AMD Opteron "Hammer" Preview

Melvin Tong writes "Hardware Extreme has posted a preview of AMD's 8th-generation processor that AMD is currently developing with a few exclusive pics of the mechanical sample. AMD Athlon processors based on Hammer technology are expected to ship in the forth quater of 2002. The preview is located over at HW Extreme."

Comments

AMD Opteron + Linux

[Strom Thurmond (R-SC]

It's a pretty safe bet such systems are already in use in some TLAs, we just don't know about it. Opteron/Hammer will be a nice step forward but obviously it's not meant as competition to big iron from Sun and others for quite a while, if ever. AMD has years and years to go before they can enter such markets successfully, they are relatively small and have to concentrate on the desktop and small to mid sized servers for the
forseeable future, plus have yet to prove themselves in the higher end, attract the appropriate support and build an image they don't yet have when compared to Intel, Sun or IBM.

It also looks like they have their work cut out for them already, with not as fast clock/rampup on 0.13 micron as expected and a tight line to get the Hammer line done properly as they are pretty strung out on cash compared to Intel, while the latter seems to have no trouble in increasing clock all the time (by throwing huge gobs of money at it of course).

Intel

[T-Kir]

Either way, It would be funny if Intel ended up having to license AMD's x86-64 technology. Even though I don't think that would happen, I suspect Intel would rather fork the 64bit platform with their Itanic (part 2) than license from AMD... but you never know!

Re:Intel

[Shadow99_1]

Actually if you remember back a year or twop ago you'd realize they already have a cross-licensing deal with AMD which would entitle them to use x86-64 if desired without further hassle... Of course Intel may prefer you forget that til they need to use it...

Nice cap!

[(H)elix1]

I was very happy to see the nickel cap on their new CPU. After crushing a couple of AMD chips, I became very weary of removing the heat sink after a successful mounting. More so than I probably should be, but after chipping the edge off of some $100+ CPU's, I was very nervous about picking up any of the cutting edge processors.

I look forward to lapping the cap to a shinny mirror finish!

Re:Nice cap!

[Artifex]

After crushing a couple of AMD chips, I became very weary of removing the heat sink after a successful mounting.

No doubt. Actually, that worked to my advantage, when I was trying to get Fry's to take back an Athlon XP that had gone bad... when they told me they had to test it, I was worried, because their idea of a testbed is another customer's board hooked up to crappy "PC Doctor" software, and has rarely caught transient errors in the past.

Wouldn't you know it, though, they cracked it during mounting, so of course it became "oops, let's get you credit for that chip" instead of "we can't find a problem in 30 minutes of running crappy test software so it must not be bad."

Notice the spread eagle in...

[MoThugz]

this picture here... on the left hand side... eh wait a minute. Sorry, just a bent pin.

Longevity of CPU w/ integrated memory controller

[wowbagger]

I have to wonder about the lifespan of a CPU that has an integrated memory controller of any type - not just DDR, but RDRAM, or FOORAM, or NARFRAM. What happens to the family when a new RAM interface comes along?

Now, for high-integration CPUs designed for embedded style apps I can see it, but for a main-line CPU it seems to me that tying the memory controller to the CPU limits the lifespan of the design.

I realize that should POITRAM become the new speed king that the RAM controller block of the CPU can be redesigned, and I understand that putting the RAM controller in the chip can increase the memory bandwidth to the CPU.

But it does cause me to think....

AMD Hammer FAQ

[antdude]

AMDZone wrote a FAQ which was a good read.

Chip comparison?

[vga_init]

How are the x86 family processors comparing to other chip architectures these days? Are other chips (ie SPARC, PowerPC) still superior, or is the x86 catching up? "Go intel or go home"? That's catchy, but my quote is far superior: "f*** intel." ;)

Re:Clearly a reprint of AMD Marketing material

[shird]

AMD's 64-bit processors extend their long, rich history of semiconductor solutions based on customer-centric innovations

Emphasis mine. Looks like they've gone and changed it now. Must be readers of slashdot.

I hope...

[ParisTG]

I hope they dont ship them like this! (Note the bent pins on the left corner :))

Re:Benchmarks on OpenSSL

[Anonymous Coward]

As a side bonus, you can find SPEC benchmarks for Itanium and Itanium IIs on that chart

I submitted a story when the results for Itanium 2 were first made available, but I guess it was rejected. Instead, we get lots of fluff pieces about Opteron.

The misinformation here is rather ridiculous. When it comes to the topic of IA-64 vs. x86-64, Slashdotters crawl out from under their bridge to slam the IA-64 often because of its "technical inferiority."

Whatever. If you want to argue non-technical issues such as compatibility, marketing, pricing, etc., you might be able to make some points, but based on performance alone, IA-64 has shown some stellar SPEC results with Itanium 2.

It currently has the highest SPEC FP results for a single CPU system. It's significantly better than a Pentium 4 2.53GHz system and even outperforms IBM's 1.3GHz POWER4. And this is all at a mere 1.0GHz. I'd say Itanium 2 is pretty darn good at getting a lot of work done on a per-cycle basis, wouldn't you?

Then why does misinformation persist here? Do Slashdotters want to remain oblivious to the facts?

Someone a while back was looking for a workstation suitable for intensive floating point calculations. Someone else responded and recommended HP's zx-series of Itanium 2 workstations. But yet another poster got modded higher when he recommended a $2 million IBM POWER4 system claiming it would offer better performance than an Itanium 2. "Where are the benchmarks?" he asked. Go to the SPEC site, and see for yourself.

Obviously some 16-way $2 million monstrosity from IBM is going to perform better when doing massively parallel calculations, but try finding a POWER4 desktop system. You can't.

Of course, SPEC marks don't necessarily translate to real world performance, but they are respected benchmarks and it's all we've got to compare right now.

POWER4 seems to be a pet favorite around here for some unknown reason. Only IBM sells them. Intel is a chip maker, not a system vendor, so if IA-64 succeeds, you'll be able to choose vendors. You won't get that freedom with POWER4, UltraSPARC, or PA-RISC.

Alpha was in a similar situation, but it failed, so clearly the technically superior solution might end up being a disaster.

This is where the real trouble lies and you can debate away on this issue. But when it comes to performance, don't be quick to slam Itanium 2.

IBM and AMD get praised for their efforts to help out the open source community, but it's not like Intel isn't doing anything. They maintain an open source IA-64 research compiler on SourceForge (ORC), have plenty of open source libraries at their site, offer free hard copies of their documentation (these are very well written references), and I think they may even assist developers who are porting software to their platform.

I don't know about the rest of you so called "geeks", but I like hearing about interesting processor technology -- so could these Opteron fanboy posts be balanced out in the future with some more information on the competition?

Benchmarks

[decefett]

The machine was running Mandrake Linux, kernel 2.4.18-24mdk, and identified itself as running at 797.7 MHz with 256k of cache.
...
And here's a comparison, openssl 0.9.6b (as shipped with Redhat 7.3) running on a 400 MHz

What was that about lies, damned lies and...

Re:Benchmarks on OpenSSL

[Zathrus]

First off, those are stellar benchs for the Itanium2. And I agree that this was a total fluff piece for Opteron -- as others have pointed out, it's nothing but rehashed info from other sites and AMD press releases.

I'm not a POWER4 advocate either -- the chip may be cleaner than x86 (wow, that's not hard), but proprietary is proprietary.

Back to Hammer vs Itanium though. I am much, much more excited about the pending Hammer/Opteron release than I am about Itanium2 or McKinley or whatever. Why? Because Hammer is made for consumer systems. Itanium (w/ or w/o the "2") is still priced somewhere in the stratosphere and it's performance on desktop systems is abysmal. Sure, the SPEC numbers are pretty, but there's no software out there, the compilers continue to suck, and I don't expect either situation to improve anytime soon. When Hammer comes out there will be a plethora of software that will already run (and probably run faster, even in 32-bit mode) and compiling 64-bit apps will be relatively straight forward. VLIW is a nifty idea, but we're nowhere close to optimizing code perfectly now. Adding on the additional layer of VLIW makes the problem even worse.

High end computing has always been a totally different realm from desktop computing anyway. I don't really expect the Hammer/Opteron to compete in that realm -- it's too limited by the load of crap that comes with x86. But it's a far better future desktop computing solution than anything Intel has to offer thus far, and that's why you see so many people excited about it.