A Digital Age Deserves A Digital Leader

Intel i7 CPU: Is hyper-threading potentially a problem?

Intel i7 CPU: Is hyper-threading potentially a problem?

Postby JabbaPapa » Tue Jul 28, 2009 10:16 pm

There is one potentially serious issue with the i7 chips, which is that Intel is using hyperthreading on at least some of them :(

My experiences on clients' rigs with the old P4 dual-core CPUs with hyperthreading has been overwhelmingly negative. I'm talking serious partitioning errors, deep Windows instability, and potential data loss negative here ... although to be fair I have also seen the odd hyperthreading rig with no serious problems at all.

It is of course possible that Intel has ironed out some bugs since the P4 version ; but generally speaking it's still the cheap-o method to get extra cores out of a piece of silicon, and therefore suffers from the limitations of being cheap-o ; and in the best of cases, it remains a form of virtualisation so that performance will in any case take a hit, because the extra virtual cores are created at the software level.

If you do decide to buy an Intel i7 for the triple-channel DDR3, then I would anyway advise you to look for a CPU that does NOT integrate this tech.
Image
PRO VETERAN
User avatar
Posts: 9538
Joined: Sun Feb 22, 2004 5:17 pm
Location: Monte-Carlo
Real Name: Julian Lord

Re: Its that time again = new desktop!

Postby Grav!ty » Wed Jul 29, 2009 12:14 am

JabbaPapa wrote:There is one potentially serious issue with the i7 chips, which is that Intel is using hyperthreading on at least some of them :(

My experiences on clients' rigs with the old P4 dual-core CPUs with hyperthreading has been overwhelmingly negative. I'm talking serious partitioning errors, deep Windows instability, and potential data loss negative here ... although to be fair I have also seen the odd hyperthreading rig with no serious problems at all.

It is of course possible that Intel has ironed out some bugs since the P4 version ; but generally speaking it's still the cheap-o method to get extra cores out of a piece of silicon, and therefore suffers from the limitations of being cheap-o ; and in the best of cases, it remains a form of virtualisation so that performance will in any case take a hit, because the extra virtual cores are created at the software level.

If you do decide to buy an Intel i7 for the triple-channel DDR3, then I would anyway advise you to look for a CPU that does NOT integrate this tech.


Tinfoil Hat Woo Woo Alert! :lol:

This prejudice against hyper-threading is totally unsubstantiated and has no foundation technologically. It has been repeatedly proved that there are substantial performance gains to be had from it, even when software is not multi-thread optimized. There is no documented evidence of "deep Windows instability" or "potential data loss" caused by it anywhere to be found. Don't you think there'd have been some sort of backlash and huge numbers of forum posts around the net complaining about it if there was? See HERE for more information on the subject.
Image

"The great majority of mankind are satisfied with appearances, as though they were realities, and are often more influenced by the things that seem than by those that are." - Niccolo Machiavelli
PROfessional Member
User avatar
Posts: 15790
Joined: Tue Sep 14, 2004 5:22 am
Real Name: Graham

Re: Its that time again = new desktop!

Postby JabbaPapa » Wed Jul 29, 2009 6:05 am

Grav!ty wrote:Tinfoil Hat Woo Woo Alert! :lol:


The worst PC instability issues that I have had to deal with in my time have in their majority involved CPUs using hyperthreading technology.

This is just my experience mate.
Image
PRO VETERAN
User avatar
Posts: 9538
Joined: Sun Feb 22, 2004 5:17 pm
Location: Monte-Carlo
Real Name: Julian Lord

Re: Its that time again = new desktop!

Postby Grav!ty » Wed Jul 29, 2009 6:57 am

I know Julian and I'm not intending to be offensive or harsh, because it's great to be able to have these debates. It's just that's not enough to trash it on or to put out "health warnings" on it. There's just too many other factors involved that could have led to issues you came up against on system using it. I've been looking for more balanced information on negative aspects of it since your post in the news section the other day, and all I find is some negative sentiment going back to the P4's that used it relative to AMD architecture of the same time, but nothing at all that's grounded in fact.
Image

"The great majority of mankind are satisfied with appearances, as though they were realities, and are often more influenced by the things that seem than by those that are." - Niccolo Machiavelli
PROfessional Member
User avatar
Posts: 15790
Joined: Tue Sep 14, 2004 5:22 am
Real Name: Graham

Re: Its that time again = new desktop!

Postby JabbaPapa » Wed Jul 29, 2009 12:32 pm

Obviously I cannot **prove** my negative experiences with hyperthreading, but I have still seen time and time again, even on some not-so-old CPUs, that partitioning has been severely damaged on those machines, requiring data recovery, partitions *deletion* (not just formatting), and reinstall everything from scratch (or otherwise, just "normal" Windows-stops-working syndrome). Same symptoms, same CPU tech.

A client of mine has three computers, ALL of which had hyperthreading CPUs (a P4 and 2 Atoms), and they have ALL of them broken down -- two of them in the typical manner that I have come to associate with hyperthreading which is that some data-corruption damage to Windows occurs requiring format and full reinstall. After some months of ongoing and continual breakdowns, which I predicted to the client unless she changed CPU, she eventually agreed to have a new custom-built computer.

I have installed W7 on the second, and we will see if W7 can manage hyperthreading and avoid the instability. (?)

The third I have so far been unable to fix.

Like I said, maybe Intel has ironed out the bugs with NuHyperthreading, and I _have_ also seen some machines with hyperthreading CPUs that work perfectly well -- and it's possible that certain motherboards are to blame rather than the CPUs. But as the chipsets on these boards are designed for these CPUs anyway, does it really make much difference ?

My ongoing hands-on experience has led me to a very, very deep distrust of hyperthreading, and I would NOT be at all surprised if I were to see clients' i7-based rigs completely and similarly broken down in future.

Comparing with equivalent non-hyperthreading CPUs is somewhat difficult, because the number of cores at any given date will be different -- but a CPU with X physical cores will at equivalent clock rates and everything provide better performance than one with X/2 physical and X/2 virtual cores simply because there will be no software-based virtualisation managing any extra cores during every single clock cycle --- BUT hyperthreading CPUs typically have up to double the number of apparent cores than non-hyperthreading ones of the same generation.

I think that most people automatically assume that when Windows breaks down it's something wrong with Windows -- which likely explains why you never see many people complaining about fundamental hardware instability. Of course, most people are incapable of detecting it when it occurs, for them it's just Windows. Another reason is that the specialist benchmarking sites that compare various CPUs use newly installed hardware -- whereas the problems I have seen involving hyperthreading CPUs all seem to surface after some months of use, instead of immediately.

---

Anyway, given that ALL i7 CPUs use hyperthreading, I personally would not touch any of them with a 10-foot pole ; and I would therefore personally wait for triple-channel DDR to be available with other CPUs before adopting this new standard ... the first triple channel -capable 6-core AMD CPUs are scheduled for the end of the year as far as I can see (which might mean early 2010 given the history), and I do not know if Intel has any plans to sell CPUs *with* the triple channel controller but *without* the hyper-threading.

Anyway, Intel has managed to screw AMD yet again, by simultaneously reintroducing the unstable hyperthreading *and* this new method of putting the memory controller on the CPU instead of the motherboard *and* winning the race to put out the first triple-channel DDR3 hardware. Talk about a triple whammy :-?

Intel have set the bar so high now, that AMD are going to have HUGE difficulties even to just catch up :/ -- but I would personally STILL buy AMD because stability is generally guaranteed with AMD, whereas with Intel it can be pot luck.
Image
PRO VETERAN
User avatar
Posts: 9538
Joined: Sun Feb 22, 2004 5:17 pm
Location: Monte-Carlo
Real Name: Julian Lord

Re: Its that time again = new desktop!

Postby Grav!ty » Wed Jul 29, 2009 1:30 pm

I think you are in deep prejudice mode on this thing Julian. I've owned and built many P4's for clients and friends with hyper-threading (probably 50 or more) and never experienced any issues such as that you describe with any of them.

In all these years since it's first introduction, I'm inclined to think that if there truly was an issue with them, firstly it would have been identified by Intel and secondly that the net would be littered with reports of the issues. There still seems to be nothing you are basing your "findings" on other than "observations" which could be caused by a multitude of factors. In the first case, it just doesn't make sense that Intel would build this technology into a totally new architecture and in the second instance it must be because there were never problems.

Besides, most if not all motherboards designed for hyper-threaded CPU's have the ability to disable that feature, which is how performance tests of with and without hyper-threading are performed.
Image

"The great majority of mankind are satisfied with appearances, as though they were realities, and are often more influenced by the things that seem than by those that are." - Niccolo Machiavelli
PROfessional Member
User avatar
Posts: 15790
Joined: Tue Sep 14, 2004 5:22 am
Real Name: Graham

Re: Intel i7 CPU: Is hyper-threading potentially a problem?

Postby imnuts » Wed Jul 29, 2009 2:30 pm

I've worked on hundreds of computers in the past when i was at college and many of them had hyper-threading P4s when I was there as that was the technology then. In that time, I never saw any data corruption/instability issues caused by the CPU unless it was because the desktop P4 was installed in a laptop and the CPU was overheating, but it never caused corruption of data. If you're seeing data corruption and other issues with the hard drive, I'd be more inclined to think you have an issue with the hard drive in some way, or the bus it's connected to. One other issue that could cause this would be an overworked power supply, but I have never seen any issues with hyper-threading such as you are suggesting, and I agree with Graham that if there really were these issues, there would be more info out there about it and they wouldn't have put it back into the chips.
Image
PRO SUPREME
User avatar
Posts: 7457
Joined: Wed Mar 24, 2004 5:19 am
Location: Boothwyn, Pennsylvania
Real Name: Mark

Re: Intel i7 CPU: Is hyper-threading potentially a problem?

Postby JabbaPapa » Wed Jul 29, 2009 5:57 pm

Well these might be the results of an unusually bad streak, and it might be pure coincidence that a clear majority of rigs with H-T CPUs that I've worked on have had these issues of low-level OS and/or database and/or partitioning instability.

Anyway if you dig around you CAN find some explanations, for example in HERE : http://www.videoguys.com/Guide/C/DIY+Sy ... dfda0.aspx (which btw is a POSITIVE review of the i7)

Videoguys wrote:What is Hyper-Threading? It’s when the processor splits itself into two virtual processors in order to share the workload it’s being asked to run. In its first incarnation, going back to P4 processor, Hyper-Threading was hit and miss. Under Windows 2000 Intel actually recommended shutting down Hyper-Threading as it caused instability. Videoguys also recommended turning off Hyper-Threading with certain NLEs and video capture cards. While the concept was great, the execution wasn’t there yet. Mainly because most software and operating systems were not optimized for more then one processor, so while they were Hyper-Threading aware, the cost in stability often outweighed any performance gains. All it took was one finicky application or device driver and Hyper-Threading became more trouble then it was worth.


Another possible cause of instability I have found mentioned on other sites is if Windows is installed with HT off in BIOS, and is then used with HT switched on -- but it is common practice among OEMs to install cloned hard drives in factory-made machines, which could occasionally lead to this cause of instability.

There are also quite a few sites of some software or hardware publishers where they recommend disabling HT as the combination of their products with HT caused instability. The wording often contained in these recommendations appears to suggest, as in the above example, that HT is unstable on some particular hardware configurations -- which could of course explain the string of bad computers that I have personally encountered, given admittedly that I tend to work with computers that are broken instead of being fully functional ones :P

And let's not forget that Intel retired the technology for several years. Why ? If there was nothing wrong with it ?

And reading between the lines of that news article on W7 and HT, well if W7 provides improved performance with HT doesn't this also mean that previous Windows versions provided inferior performance ?

----------

I'll repeat myself -- maybe Intel has ironed out the bugs. Maybe Microsoft has ironed out the types of Windows instabilities I have encountered. This could even be the reason why the technology is making its return.

Again, there's nothing I can do to prove the validity of my perceptions, all I can do is report on them.

If the i7 version of H-T is actually stable, well then good for Intel, and good for anyone buying these CPUs !!!
Image
PRO VETERAN
User avatar
Posts: 9538
Joined: Sun Feb 22, 2004 5:17 pm
Location: Monte-Carlo
Real Name: Julian Lord

Re: Intel i7 CPU: Is hyper-threading potentially a problem?

Postby JabbaPapa » Wed Jul 29, 2009 6:06 pm

imnuts wrote:If you're seeing data corruption and other issues with the hard drive, I'd be more inclined to think you have an issue with the hard drive in some way, or the bus it's connected to.


I normally assume that issues on individual rigs are local issues, yes.
Image
PRO VETERAN
User avatar
Posts: 9538
Joined: Sun Feb 22, 2004 5:17 pm
Location: Monte-Carlo
Real Name: Julian Lord

Re: Intel i7 CPU: Is hyper-threading potentially a problem?

Postby Grav!ty » Thu Jul 30, 2009 3:14 am

Hardcore AMD supporters are always going to find fault with Intel. The majority of users tend to go for performance irrespective of who it comes from, as was seen by the massive switch to Intel even by previously sworn AMD fan boys.

In my opinion, AMD made some errors in judgment taking on ATI when they did, and dissipating their core work on CPU's. It's probably the correct strategy for the longer term but short term they bit off too much in it's expensive absorption and the adaptation they were forced into as a result.
Image

"The great majority of mankind are satisfied with appearances, as though they were realities, and are often more influenced by the things that seem than by those that are." - Niccolo Machiavelli
PROfessional Member
User avatar
Posts: 15790
Joined: Tue Sep 14, 2004 5:22 am
Real Name: Graham

Next

Return to Hardware and Customizing

Who is online

Users browsing this forum: No registered users and 2 guests