Socket AM2 Has A Secret Weapon.

[chovi]

http://www.theinquirer.net/?article=32589

"It seems that all AM2 CPUs were outfitted with a support for Reverse-HyperThreading, an architectural change which enables software to think that it is working on a single-core alone. By combining two cores, the company has been able to produce the six IPC "core" that will go head to head against four IPC "core" from Conroe/Merom/WoodCrest combo.

It seems that in certain cases, even an old AMD Athlon 64 3800+ can wipe the floor with Core 2 Duo E6300 CPU.

As we all know - the results from E6700 and X6800 against FX-62 will be nice, but the real fight with AMD is the one for the Conroe with 2MB of L2 cache. The system memory avoidance technology is working flawlessly on a 4MB cache model, but the case is reversed in the two Meg cache variant, especially in cache-hit sensitive apps, such as games.

In single-treaded apps, Core 2 Duo is expected to struggle against Reverse HyperThreading CPUs, which work at higher clock frequencies and produce higher instruction per clock ratios (IPC).

AMDs Reverse-HT is a dynamic technology, and with Microsoft's Windows update and a new processor driver, the driver will copy the graphics drivers of today's 3D accelerators. The driver will detect the app, see if it is multithreaded or not and turn the ReverseHT on, or leave it off".

 

[Harrigan]

That would be so cool...but....The Inquirer? I think we will see this but not hidden in current AM2 cpus, probably not until K8-L, by which time it will be too late.

 

[lee87]

Sounds good, does this mean a dual core at 2.5ghz would perform the same as a single core 5ghz AMD in single threaded apps?

 

[Dolph]

And of course, this has all shown up in the benchmarks that have been run and consistantly shown zero tangible benefit to AM2 technology....

You do realise the Inquirer is one of the single worst resources for spouting absolute tripe on the entire internet...

 

[chovi]

yeah but at least its somn to read on this fine friday morning !

 

[lee87]

And of course, this has all shown up in the benchmarks that have been run and consistantly shown zero tangible benefit to AM2 technology....

You do realise the Inquirer is one of the single worst resources for spouting absolute tripe on the entire internet...

I've just read on another forum that you need a CPU driver and a bios update to enable the reverse HT, which isn't out yet so benchmarks wouldn't show any improvement.

It's on a few other sites as well. Though I don't know if any of these are reliable.

http://www.bit-tech.net/news/2006/04/17/amd_reverse_hyperthreading/
http://www.geek.com/news/geeknews/2006Apr/bch20060417035855.htm

 

[Explicit]

And of course, this has all shown up in the benchmarks that have been run and consistantly shown zero tangible benefit to AM2 technology....

You do realise the Inquirer is one of the single worst resources for spouting absolute tripe on the entire internet...

^

 

[BubbySoup]

IF it is true, I believe the reason we haven't seen it working yet, is that it is supposed to need some sort of driver/bios update to enable it.

It won't however, be the first time a CPU has been released with features included that have been disabled. Intel did it with HT I think, although it was actually physically disabled which is slightly different I guess.

 

[Dolph]

I've just read on another forum that you need a CPU driver and a bios update to enable the reverse HT, which isn't out yet so benchmarks wouldn't show any improvement.

It's on a few other sites as well. Though I don't know if any of these are reliable.

http://www.bit-tech.net/news/2006/04/17/amd_reverse_hyperthreading/
http://www.geek.com/news/geeknews/2006Apr/bch20060417035855.htm

Those other sites talk about roadmapped work in progress, NOT something that is fitted to existing chips but not enabled via bios/drivers, both also reference the same french source site, which is something that hasn't been substantiated on any of the more reputable hardware sites as anything more than a rumour.

To jump from an unsubstantiated rumour that something is under development, to the idea that it's already been developed and has been fitted to existing chips and just disabled, is a little bit far out for me. Unless there's confirmation of the news in an AMD press release, or by one of the bigger hardware sites directly (ie not reporting from the same unproven source) I'll stick to my guns on this one

 

[Uriel]

Well if it turns up on AM2 I'm definitely getting that upgrade board for my ASROCK...

I seriously doubt that they'll enable it for existing CPUs. Why would you want to do that when you could just release a new set of CPUs with it as a feature and charge loads more money?

It would be on par with Microsoft giving away a bunch of stuff to existing customers when they cut the price of the xbox a few years ago. I don't think AMD can afford the same kind of loss-making masterplan.

 

[Stelly]

And of course, this has all shown up in the benchmarks that have been run and consistantly shown zero tangible benefit to AM2 technology....

You do realise the Inquirer is one of the single worst resources for spouting absolute tripe on the entire internet...

I agree with that

Stelly

 

[Dolph]

It won't however, be the first time a CPU has been released with features included that have been disabled. Intel did it with HT I think, although it was actually physically disabled which is slightly different I guess.

But in the case of HT the issue was that some CPU's had it, and others had it physically disabled, as it's cheaper to make them all with the technology then disable it than to have two seperate production lines (similar to the concept of speed binning. You don't make processors of different speeds, you make lots of identical processors then set the multipliers differently based on likely performance/market demand. That's why some chips/steppings are uber overclockers, but it's in no way guaranteed)

 

[FreeStream]

I was reading that you can't actually physically make a single threaded app use two cores, something about splitting then re-combining and it would use more CPU power than it would give


I think there isn't a reason this has happened before, it aint possible?

 

[Dolph]

I was reading that you can't actually physically make a single threaded app use two cores, something about splitting then re-combining and it would use more CPU power than it would give


I think there isn't a reason this has happened before, it aint possible?

Well, what we're really discussing is CPU level out of order execution across multiple cores. It's possible, certainly, but it's going to encounter significant overheads to do it. On a standard single core cpu you already have a significant amount of branch predicting and scheduling hardware to manage multiple execution units. The way you would do what AMD are rumoured to be doing is moving that up a level, so you have scheduling and prediction logic above core level so it sees the two cpu's execution units as one big pool, rather than two discreet ones. This would also require a major set of hardware at the other end to put everything back together in the correct order again, and personally I can't see how it could be efficent. It certainly couldn't be implemented that efficiently without major architectural revisions (which is why I seriously doubt that existing K8 core chips could have it and have it do anything meaningful or useful)

For some more explaination on basic CPU architecture, I'll point you to this thread, and the diagrams, imagine moving the logic further up and you'll see why I'm very skeptical about the 'reverse hyperthreading' thing

http://forums.overclockers.co.uk/showthread.php?t=17569099&highlight=conroe

 

[killer_uk]

If this is true, does Intel have something to counter it?

Core Multiplexing Technology (CMT)

Take a look over at XS

 

[Angilion]

If it did exist in a useable form on existing chips, why would AMD disable it? Particularly with Core 2 looking rather good against Athlon 64. If AMD could implement a significant performance increase simply by releasing a driver, I think they would have done it already. Actually, I think they wouldn't have disabled it in the first place. One of the impediments to AM2's success in the market is that it offers minimal performance increase over S939, and what little there is comes from the memory. So if AM2 did offer a significant performance benefit over S939, in the shape of this hypothetical reverse hyperthreading on the CPUs, what reason would AMD have to disable it?

 

[sargatanas]

The thing i dont get, is even if it is true, why would they let Intel get away with this much hype for Conroe? So many people have already made the decision to buy Conroe that anything AMD say just now is too late to majorly affect sales.

 

[jellybeard999]

sounds a load of crap tbh...

like the brilliant hyperthreading

only time will tell

 

[NathanE]

The really funny thing about this is that it's very likely a bluff, playing into the current demand for "pro-AMD" tech' news. The irony is that Core 2 architecture _does_ have some form of hyperthreading built into it (though disabled in the first wave of chips) and Intel seems very reluctant to talk about it

The HT present in NetBurst chips was just one form. There are many different forms of HT. I'd hazard a guess that Core 2 may have some variation of SOEMT - a type of HT that works well on small-average length pipeline CPU designs.

It's possible, certainly, but it's going to encounter significant overheads to do it.

Of course. Even the HT in NetBurst had a large overhead. Anything that adds delay to instructions reaching the pipeline will. However the aim of all types of HT is to provide a performance boost large enough to negate that overhead and then some.

Also note that this type of "inverse hyperthreading" (god I wish people would stop calling it that...) has been done before on mainframe chips (e.g. PPC AS400) for donkeys years.

 

[FrankJH]

two things occur to me

1) if this tech existed in current chips , its possible AMD wanted AM2 out and stable before releasing new bios's / drivers when they saw how bad the conroe tests where (for AMD themselves)

2) Present AM2 systems surely would be littered with unknown device(s) in Device Manager if drivers were not available - even if bios was required surely it would still pick it up as "something"?