New AMD Athlon 2600 Processor Released

Ertai writes "Looks like the latest AMD processor is out today, and is taking it right to Intel! Running at 2.13 GHz, the Athlon XP 2600+ is reviewed at Amdmb.com. The benchmarks show that the new Athlon on a 'revision B' Thoroughbred core with the frequency increase is able to beat out the Pentium 4 2.53 GHz processor on almost every test. Not only that, but it is a good overclocker as well! Check it out." AMD's press release on the topic also notes a Athlon 2400 was released as well.

Comments

Re:How well will the Athlon scale?

[Jugalator]

Well, it's not too long until AMD starts using their new Clawhammer architecture with the introductory processor being similar to a 3400+. And then it's a whole new ball game when it comes to scaling.

Re:Intel has to shaking now

[Hard_Code]

Not to mention, that if you look at the comparison here you see that AMD is kicking out a chip which as 1/3 fewer transistors, and just over half the size of Intel's P4, yet is faster. I understand Intel is supposedly gearing up for higher clock speeds or multiprocessing or what have your, but still it is impressive that AMD is besting Intel with a chip which has 1/3 less transistors and is half as big. They must be doing something right.

RE: What is quantispeed?

[Vengie]

The basis for all of Athlon's "better than intel" claims.

Nine-issue, superscalar, fully pipelined microarchitecture: Provides more pathways to feed application instructions into the execution engines of the core, allowing the processor to complete more work in a given clock cycle (high IPC). The delicate balance between the depth of the pathways and clock speed of the processor produces high levels of performance.

Superscalar, fully pipelined Floating Point Unit (FPU): Completes more floating point operations per clock cycle than competitive x86 processors and permits high operating frequencies. The end result is a processor with the computing power to tackle the most computation-intensive software applications.

Hardware data prefetch: Prefetches data from system memory to the processor's Level 1 cache, which reduces the time it takes to feed the processor critical data, increasing work throughput and therefore overall performance.

Exclusive and speculative Translation Look-aside Buffers (TLBs): Keep the maps to critical data close to the processor, which helps prevent the processor from stalling or waiting when future data is requested. These TLB structures are now larger, exclusive between caches, and speculative. Larger TLB's give the AMD Athlon XP processor access to additional data maps. Exclusivity removes the duplication of information, freeing up more space in the Level 2 cache for other useful data to be used by the processor. And the speculative nature of these structures allows the processor to generate future maps of critical data quickly.

(from)
http://athlonxp.amd.com/overview/quantiSpeedArchit ecture.jsp

Anyone who has taken an OS course (ugh, NachOS) knows the pains of TLB management -- I really would like to see the 'voodoo constants' they used. (Background: the clock-hand approximation of LRU is one of those "voodoo constants" -- most of physics is filled with "voodoo constants" -- likewise...programming an OS is filled with them. If you've ever looked at SPRITE and LFS, the i/o data burst rate suggests that the time-slice should be ~8 seconds -- etc etc. I'd _really_ like to know AMD's voodoo constants.) =)

Re:Intel has to shaking now

[jpmorgan]

'I will answer you from the mouth of my canon' - Le Marquis de Montecalm to General James Wolfe

Intel releases their 2.8ghz chip next week. The speed battle trundles on.

Of course, it's getting boring now; Intel has mostly reclaimed the performance crown from the usurper AMD, and the Athlon core is showing its age, and AMD is facing new problems with clockspeed. The previous Thoroughbred core had problems increasing clock speed due to signal propagation issues and AMD had to add another metal layer to optimise the wiring layout. Think about it, at 2 billion clock cycles every second, there isn't a lot of time for a signal to get from one side of the CPU to the other, and unlike NetBurst(the P4 core) the Athlon core really isn't designed to take signal propagation issues into account - hence why Intel is rapidly scaling up the clock speed while AMD is struggling (I calculate the P4 can probably hit at least 7-8ghz before signal propagation becomes a problem again). AMD has managed to stretch out the life of the Athlon core, but I'm not sure what they're going to do next; signal propagation speed isn't easy to solve without a complete redesign of the core. Although the TB can hit 2.4ghz, beyond that I can't see the processor continuing to increase in speed. For AMD's sake, I hope that lasts them until they can get Hammer based chips on the market.

Still, it doesn't keep me up at night. Intel is ramping up clock speed as regularly as clockwork, AMD is keeping up (for now). Wake me up when something interesting happens.

Actually, I am looking forward to the 3ghz since I've heard, well, rumours that Intel is enabling SMT on it. Finally, an innovation! Seriously though, SMT is pretty cool, it gives the processor the ability to run two threads at once. The main thread is slightly slower than it would be if the chip didn't do SMT (a couple of percentage points), but the CPU can use its unused resources to run a second thread at about 15-20% of full speed.

So look back in December/January. Intel's releasing SMT chip, and AMD might be releasing Hammers. Until then the Intel vs. AMD battle will continue to be boring!

Re:AMD vs Intel -- Only we lose

[jpmorgan]

You might want to look at Serial ATA, then. And yes, Intel is one of the designers of the spec.

CPU performance will be a factor again within a couple of years. Software developers just have to get used to the headroom and realise the true implications of what they can do now. I'm working on some software right now (planning to release it under a BSD license now, but I have plans for a commercial release at some point in the future) which would heavily tax a modern CPU. And yes, it actually provides some *gasp* value. And no, it has nothing to do with video editing. :P

Be patient! We'll find something to do with your excess clock cycles soon enough.

Re:Soggy Chips?

[Brian Stretch]

AMD Saxony Operations Unaffected by Dresden Flooding

DRESDEN, GERMANY -- August 15, 2002 --AMD (NYSE: AMD) said today that its AMD Saxony operations located in Dresden, Germany - including production at the facility's Fab 30 plant - continue to operate normally despite severe flooding across Germany's Saxony region.
"Although much of the larger Dresden area is being affected by unprecedented floods, our production is running according to plan and employee morale remains high," said Hans Deppe, vice president and general manager of AMD Saxony. "Because of the preventive controls built-in to our facility and the exemplary dedication of our workforce, we expect to continue to operate normally despite the conditions."

AMD Saxony has its own on-site power plant, and remains accessible via the Dresden airport and federal highways. AMD Saxony's operations, including Fab 30, are located high up on the rim of the river valley and have not been directly affected by the flooding in other parts of Dresden and surrounding areas. The company does not expect that operations will be impacted even if the local flood situation worsens.

Re:Intel has to shaking now

[Sabalon]

Look at the Lightwave and 3dsmax benchmarks, where code optimized for the P4 kicks AMD's butt.

How likely is it that Visual Studio .NET 2 or whatever is P4 optimized? How likely is it that Adobe and other big players start optimizing their code for the P4?

Depending on how you feel about the above questions is how much it matters what AMD can do with less transistors.

Then again, cost is also a issue, and Intel just can't win there.

New scientific dating technique

[Sloppy]

2.8Ghz...my first computer didn't have that many MHz

Log2 of 1000 is about 10, times 1.5 years ("Moore generation"), results in 15 years. Ergo, qurob's first computer was likely more than 15 years ago.

Re:Alternative reviews...

[cheezedawg]

Reading the tomshardware review, I don't see this as a big advantage for AMD.

- The p4 2.5 GHz beat the Athlon XP 2600 in over half of the tests
- The fastest P4 is cheaper than the new Athlon 2600 (???- when is the last time we saw that?), and that is before Intel's price cuts they announced for later this month
- The new AMD 2600 won't be available to customers for another month or so
- Intel is releasing the 2.8 GHz P4 next week

Why Not Push Multiprocessors?

[Blahbbs]

Why the push for MHz and not multiprocessor systems? You'd think that AMD/Intel would like that solution because it would mean people would buy more than one processor from them at a time. Seems it would give their R&D folks a bit more breathing room, too.

I could see holding back on multiproc systems when the big manufacturers were preloading Win98/ME, but doesn't Win2kPro and WinXP support multiproc systems? I, for one, will likely make my next PC a multiproc machine.

Re:Someone remind me why we really care anymore..

[Anonymous DWord]

How about a hot-swapped hard drive? 100 gig drives are a hundred bucks - you could get one for each day of the week. Get two sets, so you have last week's stuff too.