Could it be? AMD chip possibly running at 5ghz and beyond!!

IBM's power6 65nm cpu has been supposedly going above 5ghz! The previous power5 core with 90nm using the same technology as amd couldn't get only half those speeds. According to overclockers.com, this could mean big news for amd, and that they might actually have a shot at beating conroe (amd's value chip @ 5ghz vs e6300 @ 3.5ghz, you do the math)

Here is the link:
http://www.overclockers.com/tips01085/

Any thoughts?

well see, if amd does go with it, they will aplly it to k8l, their best arch they are boasting about, the current 65nm will at best give a tiny bit more oc than the 90nm, they will just run cooler.

The thing is, amd has to find a way to keep their old performance per clock, but also find a way to get their cpus up to 5ghz to compete with intel. Even if the manage to get a cpu to run @ 2ghz that beats the e6600, if they can only get that cpu to 3/3.4ghz maz, they won't be able to compete with an extreme or e6700 running @ 5ghz. So I think amd really has no choice, but to go IBM's way. I think the main problem comes with ht instead of fsb. It allows performance per clock, but not high clocks unless they find a better arch that isn't solely based off ht.

Pentium d 9xx? No, they were killed even by the 939 athlons

I agree that we will have to wait a bit on this to truly see how things perform. If amd can get performance per clock back, then I am definately buying one of those at any cost. If I can replacing my 4000 with a cpu that runs @ 2ghz, but can out perform most conroes except the extreme and e6700 and run at 5ghz for only a few hundred, that's the chip for me

I think the information is unfounded and has little merit.

From an actual report where they tested it they had major problems overclocking the x2 4800 @ 2.5 ghz more than 2.838GHz at 1.5625V!

Look under the notes section

http://www.firingsquad.com/hardwar [...] page11.asp

No, you have it all wrong, of course the brisbanes are going to perform horribly, they are a new test subject for amd same way the preslers failed. I think that this technology will be probably applied to k8l, as there is no way they can revise the brisbanes in time for the actual release in stores

After reading the link I provided I don’t think this is a plug in for you my friend
I do remember reading about this now but didn’t pay attention to its function in the mainstream marketplace

No According to me this it's difficult for Amd reaching 5 ghz now as the frequency that a processor can reach is highly dependent on it's architecture , the higher the speed the longer the pipeline has to be and the longer the processor pipeline become the diffcult it is to control its efficiency.

Both the intel core and the current Amd k8 base micro architectures are nt designed to reach very high clock speed for a five ghz chip on will have to use an entirely different design approch i don't see atleast Amd reching 5 Ghz for now cause the k8l is just an incremental upgrade over the current k8 design.

If IBM announces such a chip the power 6 architecture must have been designed for that.

whatever, you guys seem to shut my idea down

All I thought is that if they can manage to perfect a ht based soi and get it to 5ghz, that's gotta be pretty damn fast

Besides, the fact they can do that with 36mb of lvl3 cache blew my mind

The power6 core cpus aren't out yet or anywhere near out, so I don't believe there are any benches, but it would probably out-perform both, because pretty only opterons double over as both server and super computers. This competes with itanium etc, so it wouldn't be a fair comparison

yeah, but the thing is the slower cache and such was actually designed to save money, so more cache could be added to the next crop of cpus if they had to in order to compete

Right now since 4x4 kinda just failed, nobody has really even heard of it anymore, amd is kinda bargaining that their k8l arch will make a difference and give them a gain over intel. So I guess your right that speed is all they have to turn to, because we all know that some idiots will buy a cpu that runs at 5ghz that is slower than a x6800 just because of the ghz, if you don't believe me, just look at the newegg pentium extreme 965 reviews, anything that happened around the core 2 launch shows how stupid people are and why the price is still higher than the x6800 8O

This is all hypothetical, amd might not even use the thechnology, not to mention, it won't be applied (if it does) until k8l at the earliest, and by then there will be programs for vista to take advantage of it

This is a terribly wrong assumption to thing 5Ghz from Amd. IBM's architecture is called a RISC processor, it can run a lot faster because it has no code optimization built into it's processor architecture. Which means a lot less transistors, a lot less power, and therefore it can go faster, but you may not see any performance difference between this and an x86 in some tests, because the x86 compensates for slower Ghz with an optimized architecture. RISC processors are based on the theory that most code isn't coded with the SSE optimizations, etc.

I thought by now people were over this BS. Power 5, which is an RISC CPU, runs at 167W @ 1.65GHz, and also has 389mm2 die. So much for cooler running, smaller RISC cores.

The REASON that IBM can do this while AMD can't for their marketplace is because the market that IBM Power is in is so aggressive and ASP is very high, and the designers put all they can in. See what the high end Power 4 became when it came to desktop. It became a much lesser CPU called a G5. Compared to Power 4, G5 has no performance, consumes no power, pea-sized die, and no cost.

You also can't have a 150W/200W CPU on a desktop or people will piss on it like Prescott, while the place Power 5/6 goes into, they don't care much.

Also, Power 6 has SAME PIPELINE as its predecessor, the Power 5, which is 16 stages. It achieves super-high clock speed by
1. 65nm process, it gives 30% higher clock speed boost at same power consumption
2. Highly-tuned circuit design, which on previous designs like Power 5, it was lot more automated
3. It's DUAL CORE, coming at 2007-2008, I guess they want to achieve both super high single thread performance, which automatically equals pretty good multi-thread performance, and dual core is easier with higher clock than quad core

It's an off-die L3 cache. I heard it runs pretty fast for its size, but it is off-die. On-die 36MB L3 would create a CPU with ridiculous die size, something like 600-700mm2(maybe more).

http://www.xbitlabs.com/news/coole [...] 75540.html

http://en.wikipedia.org/wiki/POWER4

http://en.wikipedia.org/wiki/Power5

http://en.wikipedia.org/wiki/POWER6

http://en.wikipedia.org/wiki/IBM_POWER

http://en.wikipedia.org/wiki/Power_Architecture

http://en.wikipedia.org/wiki/PowerPC

http://en.wikipedia.org/wiki/Vector_processor

http://en.wikipedia.org/wiki/IBM_ViVA

I think these will uncover more of the IBM plot.

Unlike (almost) everyone else (it appears) I've kept my eye on IBM.

Heat is only the enemy if you can not move it elsewhere at a very high speed.

ED: I'd also like to add that 5 GHz while keeping the same 16 stage (length) pipelines is a very good design move.

ED2: There are now several companies which have '32-bit PowerPC' licenses and '64-bit Power' licenses developing their own processors based on the 'Power Architecture' including Freescale, Tundra, AMCC, HCL, Culturecom, P.A. Semi, Xilinx, Microsoft, Rapport and Cray.

Tip: P.A. Semi is *THE* company to watch. - Word on the street is they can build a 4-6 core G5/Power5 based chip using well under 50 watts at an acceptable clock speed. (Imagine a hex-core G5 mobile in a laptop with 4-6 hours of battery life).

How did I make a" terribly wrong assumption" you said the same thing I did. My point was that AMD makes x86 cpus and can't use most of the optimizations that IBM can. Thats what the apples to oranges means.
What I want to see is non extension using math apps like PI, prime95 and how a desktop cpu holds up. The power6 would kill it but by how much on a price/performance level? This cpu should put up some good number just how good is still to be seen. It's a lot harder to sell hype to the IT pros then the guy next door.
It's weird when you think about the advances in cpu power and that no one wants to use it. Every one wants to off load the work on to a gpu or other hardware. Right now I'm running a P2P watching a AVI and posting and my cpu[ 939 3000@ 2.3] usage peeks at 5%.