opportunity ranter.

 

[Stupido]

... equal opportunity ranter.

LOL
this was good one, thanks!

 

[Darth Shiv]

Ah, okay, I see your point. You expect Intel to catch every erratum prior to shipping ?

Nope don't expect them to catch it all. For the record I haven't gone AMD since S939 because despite the rant, I think they are producing better than AMD. The recent SATA 3 and USB 3 issues are pretty "low quality testing" imho which is the reason for the rant.

 

[TS-56336]

Makes me feel like the old 775 days of me having to scour for E0 stepping E8400s. Guess I'll be waiting a bit longer for whenever the revisions are ...

 

[Zeromus]

Well it's a good thing, that by habit, I always check for any excess IO before performing any kind of of system power transition. Though I constantly have USB devices plugged into my system, I have good reason to trust their light indicators. I guess these issues aren't as bad as the SB9X0, where a kernel patch or revision is needed to address the issue, or a simple driver update on windows made possible by the HAL's driver isolation design.

 

[PC nerd]

First they screw up the IHS on the Ivy Bridge family, now this.

Intel are getting sloppy.

 

[Zeromus]

First they screw up the IHS on the Ivy Bridge family, now this.

Intel are getting sloppy.

Perhaps it's becoming more difficult to manage the complexity of the features they smack on these little squares. But I can't imagine them being unable to add a machine with usb 3.0 with power state testing to their testing suite.

 

[pmshah]

Ah, okay, I see your point. You expect Intel to catch every erratum prior to shipping ?
Just as a point of comparison I'd take AMD's latest southbridge (I.e. chipset for chipset)- the SB950 as an example. Launched June 2011, yet it wasn't until June 2012 that AMD seemed to point out some of the erratic USB and SATA behaviour of the chipset. Off the top of my head I can think of a couple trio of USB issues - USB2.0 not being detected after resuming from S3/S4, system resets causing BSOD's because the USB controllers aren't reset also, and data cache errors with consecutive DMA requests. The SB950 also suffers from a loss of SATA performance due to IDE timeout errors.

So while I would agree that new adopters tend to take the brunt of the troubleshooting, that is almost entirely the case with any hardware/software- even the enterprise market. While some of us realize that the issue is vendor agnostic...and generally don't make an issue out of it, you seem to think that the problem is somehow Intel's sine qua non...unless of course you've also railed -for instance- against the s****y SATA and oddball USB performance of the 990FX/X/970 series of boards...if you have (although I can find no evidence thus far) I would at least consider you an equal opportunity ranter.

I can vouch on the USB 2.0 issue from first hand experience. I have been an AMD fanboy since 1997. After I burned my hands on 5 AMD Phenom 9559 systems built on MSI motherboards I swore off the platform altogether abot a year ago. I had every kind of problem with their on board USB ports that one can think of. Finally I had to stick in via chipset based USB 2.0 addon pci cards to work around it.

BTW there was no way of contacting either AMD or MSI for solution to he problem. So much for customer support.

 

[pmshah]

Ah, okay, I see your point. You expect Intel to catch every erratum prior to shipping ?
Just as a point of comparison I'd take AMD's latest southbridge (I.e. chipset for chipset)- the SB950 as an example. Launched June 2011, yet it wasn't until June 2012 that AMD seemed to point out some of the erratic USB and SATA behaviour of the chipset. Off the top of my head I can think of a couple trio of USB issues - USB2.0 not being detected after resuming from S3/S4, system resets causing BSOD's because the USB controllers aren't reset also, and data cache errors with consecutive DMA requests. The SB950 also suffers from a loss of SATA performance due to IDE timeout errors.

So while I would agree that new adopters tend to take the brunt of the troubleshooting, that is almost entirely the case with any hardware/software- even the enterprise market. While some of us realize that the issue is vendor agnostic...and generally don't make an issue out of it, you seem to think that the problem is somehow Intel's sine qua non...unless of course you've also railed -for instance- against the s****y SATA and oddball USB performance of the 990FX/X/970 series of boards...if you have (although I can find no evidence thus far) I would at least consider you an equal opportunity ranter.

I can vouch on the USB 2.0 issue from first hand experience. I have been an AMD fanboy since 1997. After I burned my hands on 5 AMD Phenom 9559 systems built on MSI motherboards I swore off the platform altogether abot a year ago. I had every kind of problem with their on board USB ports that one can think of. Finally I had to stick in via chipset based USB 2.0 addon pci cards to work around it.

BTW there was no way of contacting either AMD or MSI for solution to he problem. So much for customer support.

Actually this is not a reply but a correction in my post as there is no edit function available.

I can vouch for .... The systems were built on Quad Phenom 9550 and not 9559. Mobos used were K9A2GMxxx.

 

[mailpup]

Actually this is not a reply but a correction in my post as there is no edit function available.

Yes, there is. If you enter the TechSpot forums, you can correct your posts from there. Look for the "Edit" button at the bottom of your post.