Slashdot Mirror
Intel Delays Dual-Core Processor, Plans New Server Chip
Kajakske writes "Intel said Thursday that it is pushing back the release of its first dual-core processor by a year to 2005 and adding a new microprocessor for servers to its Itanium II lineup. On the other hand, Intel is moving forward in the area of new technologies."
The Motorola 85xx chip might be going for use in embedded devices, but I can almost guarantee you that it will *not* be used in a cell phone. A PDA? I doubt it, unless it's compatible with the chips that are already used in PDA's. This chip is more for things like network hardware, cable boxes, cars, and the like. It draws too much power to put it in a cell phone, and it's not quite powerful enough to put in a desktop.
If you're looking for the next generation of PowerPC chips, look to IBM's PPC970.
"The image is a dream. The beauty is real. Can you see the difference?" -- Richard Bach, Illusions
So, listen up, Intel - the server market may pay more per chip, but we "mere" home users buy a HELL of a lot more of them Server chips are sold at an unbelievable markup, though, so they make more profit from them. It's not uncommon to spend a couple grand on a CPU module for a server.
"The image is a dream. The beauty is real. Can you see the difference?" -- Richard Bach, Illusions
IBM ($81.19billion FY 2002) is four times the size of Intel ($26.76billion FY 2002) in revenue terms.
Bad analogies are like waxing a monkey with a rainbow.
From what I've read, HT doesn't even have a possibility to slow things down. Do you know how multithreading works in an SMP environment?
What HT does is allows this single CPU to pretend to be 2 independent CPUs, effectively splitting it in half (but not necissarily down the middle). The upshot of this is that it can more effectively deal with cache bubbles and all those horrible performance-draining problems Intel chips, with their insanely deep pipelines, are vulnerable to.
Basically, if you only throw a single thread at the processor, only the first virtual processor does the work and the other virtual processor is idle, allowing the entire processing power of the computer to deal with one problem, instead of half of it sitting idle. This is an advantage because HT only requires 5% more transistors, and the net benefit is something like a 20% performance increase. Of course, if you're not doing any work where you actually *use* multithreaded apps, you'll never understand why HT is a big deal.
This post has gone way beyond what I originally intended to say, and instead of rescuing it, I'm just going to kill it now.
"The image is a dream. The beauty is real. Can you see the difference?" -- Richard Bach, Illusions
That you were running a single thread computationally-intensive task as a benchmark.
Dual CPUs are there to help parallism. They won't show great increases on pure number-crunching. For example, my previous machine was a dual-533 Celeron, and it would be nice and responsive whilst running multiple apps, even if one of them (say, my MP3 encoder) decided to max out one of the CPUs.
Cheers,
Ian
c't magazine
translation of a short except: even early prototypes of amd opteron can win over all competition in four ways systems - either 32 or 64 bit - at the sap sd benchmark. and that with only 1.6 ghz (planned to launch at 2 ghz)
i think the chart says it all. go amd!
"It's such a fine line between stupid and clever" -- David St. Hubbins, Spinal Tap
For a good analysis, read this article over at Ars. In particular, it does point out that the likely cause of slowdowns in some apps is down to cache contention. Near the end, it also says:
In short, sometimes it helps, sometimes it hinders.Finally, you don't need multithreaded apps to take advantage of SMP/HT; if you're running a cpu intensive application on one CPU, the other is free for interactive stuff. You do, however, get much more benefit in a multi-threaded application.
What is this, AMD cheerleader day or something? Ok, understand that new technology, espically something as big a change as the Itanium takes TIME to develop. The Itainum is NOT for desktops right now, and if you think that's who Intel is targeting, you have a poor understanding of business.
The Itainum 1 was mainly a research chip, a first generation to let people start to develop and test on real hardware. MS took advantage of this and rolled out an IA-64 version of Windows. Intel was hoping for some server sales, but the real goal was getting the new IA-64 system into production silicon.
The Itainum 2 is a much more practical chip. It is something that peopel will probably seriously look at for high end server as it is competitive with 64-bit chips from Sun. You may see it in a few workstation, but probably not many, it's mostly a server chip. Remember, we are taling competiton with big iron here, not desktop system.
Now, as time goes on, the technology will become much more mature and cheap and will eventually filter into desktops. Hopefully, that will happen before we start to hit the 32-bit crunch.
The idea here is not wait until the last second for people to need a 64-bit chip, but to get it to market sooner so you can start working on it.
This, by the way, is not the first tiem Intel has done something like this. The Pentium Pro was blasted when it came out because it's 16-bit performance sucked. Sure it did great for 32-bit but who teh hell used that? Well tehn along came Windows 95 and teh PPro architecture was refined into the PII and it was a great chip since 32-bit was rising rapidly and it smoked at that. The P3 is the third incarnation of the PPro architecture. It's optimised and enhanced (ala SSE) but the same fundimental architecture. The P4 is the first brand new architecture since the P3.
The Itainum is a much larger change than the P4 since it is not only going to 64-bit but a new ISA (EPIC instead of CISC). It needs time and testing before it will be real.
However, Intel is certianly NOT ignoring the home market. The P4 is going to continue to be refined (we are on the 3rd revision of P4s and a 4th is soon comming) and should scale up to around 10ghz. There is plenty of life left in it (and probably subsequent chips based on its architecture). Then, by the time it is getting ready to be replaced, the then current Itanium chip should be ready for prime time.
So quit your bitching. If you don't want a P4, fine, stick with a P3. Why the hell do you care WHAT Intel is doing if you don't want a new chip? When you do decide you want one, get a P4, you ahve no lack of options with them and they scale to rather high speeds already and are not stopping.
from what i have read from now, it seems that some readers are looking at the itanium 2 as a chip for consumer. in this case, it might never be. some comments just seem to be flamebait.
:)
the itanium series is designed with special applications in mind including scientific work and datamining applications. keep in mind that 9mb of cache may be too big for the typical application but for those high end where you would want to let say analyize an entire database and get statistics to determine trends, then you might want to think again. faring the cpu even with a higher clock rate but with a small cache won't keep up with the competition.
i would be pleased to see an amd opteron chip with at least 3mb cache in the market (maybe i can think about getting one of them.)
with competition, i believe there are just three right now, with ibm's power, and sun's ultrasparc to make the rest. this is for the high end arena.
and of course, the processor is just a variable to the equation. in the enterprise arena, you must need a good platform. that is it should be very scalable (with hundrends of processors in a system and upgradability) and reliable (with 99.999% uptime and hot swap components including cpu, memory, i/o cards, etc.). intel has good tools and partners for these and amd will take some time to catch up (but i believe they would.)
intel has some good plans for itanium including the dual-core cpu and even the same pin compatibility (although it doesn't mean it can be fitted into the old ones.) the thing is, intel is already gearing a battle in the enterprise arena. with its resources, it will be able to deliver quite better products in the future.
i believe intel has lots of technologies lying around that we do not even know. of course, currently, you will not put all your cards. wait for some threat and put it down one by one.
with the latest results, intel is doing well financially compared to a greater loss for amd. their new hammer line will be a saving factor for them (question still to be answered this year - and i'm excited about this.) and i'm sure intel already has a pentium 4 running at 5ghz lurking around their labs. they are just waiting for the new processor before we start a new ghz revolution.
Live your life each day as if it was your last.
> Wow - way to make up stuff.
:(
> The Itanium II is certainly not a dud
Agreed. From an engineering standpoint, it's quite a nice chip. I don't agree with some philosophical stuff in the ISA (I'm not that much of a VLIW-for-general-purpose fan, but hey), but the microarchitecture and implementation seems very nice. I do wish that it was easier to implement OOE on IPF, though.
> x86 compatibility is worthless ina high end 64-bit machine, somethin AMD doesn't seem to grasp.
> They're marketing a high end technology (consumers and normal business users don't need
> 64-bit technology and won't for a while) to the mainstream market. Morons.
Feh. A big "screw you" on that. AMD isn't catering to the high end server group. They obviously can't just teleport into that market. Their catering to the smaller business that uses Xeon servers. Backwards compatibility with x86 is of the utmost importance in this market. Basically, they're marketing x86 workstations and x86 servers that happen to allow you to enhance performance of some types of programs with simple recompilations. There is a good chance that I might get the lower end version of this product when it comes out, as I use Linux, which may strongly benefit from those extra registers in x86-64, on my home machine. We'll have to see, of course, before I pull out the green.
> And you seem to be ignoring the numbers (remember that 'reality' the rest of us
> live in matters to us, if not to you). AMD is going broke. Intel isn't.
That's a bad measure to use. You don't have any controls in this analysis. There are a lot of reasons why AMD is losing money (poor management a la Hector Ruiz, inability for a relatively small company to handle a very harsh recession, etc..), and there are a lot of reasons why Intel is still doing phenomenally (people buy Intel no matter what, currently excellent execution, they can afford to strongly diversify). Many of these reasons have nothing to do with the technical/engineering side of the equation. IMHO, both AMD and Intel have incredible engineers, and frankly AMD especially warrants respect for being able to ramp technology at *approximately* the same rate as Intel despite having a very, very miniscule fraction of their resources. That is why I was a big AMD fan a couple years ago, at around the time when the company was dominated by the excellent triumverate of Sanders, Raza, Meyer as well as a couple critical folks like Norbert Juffa and Paul Hsieh. At this point in time, AMD was a quantum of a company that somehow managed to produce a piece of engineering that allowed them to, for a brief time, outdo the capabilities of a company fifty times their size. I am somewhat dismayed that AMD turned into a more traditional company over the last two years or so.
-JC
Don't think of hyperthreading as an alternative to dual-core or 2-way SMP. It isn't. Think of hyperthreading as a way of squeaking out more usage from the existing execution units in your CPU.
Lets say you have a hypothetical CPU with n execution units. (For simplicity, we won't distinguish between types of execution unit, such as integer, floating point, branch, etc).
You fetch and decode a bunch of instructions, and then issue them n-at-a-time to these execution units, for maximum performance.
But, the instruciton stream has some inherent limitations on which instructions can be issued concurrently, due to dependencies between instructions, instruction type mix mismatching available execution unit type mix, and instructions waiting on loads, etc. Even with control and data speculation, there may be fewer than n instructions READY to issue on the next clock cycle.
So, you have three choices:
1. Just issue the ready instructions, and let the other execution units go to waste.
2. Switch to another thread, maybe it has n instructions READY to run. (This is usually called on-chip multithreading).
3. Issue a mix of READY instructions, some from one thread, and some from another thread, which combined together use all n execution units. Both threads get to make some forward progress, and no execution units are "wasted". (This is usually called on-chip hyperthreading).
So, back to the big picture: Hyperthreading isn't a replacement for a second CPU or core, because it does not provide any more computation resources. It's a way of using the available resources in a CPU more efficiently, so that fewer computation units are likely to go to waste on any given clock cycle.
A dual core chip typically duplicates almost ALL the circuitry on the chip, often even including the caches. Big chips have low yields and cost a lot. Dual core is a way of throwing a lot of money at getting more parallelism. Kind of like having multiple CPUs in separate sockets, but with both advantages and disadvantages coming from the closer coupling. Hyperthreading is a way of throwing far less money at the problem of squeaking out some of the wasted performance in an existing CPU design.
It isn't free, by the way. Hyperthreaded CPUs do have to duplicate some hardware on a per-thread basis. Obviously, thread context registers like program counter and stack pointer have to be duplicated, as do application registers. But they share caches, execution units, decoders, memory management units (mostly), bus interface logic, etc.
Hope this paints a clear picture.