Sign in with
Sign up | Sign in
Your question

Hyper Thread is broken?

Tags:
Last response: in CPUs
Share
December 5, 2003 10:30:34 AM

HYPER THREADING (HT) PROBLEMS
I have a P4 3.06GHz Intel HT CPU that seems to have strange issues with the hyper threading feature.
First, I checked all the normal HT technology indicators were set correctly.
Intel 845 chipset HT enabled ASUS P4PE motherboard has most current BIOS (1006).
Multi -Threading or Hyper-Threading is turned on in BIOS.
Initial boot screen shows both processors as CPU1 and CPU2.
XPSP1 HT aware operating system shows both processors in the Hardware Manager within XP.
So far so good it seems.

The problem is when I try to use the Outlook Express / MSN mail program in XP. If I try to open the program, it doesn’t respond at all. I have to close Express / MSN with the CTR/ALT/DEL end now function and open my IE6 Explorer, and then open MSN mail from the read mail feature. Then it seems to open. I use this program every day, so why won’t it work with this CPU with HT enabled? Going into BIOS and disabling multi-threading fixes the problem. Reboot and the Outlook Express mail program opens fine in any order with no “not responding” issues.

What is broken the CPU, the program(s), the Operating System or the associated hardware and how to tell? Right now I paid a lot of money for a CPU that doesn’t do what it is supposed to do.

Are their any benchmarks that are HT enabled? Is Mad Onion’s 3Dmark 2003 HT aware? Is my second CPU per say working? With “one” CPU enabled the system runs fine. I really can’t explain Outlook Express / MSN’s issues except it hates HT turned on. This old program can’t be HT aware, yet it locks-up with HT enabled. I thought things were to run as though no HT existed with any non-HT aware programs. Not so, it seems. As is, I have to disable HT to use my computer.

Comments? Anybody have issues like this? Could my CPU be broken and how to tell what is if it’s not?

More about : hyper thread broken

December 5, 2003 5:11:08 PM

Microsoft.com best place to search

I dont like french test
December 5, 2003 9:02:14 PM

Well, still no luck after I have updated the following.
1.0 Searched XP update for new CPU HT driver (none).
2.0 Updated Modem drive.
3.0 Updated Soundcard (soundMax) driver.
4.0 Updated Intel 845 chipset drivers.
5.0 Update Intel Application accelorator program.
7.0 Update ASUS P4PE BIOS to 1007.

So far, in a rough search, the HT CPU seems to be causing modem and sound card related problems. I have both. I get stuttering and missing sounds in games under XPSP1 home with HT enabled and lock-ups when a program (Outlook Express in my case...so far)tries to connect to the internet.

I also get strange freezes in XP for no apparent reason. None of which I get with HT disabled in BIOS. As much trouble as this technology causes, I consider it broken. Intel, let me introduce you to Microsoft. Oh, I guess you already know them!
Related resources
December 6, 2003 1:12:17 PM

The technology is very sensitive to program bugs. That's because it effectively runs threads in parallel instead of closely behind each other. So, in a way it's very good that this happens because it's easier to detect bugs with multi-threading.

I'm quite certain that it's not a hardware fault. If it was, you'd probably not even be able to boot, especially because the technology is this sensitive.

Try throwing all your extra hardware out and reinstall Windows. Run Outlook Express and if the problem is gone, add your hardware one by one and you'll probably find the culprit. If the problem is not gone after the reinstall, the only other thing that might be broken is your motherboard.
December 6, 2003 3:08:15 PM

First off its not the CPU with so many system checks in the BIOS and OS boot cycle windows XP would display a big blue screen showing a General Protection Error was encountered.
I am going to hedge a bet here you have Symantec software for both a fire wall and Norton anti virus.
It sounds to me that you might have updated Windows XP and now the default settings for the anti virus software is corrupted.
The anti software has for lack of a better term gotten LOST. When you update at the Windows update site they re-enable secure logon for windows XP and some of the patches get in behind the original fire wall and anti virus software and those programs do not correctly function any more and deny access to the hard drive or anyone trying to use the computer main drive such as opening up Email accounts that have user authorization attributes.
If you are using a virus protection software package totally remove it and clean the registry then reinstall it the problem with MSN and your mail should go away.

Barton 3200+ 400MHz
A7N8X Deluxe
Liquid
2x512 KinstonHyperX PC3200
GeForce FX5900
Maxtor DiamondMaxPlus9@80Gig
SONY CD 52x
SONY RW 52x/24x/52x
SONY DVD 16x/40x
December 6, 2003 5:20:43 PM

I agree that the CPU is probably fine given the odds of booting with a bad CPU. That's why I'm so disappointed with HT technolory. XP isn't "broken" either but its full of bugs. The software compatability end of HT seems to follow a similar path. And no, I don't agree that making a system (hyperthreading) so sensitive that it crashes at every opportunity as a good thing. Too many systems have to run simultaneously to allow that kind of sensitivity. How about I design a car that's so sensitve to steering input you can't keep it on the road and then say you can't drive?

I use PCcillin anti-virus siftware and firewall. Your points are valid that it may be the problem with HT enabled. I can't see why just Express is affected, though. A reinstallation of XP will most likely only leave me sitting with the same issues in the end. I have solved few to no problems with windows reinstallations. Call it happen stance, but it's just a method to I use to admit that I don't know what's going on. When the problem persists after the reinstallation of XP, I know I don't know what's going on.

HT seems to be driver sensitive based on google sesarch on HT problems. And no, I doubt that the motherboard is broken for the same reason I doubt that the CPU is bad, the odds of booting with HT enabled in BIOS with a bad HT hardware level defect in the motherboard is slim to none.

My guess is poor software compatability with HT by about every software program that has been designed to run WITHOUT it. And why would software work with it when it was never been multi-proc aware? HT technology is responsible for staying out of the way of the existing software, not the other way around. I don't mind a minor "slowdown" with some apps in HT enabled mode, but to plain crash the system?

I'd be curious to compare notes on HT enabled systems out there. Intel has been known to make less than optimal decisions before. HT isn't optimal and so the decision wasn't either. MMX and SSE2 ETC were good designs that didn't crash the system and knew how to stay out of the way.

Hardwarev companies that incorporate software into their hardware aren't the best at it. It will improve, but it will take some time.
December 6, 2003 5:46:44 PM

I had the same problem with my Mail in XP that you say you are having.
I reinstalled the Anti-Virus software and the problem went away.
The thing that makes me almost sure its not the Hyper Threading is because when I had the same problem you describe I was not using a HT system.........

On the other hand I am still not using a Hyper Thread P.C for home use so I have very little knowledge of those systems. Other then installing the XP OS and getting them out the door to a customer.
We have had a few returns on a couple of HT systems but only because the customers tried to install Hardware themselves after the system was sold.
One guy keeps trying to run his P.C as the household EVERYTHING system to run his entertainment systems to I guess his garage door to ordering food through his fridge via a console built into the door.

The guy has his house wired.........but the P.C cant keep it all going and gives up that is how hard you need to bang on a HT machine before they crash.

It could be a hardware failure you are getting but it would have to be a very slight defect that does not interfere with booting the P.C but causes corruption later at the desktop.

I know that some systems when overclocked can cause data to be corrupted when written to the hard drive.
Perhaps your system is experiencing a MHz issue with the Bus for the AGP or CPU this would simulate an overclocking and cause the same corruption of the OS.

If you are going to do a complete reinstall try to download the MAX Blast software if you own a Maxtor Hard Drive it can be put into a floppy and used in DOS to diagnose the hard drive. Again more work but at least you would know the drive is good or bad before you try the reinstall.
Most drive makers provide a free diagnostic software download for their drives.

Barton 3200+ 400MHz
A7N8X Deluxe
Liquid
2x512 KinstonHyperX PC3200
GeForce FX5900
Maxtor DiamondMaxPlus9@80Gig
SONY CD 52x
SONY RW 52x/24x/52x
SONY DVD 16x/40x
December 6, 2003 8:34:27 PM

Quote:
And no, I don't agree that making a system (hyperthreading) so sensitive that it crashes at every opportunity as a good thing.

I completely disagree. See, without Hyper-Threading or multiple processors, a program can contain bugs with multi-threading that only very seldom causes a problem. But when it does, the program crashes or the system hangs. That's terrible, because the cause is extremely hard to find. With hyper-threading, a programmer can detect bugs which originate from bad multi-thread programming as soon as he runs his new code. So he can just rethink his last steps and fix the bug. Without it he had never known where or even why it happens so he wouldn't know how to start looking for it.

The downside is that programs that were written on machines without Hyper-Threading or multiple processors can contain these bugs that the programmer did not detect, and in practice might not cause a problem. But however you look at it, they are still bugs and should be detected. Even programs that were never meant to make use of multi-threading should still be aware of since it exists for more than 10 years now. It's only recently that these bugs are not tolerated any more.

So I think we're just in a transition state where all programs, and mainly drivers, will be rewritten to be more stable with respect to multi-threading. In the meanwhile, you just have to avoid using programs and drivers that have bugs, or disable Hyper-Threading if you really do want to use them.

The analagon is... (Ok use your imagination now). Suppose you've just bought a four wheel drive car. Damn cool compared to your old two wheel drive, huh? Only problem is, it's not allowed for any of your wheels to be lifted from the ground (e.g. on a ledge) or you'll get stuck (not 100% true but suppose it is). It's a rare thing to happen, but it occurs more than with the two wheel drive car because it allowed the other wheels to be lifted and still have traction. So what do you have to do with your 4WD? Refine your driver skills to never tolerate that any wheel doesn't touch the ground.

In computer world having no cardan joints would be like having no multi-threading (i.e. DOS). It the solution to all bugs related to multi-threading but you don't want it just as much as you don't want a car without cardan.

Ok it might be a bad example, but I hope you realize that blaming Hyper-Threading for program bugs is wrong. Being less tolerant for bugs is a good thing and will eventually lead to better programs everywhere. It's a good evolution but you just have to give it some time to weed out the bad seeds.
December 7, 2003 9:57:32 PM

A non multi-proc "highway" software program should never go "off-road" and get "stuck". A bad HT enabled CPU sticks it off road and gets it stuck. Again, the HT code in the CPU has to stay out of the programs way when it isn't needed. Not the other way around. To tell me that because multiproc PC's have been around confirms all programs are multi-thread aware is not true. Just the opposite is true. HT tries to "trick" non multi-thread programs into using wasted CPU cycles (a P4 wastes plenty compared to an AMD proc)and in the process, crashes programs when it takes some where they can't go. Either the program slows down, or it crashes.

I don't buy the lame explaination that multi-proc systems should have to be used to improve code on non multi-proc aware software. If so, and to accept your suggestion that multi-proc systems have been aroung for so long, programs would already be excellent because only a fool wouldn't use a mutli-proc system to write code!

I do agree that the attempt, and eventual improvement, of this technology is admirable.
December 7, 2003 10:44:28 PM

Quote:
...only a fool wouldn't use a mutli-proc system to write code!

Ok you've just called 99% of all programmers a fool...
December 8, 2003 12:21:32 AM

Stranger,

If the assumption is correct, that multi-proc systems are used to write ALL SOFTWARE (you are assuming they are in order to throw out an unsubstantiated silly comment)why would a program as common as Outlook Express not run with a HT processor enabled, and run fine with it disabled?

I'll answer that for you, because an HT type Intel processor ISN'T really a multi-proc processor. It interleaves instructions into a single processor with software imbedded into the CPU like MMX or SSE2 does. And, as such, it has issues in the code inside the CPU itself that needs some ironing out.

Go to Intel's own web site for MANY instances of this CPU either crashing or locking-up applications in XP. Or, run a google search on Hyper-Thread problems and read away.

I have said nothing that isn't true. The PREVIOUS comments about multi-proc systems finding errors in code is what is erroneous. Why? If multi-proc systems were ALWAYS used as suggested, and was used extensively, code would be better than it is right now. But, just as he suggested, not me. So to say programs will "improve with multi-proc usage" say that multi-proc systems aren't used extensively right now. Or, how could software get any better than it currently is with its use?

I don't care how code is written and I shouldn't have to. A CPU should stay out of the way of my programs. A CPU is designed to execute the code, not trip it up! If it does, well, we still have work to do. No crime in that.

I just asked for a work around if someone knows of one. So far all the answers point to some fundamental problem with hyper-threading.
December 9, 2003 1:32:36 AM

An example of HT issues on Intel's site.

http://intel.forums.liveworld.com/thread.jsp?forum=242&...

"I got a personal reply from Karl that gave some more details on the "solution" to his problem. I asked him to post that here, Karl said he would, and then there was the system shutdown. So, let me post what Karl wrote about this.

Thanks for getting back to me about the hyperthreading issue. Unfortunately (as advised by a microsoft posting), on computers faster than 3.00 Ghz running hyperthreading, some applications will fail to run or have unexpected fatal errors. This apparently is the case with our application.
The end result is that we have to run the computers and swap them for non-HT enabled computers. I'm rather surprised that this issue hasn't been formally discussed or fixed given that HT technology has been around for more than a year. Additionally, 95% of all computers now are being made to support it by default.

If anyone has any more details on this issue, please post them here. Is there some class of applications that are affected or some operating system services that don't work as expected at high processor speeds with HT on? Where was the Microsoft posting that detailed this information?
-- clay"

Now for you do it yourselfers (and you thought all software was compiled under dual processors, this say NO, NO, NO and they aren't any better at error detection than a single proc until......)

"Intel intros Pentium 4 hyperthreading bug tool
Takes apart threads to get closer to bugs
By INQUIRER staff: Tuesday 08 April 2003, 17:06

THE CHIP GIANT said that it has released a software tool to help developers isolate bugs in hyperthreading applications for Pentium 4s and for Xeons.The tool – called Thread Checker 1.0 – will track down threading problems in hours compared to what Intel said would otherwise take days or weeks.

It works by finding particular lines of code causing threading bugs and then categorises them showing different types of information relating to the problem.

The tool will become available this quarter for Windows XP Pro and Windows 2000 systems and costs $1,198 for a single user licence."
!