Hyperthreading & Velocity Engine

No, not at all. Altivec is for floating-point calculations, while Hyperthreading controls how the processor uses running threads. As you can see, I'm no expert at all, but AFAIK they are not comparable. I think Altivec is the equivalent of SSE (someone correct me if I'm wrong here).

Hyperthreading improves the performance of multiple threads running on the Intel processor with this technology.

The Velocity Engine is another name for Altivec, which is where you take a single instruction and have it operate on multiple sets of data at the same time.

So no, they are two different concepts.

Altivec is the equivalent of SSE

Click to expand...

Altivec is similar to SSE, but in the PPCs we use the Altivec unit is independent (might not be the case for all the Freescale imbedded units.)

SSE share resources with the FP unit on Intel CPUs (don't remember if its just registers or a bit deeper), so there is some penalty for its use.

AltiVec doesn't hurt lots of programs either, unlike hyperthreading. How many times have you seen the release notes for a driver or fairly complex program say "major bugs with hyperthreading enabled". Anyways, intel's dropping the technology in favor of dual core CPU's.

BEN

saabmp3 said:

Anyways, intel's dropping the technology in favor of dual core CPU's.

BEN

Click to expand...

Nope, its sticking around.

actually, hyperthreading is not that bad a technology, if intel could do it right... It really helps when you're trying to do a couple things at once, otherwise known as preemptive multitasking. So, lets say you're burning a cd and you wanna play a game at the same time. Hyperthreading will help. The problem is, when you're only trying to do one task. Hyperthreading actually in more cases than not (as of the current processors out) cripple a single apps performance.

Altivec, or the velocity engine, when used with certain apps, act similarly to a 128bit processor, and uses that as its engine, instead of a 32 or 64bit engine. The problem with the velocity engine is that not all apps are designed for it, but when an app is designed for it, the processor will scream. That's why a G4 today, is still fast, maybe not as fast as a G5, or an AthlonFX or a top of the line P4, but it's still fast.

Hope that answers some your questions.

One more thing, the new POWER5 has SMT support (simultaneous multi-threading) which is what hyperthreading is... so, in essence, with the POWER5, you have 8 physical processor, and 16 virtual processors. I wonder if the PPC980 will have SMT, or "hyperthreading."

The problem lies with Intel, they never patented SMT, so it's open technology. Which means, IBM will surely put it in their next processor, which they did. Oh yea, for all you fans of AMD, the PPC970FX has a hypertransport link... 970FX link it also looks like they increased the Front Side Bus too...

try and remember to post these in the right places...

IBM has plans to introduce its own version of hyperthreading in future generations of its Power5 chips. Its called SMT - simultaneous multithreading. Hopefully this will be included in the Power5 derivative for the next gen Apple Processor.

*edit* damn, I go and move the thread and I get beat on the SMT post */edit*

D

Mr. Anderson said:

try and remember to post these in the right places...

IBM has plans to introduce its own version of hyperthreading in future generations of its Power5 chips. Its called SMT - simultaneous multithreading. Hopefully this will be included in the Power5 derivative for the next gen Apple Processor.

*edit* damn, I go and move the thread and I get beat on the SMT post */edit*

D

Click to expand...

sorry buddy...