NotebookForums.com › Forums › General Notebook Discussions › Notebook Forums - General › I feel sorry for Intel... NOT!!!
New Posts  All Forums:Forum Nav:

I feel sorry for Intel... NOT!!! - Page 4

post #61 of 71
8/7/04 at 7:50pm
Quote:
Originally Posted by Superguy
Thing is, though, the average person isn't going to buy an FX or an Opteron.

The place where they're going to get confused is in the mainstream and value processors.

People are going to see a big difference in performance between a 2800+ Sempron and a 2800+ A64 much like a 3.0 P4 and a 3.0 Celeron. Clockspeeds are roughly the same (or at least the PR is), but they don't perform nearly the same by a long shot.
AMD's performance rating (PR) only denotes relative 32-bit performance, and processors with the same rating should perform on average roughly the same. Anyways, I hope this rating system will be replaced with model numbers like on Opteron & Athlon 64-FX soon.
Reply
Reply
post #62 of 71
8/7/04 at 7:54pm
Quote:
Originally Posted by Superguy
I have to find an AMD chipset that can offer me everything I want in a motherboard. Intel can do that, and it works very well.
I really would like to know what exactly the chipsets for AMD processors lack and what dosen't work very well on them in your opinion?
Reply
Reply
post #63 of 71
8/7/04 at 8:16pm
Quote:
Originally Posted by Superguy
Yes, it's SO dead end and such a bad idea that IBM uses it in its POWER series.

Give me a break.
You definately should read this article:
http://arstechnica.com/cpu/004/presc...rescott-1.html
Quote:
IBM's POWER5 uses hyperthreading, and it certainly doesn't have the outrageous 31-stage pipeline length of Prescott. In fact, the POWER5's 16-stage pipeline isn't much longer than the Pentium M's speculated pipeline length of 12 to 14 stages. I say this only to illustrate the point that there's nothing in the lower number of pipeline stages that somehow magically makes the Pentium M a poor candidate for hyperthreading.

The Intel rep's statement only makes sense if you think of hyperthreading in the terms I suggested in my article on the subject: as a latency-hiding technique. Sure, hyperthreading improves execution efficiency, or "throughput," by (ideally) keeping the processor's pipeline filled with instructions from multiple threads, but where it really shines is in situations where one thread is stalled and the other is still moving. In a normal processor, a pipeline stall would mean that multiple clock cycles could pass without doing useful work. With a hyperthreaded processor, instead of waiting around for the pipeline to come unblocked the chip can just continue executing instructions from the non-stalled thread; so instead of the processor's throughput going to zero as a result of a stall, it merely slows down.

In a processor that's as deeply pipelined as Prescott, waiting on a cache access means multiple dead cycles, especially considering the fact that Prescott's L1 and L2 cache latencies are worse than those of Northwood (significantly worse in the case of the L1); and waiting on a main memory access means lots of dead cycles for a processor that's running at upwards of 3GHz. So what the Intel rep was really saying was this: Because we got obsessed with MHz as a marketing number and made Prescott's pipeline so ridiculously long, Prescott benefits much more from a latency-hiding technique like hyperthreading than a saner design like the Pentium M.

There's actually another reason why the Pentium M won't benefit as much from hyperthreading. The Pentium M's branch predictor is superior to Prescott's, so the Pentium M is less likely to suffer from instruction-related pipeline stalls than the Prescott. This improved branch prediction, in combination with its shorter pipeline, means improved execution efficiency and less of a need for something like hyperthreading.

One question that the preceding discussion might raise for the casual reader is this: if the POWER5's pipeline is roughly half the length of the Prescott's, why does it need hyperthreading? The answer is that the POWER5 is extremely wide, pipeline stalls translate not into tons of unused clock cycles but tons of unused execution hardware. Or, here's another way of phrasing it: in the event of a pipeline stall, Prescott has a few execution units sitting idle over a large number of clock cycles; the POWER5 has a large number of execution units sitting idle over a few clock cycles. Either way you slice it, that's wasted resources.
It's almost for the exact same reasons as Pentium M that Athlon 64 wouldn't benefit much from hyperthreading, since its pipeline length is only 12 stages and much narrower compared to that of POWER5.
Reply
Reply
post #64 of 71
8/7/04 at 11:24pm
Quote:
Originally Posted by OGREtheBUFFOON
Still, HT isn't all Intel cracks it up to be. It really only makes a significant difference in apps that are written for it, and most aren't.
But it's not as little as the deriders make it out to be either.

There are instances where the OS can schedule two different threads to simultaneously. These don't need to be recompiled and can make multitasking smoother.

HT definitely helps in the video encoding I've done, so it isn't useless.

For an individual app to take advantage, yes, optimizations need to be done, but that's really no different than any other instruction set out there either.

Super
Reply
Reply
post #65 of 71
8/7/04 at 11:27pm
Quote:
Originally Posted by snorre
AMD's performance rating (PR) only denotes relative 32-bit performance, and processors with the same rating should perform on average roughly the same. Anyways, I hope this rating system will be replaced with model numbers like on Opteron & Athlon 64-FX soon.
There's still a very real difference between what the Semprons and XPs can do at the same rating versus the Hammers, which was my point.

They can keep the xxxx+ rating which will help differentiate within families. They just need a scheme to differentiate them.

Opterons and Intel's scheme make more sense than the FX IMO.
Reply
Reply
post #66 of 71
8/7/04 at 11:43pm
Quote:
Originally Posted by snorre
I really would like to know what exactly the chipsets for AMD processors lack and what dosen't work very well on them in your opinion?
Here's my beef with chipsets on AMD:

AMD doesn't make their own chipsets, and I think this is a problem. It's less of a problem with the Hammers as it doesn't need the northbridge now, but it can still be a problem.

I've always believed that the processor maker, regardless of whose it is, can make the best chipset. Why? They know the chip more intimately than anyone else. The chipsets that they have made have been good. Problem is that they don't make very many of them and rely on third party vendors to supply the chipsets. While 3rd party chipsets are nice for choice, I think AMD's taking an easy way out in not supporting them with their own platforms, rather "seeding" the market with their own chipsets until others could mass produce them. This happened with the 750 and the Athlon and the 760 when DDR memory was supported. I think AMD's been shooting themselves in the foot by not more widely marketing the chipsets after they spent all that money in R&D and creating a market for their products.

That said, we are left with 3rd party chipsets. Via was the chief offender for a long time. I've hated their Intel products and their AMD products were nightmares for many. Stability problems, buggy 4 in 1 drivers that never worked right, Sound Blaster Live problems, data corruption on hard drives, incompatibilities with hardware, etc. Via has cleaned up their act, but I still won't touch them. The 761 with the AMD northbridge and Via southbridge was a bad combo. I spent a lot of time on my Dad's machine that had that.

Nvidia probably comes the closest to the best, but they're chipsets have been somewhat buggy, specifically when it came to IDE performance. Their sound was the best onboard solution around. If I built an AMD machine, I'd probably put it on an Nvidia board.

SiS has come a long way, but I don't think they're there yet.

Intel based motherboards tend to be the most feature packed and well implemented. I love the onchip serial ATA RAID, Intel NICs perform very well with less CPU utilizations, their drivers are updated more regularly than most, they're stable, and it just works and works right at that.

AMD boards can offer similar features, but they're implemented with chips offered by a mish mash of companies (ie NIC can be from one company, SATA from another like Silicon Images, etc).

One of the biggest reasons I've gone with Intel the last few times was due to motherboards and chipsets, not processor. I think Intel's just good at it because they've done it for a very long time.
Reply
Reply
post #67 of 71
8/7/04 at 11:45pm
Quote:
Originally Posted by snorre
You definately should read this article:
http://arstechnica.com/cpu/004/presc...rescott-1.html

It's almost for the exact same reasons as Pentium M that Athlon 64 wouldn't benefit much from hyperthreading, since its pipeline length is only 12 stages and much narrower compared to that of POWER5.
I read it when you posted it the other night and it was a good read.

They also said in that article that the reason why Prescott got it and Dothan didn't was that Prescott needed it much more than Dothan did.

I still think it could help other procs. I think if AMD used it it'd blow Intel away.
Reply
Reply
post #68 of 71
8/8/04 at 11:57am
In other news, Intel to Discontinue Pentium 4 Extreme Edition Processor. (Intel to Remove 3.20GHz Extreme Edition Processor from the Family)

[edit] I originally saw this on Slashdot, but I cannot find anything on Intel's site about this, so take it with a grain...
Reply
Reply
post #69 of 71
8/8/04 at 3:15pm
i bet we don't see any price drops because of this. Bah the worst thing bout Intel is that they charge more than they need to for chips.
Reply
Reply
post #70 of 71
8/8/04 at 3:38pm
Which processor would run Mysql or some other database better?
Reply
Reply
post #71 of 71
8/9/04 at 11:50pm
Quote:
Originally Posted by 46and2
In other news, Intel to Discontinue Pentium 4 Extreme Edition Processor. (Intel to Remove 3.20GHz Extreme Edition Processor from the Family)

[edit] I originally saw this on Slashdot, but I cannot find anything on Intel's site about this, so take it with a grain...
One processor, not the family. Chips normally get discontinued.

They're releasing new ones with the 925XE with a 1066 bus:

http://www.theregister.co.uk/2004/08/09/intel_p4ee_720/
Reply
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Notebook Forums - General
NotebookForums.com › Forums › General Notebook Discussions › Notebook Forums - General › I feel sorry for Intel... NOT!!!