Hammer Prototype Test

[funky-d-munky]

Dosn't seem to convincing to me. I compared the amd 2800+ and the 2.8 p4. The winners of each program follow.

Quake 3: 2.8 p4 and 2800+ virtual tie.
3D mark2001: 3.06 p4 and 2.8 p4 and 2800+ virtual tie
Commanche: 2.8 p4
Unreal 2003: 2800+
Lame 3.92 : 2800+
MP3 maker: p4 2.8
Divx: 2800+
Mpeg 2: 2800+
spec view 3D: 2800+ 10 fps faster then 2.8 p4 and 9.5 fps faster win ace: virtual tie
Light wave: p4 2.8
4d xl: 2800+ win
3d studio max: 2800+ win

Well they seem pretty even. The p4 3.06 w/ HT enabled is just slightly faster then the 2.8 p4 and often beat by the 2800+. I don't give a crap about Sandra and such as there were plenty of actual programs compared.
In an over all picture Intel is in no way far out in first. And as far a Video editing and the majority of 3d AMD seems to be the winner.

If you really believe the p4 3.06 is worth that much money based on those benchies, you have too much money. Hey I got some swamp land in Florida for sale ;>)
funky

[Brian Ellis]

Funky

To benefit from the P4 HT, the software MUST be written for it. Please tell us which of your tests had this HT-compatibility. HT-enabled may actually be slightly slower with non HT-software than with it disabled, as there may be a tendency to have parallel instructions calculated.

The important point for here is that MSP7 is written with HT instructions in the software and I'm told that it simply z-i-i-i-ps along (not being a billionaire, can't afford it to let you know myself, but I can say that MSP7 is fast on non-HT machines, as well). From what I've heard, you won't have too much money if you buy an HT for video editing.

Will you make a similar comparison with MSP7 when you can get your copy?

[Topha]

Originally posted by Brian Ellis
And where are the 64 bit O/S and software? 64 bit CPUs are next-to-useless without these, even if they can emulate 32 bit mode. And I mistrust emulation

Linux (at least suse) already has support for the hammer cpu's, and i think even microsoft announced a 64-bit windows xp. i think there is a 64-bit ut 2003 in the make.

as for the 2800+ being as fast as the 2,8Ghz P4, well thats why they called it 2800+ but thats in 32bit apps, and as soon as we will see 64 bit apps for consumer pc's, intel will have to do something about it

[Dr Mordrid]

Not to mention that if an app (MSPro7 for example) supports HT and/or SSE2 that speed advantage of the AthlonXP's disappears fast.

Dr. Mordrid

[Jerry Jones]

There's no question that hyperthreading has given the 3+ghz Intel chips a speed advantage for now.

But we've all seen the race flip-flop back and forth between AMD and Intel over the past two years and if anything emerges as 'true' it is that today's speed demon is tomorrow's dog.

A real question for any individual is whether it makes sense in this quantum flux environment to invest large sums of money toward the purchase of legacy 32-bit processors or not...

...even if a 64-bit Windows OS and a 64-bit MEDIASTUDIO PRO (or whatever) are not yet available, they will surely be available within the next 18 months...

:-)

I understand SSE2 is fully implemented in the Hammer chips - so AMD regains that speed advantage with Hammer.

In my opinion, I can't help but believe Intel's 32-bit hyperthreading advantage will be short-lived.

The AMD 64-bit technology seems to me - on paper anyway - to be better implemented than the Intel 64-bit (Itanium) technology.

On the other hand, if AMD prices the new 64-bit chips too high, then many might be reluctant to upgrade.

The Intel hyperthreading is interesting... but... in the end... a 32-bit chip is still just a 32-bit chip no matter how much faster it is than the current AMD XP2800+.

64-bit technology is pretty cool.

I can't wait to see the first Hammer motherboards.

Jerry Jones

http://www.jonesgroup.net

[Topha]

Hammer (or athlon 64) has sse2 on board. and as far as i heard they wont be priced much higher than current athlon xp's (but i guess they will be a little more pricey, not to mention mobo's etc.)

and also, i think i read somewhere that amd has a HT-link patent, and they are thinking about realeasing dual-core hammers once they move to 0.09 micron.

personally, im really exited about the hammer

[the maddman]

Originally posted by Brian Ellis
And where are the 64 bit O/S and software? 64 bit CPUs are next-to-useless without these, even if they can emulate 32 bit mode. And I mistrust emulation

AMD has demoed 64bit windows, and Linux will be ready to roll as soon as Hammer is out. And AMD's hammer doesn't emulate 32bit x86 like Intel's Itanium does, it has a true 32bit operation mode, just like a 386 has a 16bit operation mode. I'm just upset Hammer has been delayed so long, I want 64bit linux!

[Brian Ellis]

Let me be the devil's advocate, yet again.

"64-bit technology is pretty cool.". Please tell me why. If, for example, you are doing maths operations at 64 bits or complex matrix maths at a lower level, yes, I can see that there is an advantage. But typing an e-mail, using a word processor or editing videos are not like that (I might make an exception for a few, very few, video operations). Your data will not be 64 bits long: it's rarely more than 16 or 24. If you are doing a 16 bit operation, it ain't going to pass through the 64 bit CPU any faster for a given clock speed. OK, you might be able to fiddle the software to do two operations in parallel, but this is not going to increase the reliability.

IMO, the HT approach or, better, 2 32 bit CPUs will work faster for most non-mathematical applications, because, they can REALLY work reliably in parallel: they are designed for that. If you like, the HT is a 32 bit to which you can tack on a second 32 bits and 32 + 32 added up to 64 when I was at school.

My guess today is that 2 HTs or 4 ordinary P4s will give you a performance that would leave any 64 bit computer standing. I support this by the PCPro tests on a prototype 64 bit. OK, I agree the OS was not optimised and they had little software. On 32 bit emulation, they found it was actually slower than a Pentium of equal speed. On true 64, the testers were decidely underwhelmed. I agree that this may change but "64-bit technology is pretty cool.???"

[the maddman]

Originally posted by Brian Ellis
Let me be the devil's advocate, yet again.

"64-bit technology is pretty cool.". Please tell me why. If, for example, you are doing maths operations at 64 bits or complex matrix maths at a lower level, yes, I can see that there is an advantage. But typing an e-mail, using a word processor or editing videos are not like that (I might make an exception for a few, very few, video operations). Your data will not be 64 bits long: it's rarely more than 16 or 24. If you are doing a 16 bit operation, it ain't going to pass through the 64 bit CPU any faster for a given clock speed. OK, you might be able to fiddle the software to do two operations in parallel, but this is not going to increase the reliability.

IMO, the HT approach or, better, 2 32 bit CPUs will work faster for most non-mathematical applications, because, they can REALLY work reliably in parallel: they are designed for that. If you like, the HT is a 32 bit to which you can tack on a second 32 bits and 32 + 32 added up to 64 when I was at school.

My guess today is that 2 HTs or 4 ordinary P4s will give you a performance that would leave any 64 bit computer standing. I support this by the PCPro tests on a prototype 64 bit. OK, I agree the OS was not optimised and they had little software. On 32 bit emulation, they found it was actually slower than a Pentium of equal speed. On true 64, the testers were decidely underwhelmed. I agree that this may change but "64-bit technology is pretty cool.???"

First of, HyperThreading doesn't double the speed of the processor, it's just a fast way of switching threads to fill bubbles in the pipeline. The P4 and the Athlon already have more than one 32bit pipeline, and can sometimes execute more than one instruction at a time, Hyperthreading just increases those chances on some software. HyperThreading is NOT smp on a chip, 4 ordinary P4's is a lot more processor power than 2 HyperThreading P4's.

Hammer will be able to execture more than one 64bit op at a time as well, and AMD does have patents for multicore and multithreading too, so you may see something like Hyperthreading in a 64 bit chip as well.

But that's not where Hammer gets it's speed boost. It's the on die memory controller that's the big deal. You're dropping one hop from every read and write on the system, and adding NUMA like features for SMP. Hammer is gonna be fast, 64 bit is mainly needed because as you add processors to a hammer system, you add more channels of memory. A Quad Opteron (the server version of the chip) would have 16(!) 333mhz DDR channels, and you could easily stack in more than the 4GB of ram a 32bit address space gives you.

And what processor prototype did PCpro review? I bet it was Intel's 64bit chip, which is a very very different beast from anything you've ever used. It's not x86, so when you run x86 software on it, it gets very poor usage of the silicon there. We'll see how well Intel can pull it off, but with the Itanium running multiple thousands each, not many people buy them from running Word.

If you have any other questions about either Intel's or AMD's 64bit stuff, ask away. Personally, I want an Alpha EV8.... I got to touch a working CPU once, but they'll never see the life of day.

[Jerry Jones]

I think the following exchange was informative enough for the wary reader to answer the question regarding why 64-bit technology is cool:


Dave Haynie (who I do not consider to be a credible source of information) wrote:


"32-bit CPUs have been doing 64-bit floating point, and later, MMX, ops for quite some time now."


"Some ops, in SSE or AltiVec (on the PowerPC) happen with 128-bit registers, today, on 32-bit CPUs."


"So it's important to know what you're talking about."


Martin Atkinson-Barr (a PH.D. physicist & programmer & truly brilliant fellow) responded:


"It's true that x86 processors can move 64-bits, even 128 bits at a time using MMX or SSE but these are tough to use in general purpose routines because those registers are fixed at 64 and 128 bits respectively."


"Generally, in a string move (or compare or scan) you drop off the odd bytes, or words and do a REP MOVSD (also CMPSD, SCANSD) for the bulk of the transfer."


"Then you do a MOVSW and a mov for the remaining word or byte."


"That is a little difficult for an arbitrary string using MMX or SSE."


"Its not a problem using RAX in x86-64 since mostly the registers are addressable as 64, 32, 16 and 8-bits wide as RAX, EAX, AX, AL and so on for other GP regs."


"Having written some assembler MPEG2 routines I can assure you there is a BIG advantage to x86-64."


"If you want the source just e-mail me."


Dave Haynie:


"64-bit machines don't help much, if any, in string processing."


"In the simplest case, you're reading or writing a single 8-bit or 16-bit character, one-at-a-time."


"There's no need for larger registers, and in fact, unless you have special, ultra-CISCy string instructions (sure, the x86 has a few of these, don't know about x86-64) or some serious optimizations, that's precisely what the programmer does in their code."


Martin Atkinson-Barr response:


"No you generally work a row at a time, or a column at a time."


"In integer format these are often 128-bits wide as 8x16-bit integer."


"In my routines I convert these to floats for processing."


"Go to AMD's website and they will send you, for FREE, the manuals for x86-64."


"Great manuals too."


"I'm not sure what you mean by CISCy string instructions, I presume the REP type I mentioned above?"


"These are fully 64-bit in x86-64."


Haynie:


"Code isn't the whole story."


"For one, all of the modern desktop computers, x86 and PPC as well as MIPS or Alpha, have been running on 64-bit (or greater) buses for years."


"In fact, the Alpha went from a 128-bit conventional bus to the 64-bit, double-clocked EV6 bus that was adopted for the AMD Athlon."


Martin Atkinson-Barr:


"Well, there are busses and there are busses."


"The internal chip busses may be 128-bit or 256 bit wide."


"The external bus to memory is 64-bit currently in Athlons/P4s etc."


"Hypertransport allows busses to be up to 32-bits wide but implemented as LVDS (low voltage differential) and so the clock rate is MUCH higher, giving up to 12GBytes/sec - compare that to the memory interface of current x86 which is maxed out at about 600MBytes/sec - TWENTY times slower."


Haynie:


"All modern memory for the last 10+ years has been able to go faster, taking advantage of locality of reference in the memory system."


"This started back in the 80386/MC68030 days..."


"I designed my first '030 product back in 1988, pretty long ago."


"Anyway, when an Athlon reads or writes memory, it's optimized to run four-clock cycles, but each of these is composed of two data cycles of 64-bits each."


"So the usual memory access hits 64 * 2 * 4 = 512-bits of memory per full cycle."


"The Pentium 4 is optimized for running four-cycle cycles as well, but on its preferred bus, there are four data cycles per clock, so it's actually grabbing 1024-bit of memory per full cycle."


"Essentially, whether you grab that first character in a sting as a byte, word, longword, or quadword in software, the hardware's going to suck 512 or 1024 bits right into cache for you."


Martin Atkinson-Barr:


"I think your math is off here."


"Usual cycles are 64-bits wide even if the bus rate is running at 533MHz - actually 4x133 MHz, a quad pumped bus."


"One read is ~4 cycles so you get max 64-bits at 133MHz or a maximum of 1066MBytes/sec."


"Bus overhead drops this to a real, measured peak of 600MBytes/sec."


"This is the rate from main memory to cache, which is important for video."


"It ignores latency which is the killer (how fast to start each transfer, not the max when you get going)."


"The Hammer chips have an on-board memory controller which really drops latency."


Dave Haynie:


"No one should be hand coding for modern processors -- the compilers do it better."


Martin Atkinson-Barr:


"Not so, the critical parts are still hand-assembled."


"You get a factor of ~4 improvement by hand."


"Compilers spend a lot of time running back to memory for data and writing back variables, you can see the assembler code they generate easily."


"That puts pressure on the memory bus that is talked about above."


"Keeping data in registers eases the critical memory bottleneck that so affects video processing."


"It's one of the reasons why the Alpha does so well, because it has so many more registers."


"In my MPEG2 routines, all of the data is kept in registers throughout, something a compiler could never do."


Ulead came up during the discussion; Atkinson-Barr wrote the following


"I suggest that Ulead go with the x86-64 architecture and standardize on that."


"Nobody serious about video processing is going to want a 32-bit processor come next summer."


Jerry Jones

http://www.jonesgroup.net

[the maddman]

Originally posted by Jerrold Jones

Ulead came up during the discussion; Atkinson-Barr wrote the following


"I suggest that Ulead go with the x86-64 architecture and standardize on that."


"Nobody serious about video processing is going to want a 32-bit processor come next summer."


Jerry Jones
http://www.jonesgroup.net

Mediastudio Pro for x86-64 would be one thing to make me consider buying one more copy of windows... I'm on 2k for now, and wasn't planning on shelling out again, but that might convince me.

[funky-d-munky]

Brian-
"To benefit from the P4 HT, the software MUST be written for it. Please tell us which of your tests had this HT-compatibility"

These are not my benchies. They are just my breakdown of Elie's link to Toms Hardware. Elie provided this link to show Intels huge performance andvantage. I believe my breakdown shows otherwise.

New Athlon's will ship with 512kb L2. This will balance out the gains made by the P4 HT. They are virtually tied when using AMD's + rating. Except P4's need SSE2 and HT to be competitive.
funky

[KRSESQ]

Just out of curiosity, has anyone here actually tried clocking their XP xx00+ at the plus-rated speed? For example, how does an XP 2000+ run at 2000 MHz?

Kevin

[Elie]

I was specifically comparing Intel's latest chip with AMD's latest offering, in response to Jerrold's note regarding fastest chips for 2003.

Cache will certainly boost performance, but DO NOT under estimate HT, especially when running an application like MSPro 7.0!
As Doc said, it will render all existing single AMD solutions obsolete, however I still think my Dual AMD system will remain competitive as long as AMD keeps releasing MP versions of their XP chips.

Cheers,
Elie

[Hulk]

Lots of good points here.

Funky -

I agree with you on two important points:
1. AMD and Intel are very close in performance right now. Intel has a lead, but it is slight.
2. Synthetic benches are nice to look at, but the real deal is performance on actual software.

As far as purchasing a system for what it will do in a year or a year and a half, I think that's crazy. Technology moves too fast. I buy a system based on what it can do NOW!

Mark