hyperthreading

[Steevo]

I chose a pantium 3.4Ghz HT single core for our server at work due to the cost(I didn't pay for it so who cares0, reliability, and the folding power. Mostly the folding power with a couple gigs of RAM.




But honestly, it kicks the shit out of most gaming rigs I have used, and my own, unless you actually start a game, then, mebey not so much.

 

[AshenSugar]

from what my buddy who works for intel says the shared cache is good for some stuff, like if one core only needs xxxk cache but the other cores working on something heavy needing xxxxk then it can scale, intels dual core arent platform native, 775 wasnt designed for muticore orignaly, intel only stated doing multi core because they where loosing sales to amd because of dual core, same with 64bit intel only went 64bit because they lost sales by not having it(couldnt let amd look like they had a leg up u know) em64t is a partial implementation of amd64(read amd white papers/data sheets then compare with intels em64t)

from what i understand we wont see a native multicore setup till the next new cores come out from intel, by then amd's k8l native quadcores will have been on the market a good while..

its a constant back and forth, amd gets the upper hand, intel catches up/slitely surpasses them in benching, amd comes back with another change, intel was working on a quad"core" p4 design, but it saturated the FSB and bottlenecked the system hehe, by quad core i mean 4 cores on 1 cpu package not 1 solid peice of sli.

im glad they finnly moved back to the p6 cores insted of nutburst, at least in a few years i wont see anymore nutburst systems needing fixed......i hate how slow they feel after being a long time amd guy.

 

[Steevo]

The list of the first supported 775's were IIRC Prescott HT chips.




Like the one I got 2 and some years ago.

 

[Wile E]

intel was working on a quad"core" p4 design, but it saturated the FSB and bottlenecked the system hehe, by quad core i mean 4 cores on 1 cpu package not 1 solid peice of sli.

Their current quad core isn't suffering from that malady tho, that leads me to believe that there has to be some sort of cross core communication.

 

[AshenSugar]

I loved the days of my P4northwood clocking from 2.6 - 3.6ghz and beating the pants off of all of my buddies normal athlons (3000+) ( cause around that time that is what they were meant to compete with.... and we all know that p4's vs the 64s wasnt a fair fight ). And even now with my own x2 4200+ system oc'd to 5000+ ish speeds, that old northwood still is very snappy! I sold the entire system to a friend who is still using it with a 7800gt agp w/ 1gb of ram and the thing runs great. In fact I wish I hadn't sold that northwood setup, not just cause it was my first build , but would have been a very good second pc to have around. Of course your AMD boxes will perform better, at least in gaming, cause thats really where HT has no effect, but for everyday usage and multitasking, it worked, IMO, better than the athlons of the day. Not to mention, cooler. Then came prescott....

first you must be talking a stock 3000+ because my 2600+@2300mhz was faster then an overclocked p4 at 3.x for gaming and genral apps(u know those things that most ppl use and that are not multi threded), also your assurtion about athlons running hotter then p4 makes me laugh.
tbird and spitfire where no hotter in my exp then any p4 of the day, infact the spitfire(duron) ran VERY cool even overclocked from 650 to 864(100fsb to 133, max my board could do)
palomino was a HOT chip, so was tbred-a, but tbred-b/barton/thorton/applebread ran very cool at stock speeds, hell my old applebread system was running at 2.8gz with an old dragon orb(thermaltake) i had used on a socket7 system. my buddy still uses taht box to game, sure its not as fast as my a64's but its a good bit faster the this 3gz p4 next to me for most stuff(encoding is where it falls flat on its face )

and prescott, oh lord, dont get me started, we used one in a small board to brew coffie at a shop i worked for!!!!!(ran the coffie water thru a high end water block) that damn thing was so hot u could feel the heat a foot or more away, the thing was a show peice the owner made, it was litterly an internet coffiepot

oh and most SMP games seem to dissable SMP mode on ht systems, quake4 from something i was reading acctualy has a special code path to keep perf from being hurt by HT's fake 2nd cpu.

 

[Deleted member 3]

from what my buddy who works for intel says the shared cache is good for some stuff, like if one core only needs xxxk cache but the other cores working on something heavy needing xxxxk then it can scale, intels dual core arent platform native, 775 wasnt designed for muticore orignaly, intel only stated doing multi core because they where loosing sales to amd because of dual core, same with 64bit intel only went 64bit because they lost sales by not having it(couldnt let amd look like they had a leg up u know) em64t is a partial implementation of amd64(read amd white papers/data sheets then compare with intels em64t)

from what i understand we wont see a native multicore setup till the next new cores come out from intel, by then amd's k8l native quadcores will have been on the market a good while..

its a constant back and forth, amd gets the upper hand, intel catches up/slitely surpasses them in benching, amd comes back with another change, intel was working on a quad"core" p4 design, but it saturated the FSB and bottlenecked the system hehe, by quad core i mean 4 cores on 1 cpu package not 1 solid peice of sli.

im glad they finnly moved back to the p6 cores insted of nutburst, at least in a few years i wont see anymore nutburst systems needing fixed......i hate how slow they feel after being a long time amd guy.

It's not really a constant back and forth since AMD has only had 1 series of chips which actually managed to outperform the Intel chips. (Athlon XP performed similar, Athlon 64 took a lead) And Core 2 Duo is actually a native dual core chip, unlike Pentium D which are 2 Prescotts glued together. However, Kentsfield and Clovertown are again 2 C2D's glued together. Can't argue about that working out well though. Looks a lot better than relabeling workstation chips and glueing chipsets as well.

As for the 64 bit issue, Intel has had its own 64 bit instruction set (besides IA64). However Intel said the customer does not need it yet and therefor never enabled it. Apparently the instruction set is baked into some P4 cores (so they say). The thing is AMD released their 64 bit set first and hyped it so Intel had to follow. Microsoft however would not make yet another version of Windows for Intel and Intel was forced to adapt to AMD's instruction set. That is a major part of what went wrong with the 64 bit story. And when considering that, Intel did a very impressive job by adapting their cores within a year. And as we all know, nobody actually uses 64 bit, so Intel was right about the customer not needing 64 bit.

Your hate towards "the fake 2nd CPU" isn't that logical either, it's not SMP, it's SMT, do some reading on that yourself. High end specialized chips like Niagara and Power5 use similar technologies, you really want to argue about SMT being a stupid move since AMD is about the only company NOT using it?

And finally about your hate towards the hothead Prescott, I would say that is yet another feature Intel copied from AMD *coughThunderbirdcough*

 

[AshenSugar]

Their current quad core isn't suffering from that malady tho, that leads me to believe that there has to be some sort of cross core communication.

diffrent design, p4 need MASSIVE bandwith to keep perf from being to sucky, hence the hugely high clocks, very diffrent designs, the conroe cores are more like p3 or athlon then they are like p4/nutburst.

core2/conroe chips do more work per clock and have shorter pipelines, intel basickly took a step back to take a huge step forword, a good move, one they should have made back when they dumped socket 423!!!!!!

read some articals about how netburst works, there is a reasion its despised by those who disslike stupidity and ineffincy!!!!!

 

[Steevo]

read some articles about how netburst works, there is a reason its despised by those who dislike stupidity and inefficiency!!!!!(spell checked)

It performs quite well in looping tasks such as F@H and the like. The main reason that Conroe performs as well as it does is due to higher clock rates on the memory and better cache management-prefetch logic.

 

[Wile E]

diffrent design, p4 need MASSIVE bandwith to keep perf from being to sucky, hence the hugely high clocks, very diffrent designs, the conroe cores are more like p3 or athlon then they are like p4/nutburst.

core2/conroe chips do more work per clock and have shorter pipelines, intel basickly took a step back to take a huge step forword, a good move, one they should have made back when they dumped socket 423!!!!!!

read some articals about how netburst works, there is a reasion its despised by those who disslike stupidity and ineffincy!!!!!

I understand the inefficient design of netburst, I can't stand them either. I hated my Northwood, overclocked from 2 to 2.4, it still felt slow. My X2 destroyed it in single threaded video encoding at stock speeds. But how does the netburst architecture relate to or effect bus saturation? (Honest question, btw. Just want to figure this out)

 

[AshenSugar]

It's not really a constant back and forth since AMD has only had 1 series of chips which actually managed to outperform the Intel chips. (Athlon XP performed similar, Athlon 64 took a lead) And Core 2 Duo is actually a native dual core chip, unlike Pentium D which are 2 Prescotts glued together. However, Kentsfield and Clovertown are again 2 C2D's glued together. Can't argue about that working out well though. Looks a lot better than relabeling workstation chips and glueing chipsets as well.

As for the 64 bit issue, Intel has had its own 64 bit instruction set (besides IA64). However Intel said the customer does not need it yet and therefor never enabled it. Apparently the instruction set is baked into some P4 cores (so they say). The thing is AMD released their 64 bit set first and hyped it so Intel had to follow. Microsoft however would not make yet another version of Windows for Intel and Intel was forced to adapt to AMD's instruction set. That is a major part of what went wrong with the 64 bit story. And when considering that, Intel did a very impressive job by adapting their cores within a year. And as we all know, nobody actually uses 64 bit, so Intel was right about the customer not needing 64 bit.

Your hate towards "the fake 2nd CPU" isn't that logical either, it's not SMP, it's SMT, do some reading on that yourself. High end specialized chips like Niagara use similar technologies, you really want to argue about SMT being a stupid move since AMD is about the only company NOT using it?

And finally about your hate towards the hothead Prescott, I would say that is yet another feature Intel copied from AMD *coughThunderbirdcough*

acctualy the biggest thing to go wrong with 64bit was IA64, intium/itanic was/is a nasty design, only good for things like HUGE database servers!!!!!

and intel clamed that they had working 64bit instruction sets, but from what i have been told and read it was dissabled because IT DOSNT WORK RIGHT!!!!!! and it was first introduced in prescott cores.

and people do use 64bit, just not windows workstation users, well other then ppl who need massive ram for stuff like rendering anyway.
i have seen alot of linux servers using 64bit os and 8gb or more ram(and taking advantage of it to) 64bit windows fell on its face because MS didnt push it, they didnt really want another windows between normal xp and vista, driver support was pathetic to saythe least.

from the things i have read about intels 64bit x86 its been abandoned for workstations(may still be emplimented for servers in the future, just not for windows servers since ms isnt gonna support it)

as a friend of mine says regularly, 64bit x86 is a comprimise, its not the best solution for 64bit, but its the only way to keep people happy during the move to 64bit, mainly because MS wouldnt support a cpu that didnt give them dirrect lagacy support for dos/win9x apps, again this is a failing of the market and MS, they REFUSE to dump lagacy 16bit app support!!!! (good part of why windows is as buggy and bloated as it is)

amd would have liked to gone pure 64bit, as would intel, but ms is the market domonator for pc's, as such they gotta do what ms is gonna support.

this is also why EFI isnt catching on in pc's, ms wont support it, they dumped support planned for VISTA, I acctualy like most of the ideas behind EFI vs current bios, but for it to catch on you need to get the market leader to support it.

i know we are taking this topic way off its orignal point, but the orignal point is mute now because the poster got his answer!!!!

amd and intel need to do something they have never been able to do b4, work as a team, design a new standred for cpu's on the future, something they can both build their own versions of like x86 has been, then FORCE ms to support it.
and they could do it, because ms couldnt survive with just VIA making x86 cpu's!!!!!

it probbly wont happen because intel has never been able to play well with others, look at their development platform(for hardware) you gotta kiss intell arse to be allowed to make a chipset/board for their chips, on the other hand amd lets anybody develop a platform for their chips, just a totaly diffrent platform concept.

intel wants to have their asses kissed by everybody, amd just wants to be loved

 

[AshenSugar]

I understand the inefficient design of netburst, I can't stand them either. I hated my Northwood, overclocked from 2 to 2.4, it still felt slow. My X2 destroyed it in single threaded video encoding at stock speeds. But how does the netburst architecture relate to or effect bus saturation? (Honest question, btw. Just want to figure this out)

ok p4 dose less work per clock, but does it at high clocks, do it needs a constant feed of data, hence the use of RDRAM on older p4's, it has MASSIVE bandwith but poor latancy, only after the advent of dual chanil ddr was intel able to ween them selves away from rdram, because dual ch ddr was somewhat close to rdram in perf( but for servers/high end workstations u still wanted rdram) since then they moved to ddr2, high latancy but massive bandwith, the problem comes in when you put multi p4 cores on a buss designed for 1 core, where they are FORCED to share the limmited bandwith of that 800mhz buss not only for talking to eachother but for communicating with ram and system devices.

the analigy my buddy likes to use is think of the p4 like a big articulated buss, it can carry alot of people/data at once down a very nerrow highway with very few turnoffs, its not very agile, and if it crashes/has a flat/breaksdown it takes a long time to move it out of the way(cache misses/branch miss predictions), these busses are also moving at very high speed, now think about the highway being designed for 1 large articuated buss now having 4 of them on it trying to go both dirrections at once, you may start to see the problem now.

now think of athlons buss as a bunch of vans moving down a very wide road with lots of turnoffs, if one crashes(cache misses/branch miss predictions) it can be flushed off the road and replaced pretty quickly unlike trying to move a large buss off a narrow but very long road(netburst buss).

now also remmber that athlon64 and netburst have one VERY large diffrance, k8 has an onboard memory controler, so the highway isnt saturated with memory communication and communication with the memory is very fast/low latnacy, netburst use the chipset and FSB(the highway) to carry data to and from ram to each core and to let each core talk, as well as letting them talk between cores.

humm think of the intel chipset based memory controler on a netburst system as an old skool train switch house, its got limmited tracks and has to prioritize stuff the best it can to keep data flowing from the cpu to ram, cpu to cpu(core to core), from devices to cpu/ram and cpu/ram to devices, thats alot going on over a very nerrow road.

amd's chips take the memory communication out of the equasion, the chipset becomes just a bunch of devices and a way to communicate with devices over a very bi dirrectional highway(1000mhz each way)

its hard to explain, if jake or teddy where around i would have them explain it, they are better at explaining it in ways other ppl can easly understand.

 

[Deleted member 3]

Thats of course a bunch of sillyness.

I would suggest reading up on IA64 and perhaps you'll notice how it's far from dead. In fact, it's gaining popularity.

64-bit isn't required to use 8GB of RAM, PAE support in the application is all that is required.

Intels 64bit for desktops has never "officially" existed.

If AMD wanted pure 64bit they could have just copied IA64, since IA64 was originally meant as replacement for x86 in the early 90's.

EFI is supported by Vista as far as I know. If not, you can always get a Mac, they use it already.

Intel never had a reason to work with other companies on developing platforms. They own the x86 platform, they've been the biggest player in the market by far (AMD hasn't been much until the last generation of chips). Same with Microsoft, they often don't have to listen to the rest of the market since they control it.

 

[Wile E]

Nope, MS took EFI support out. They hinted at a possible later implementation, but no confirmation, as of yet.

 

[AshenSugar]

i know about ia64, and its not very popular, i have seen a few itanic servers, they are only good in slect applications, systems using intium chips respond slowly to commands from consol or from gui(depneding on the unix variant your using) compared to their x86/x64 equivlants.
its why super computer makers didnt/dont use them to build super comps, its why most server companys dont even offer them, not even by special order.

they have a small niche and are filling it, on the other hand opteron sold more units in its first day on the market then intel sold intiums for all the years it was on the market!!!!

and i remmber intel making a statment that prescott had 64bit but it was dissabled because it wasnt needed by anybody, and then it was leeked that it also didnt work properly, thus amd beet them to market with a 64bit x86 design that worked, MS chose AMD because they offered the best option at the time, they had a working design when intel didnt, and MS hasnt supported IA64 in any os i have seen, and i have seen some non x86 windows nt versions.

for example, i have seen the dec alpha version of windows 2000(on a dual hp alpha workstation), i have seen nt4 and 2k on sparc systems, hell i have setup nt4 on powerpc systems(was better then mac os7....hehe) ms dumped official support for all but x86, but made and updated HP versions of 2k for the alpha after hp bought the alpha out.

whats weird is that i have never seen an IA64 compatable version of windows, even datacenter, and thats one place you would think u would see it since Intium would be good for datacenter use......

and vista dosnt support EFI, thats why u cant run it dirrectly as the main os on a mac, because ms dumped support for it, mac users where up in arms, ms got alot of hate mail from people who bought intel macs then found out they wouldnt beable to run vista native they would have to use a new version of bootcamp!!!!!!!! ( i got a laugh listening to mac users bitch that they wouldnt beable to run windows native on their mac.......after these same people blame windows for every bad thing in the pc world)

and as to the way intel/ms act, well unlike ms, intels been slaped around a bit by somebody who had a better mouse trap, nobodys come out with something that can run windows apps and games as well or better then windows and be as easy to use.
and if intel and amd designed a new prosessor architecture(sp) then informed MS that they had XX ammount of time b4 they had to eather support it or not have any chips to run their software on, ms would fold like a cheap shirt, because they wouldnt have a choice, they couldnt survive on via and nvidia cpus(nvidia is working on an x86 cpu now u know.....bet its gonna suck lol)

 

[AshenSugar]

oh and AFIK amd cant copy IA64 its not part of the cross licencing deal they made with ms, also they wouldnt want to, due to how poorly its looked apon by companys. (i remmber the first intium servers that u had to flash a special bios to because if u didnt they could loose data and crash, most ppl i knew at the time sent the servers back for a new version!!!!)

 

[Wile E]

whats weird is that i have never seen an IA64 compatable version of windows, even datacenter, and thats one place you would think u would see it since Intium would be good for datacenter use......

There are IA64 versions of Server 2003. But by and large, I agree. IA64 is an inferior architecture, easily outperformed by AMD64.

 

[AshenSugar]

there afterall is a reasion that in 1 day opteron sold more chips then intium did in all the years it had been out!!!!!

never seen server 03 for ia64, seen a version of solaris for it tho, ran about 1/4 the speed the same pc version did on my old test box(p3,768mb ram,12gb bigfoot hdd) the gui was just sluggish, and the server responce time to basick calls was slow, it was fast at database searches tho....LoL.

when i was in school the skill center had a company donate a first gen intium that was less then a year old, they gave us the system and a bunch of stuff with it, the teacher tryed and tryed to find a use for it, file server, domain server, he tryed IA64 versions of netware,unix,linux and all he could say in the end was, "i dont understand why its so slow" from what i been told its still on a shelf there with its screensaver going and nobodys tryed to find a use for it since, personaly i wanted the hdd's out of it, 6 15k rpm seagate drives!!!!!!

 

[Wile E]

I ran across IA64 03 when I was considering buying 03 to run a file server in my house. Ultimately decided against it tho. Couldn't justify the expense.

 

[AshenSugar]

file server, windows 2000, domain server/active dirrectory server 2003(easyer to setup)

03 also makes a great workstation ^^ its all i run, corse i got a copy of each 03 version from ms via vouchers their reps gave us when i was over at my buddys old job at a college(he was the IT manager effectivly, his boss sighned off on whatever he wanted done, hell the boss told him to throw a couple extra benq flatpanils on the order as a bonous!!!!!)

 

[Deleted member 3]

http://www.microsoft.com/windowsserver2003/64bit/ipf/default.mspx

As for IA64 being inferior to AMD64 is like saying a DAF is inferior to a Ferrari. They're for completely different markets. In fact, if you talk about the architecture IA64 is everything x86 is not. It does not have all the problems of backward compatibility. (it can run x86 code emulated though)


If you would bother to read the Supercomputer top 500 you would notice there being quite some Itaniums in the list. In fact, the Itanium has been hard to come by for a long time because most produced Itaniums ended up in supercomputers early on.

Fact is, Itanium became a very specialized platform in which it is a very mighty one. And it's sales are still growing, even though not as fast as initially planned.

 

[Wile E]

file server, windows 2000, domain server/active dirrectory server 2003(easyer to setup)

03 also makes a great workstation ^^ its all i run, corse i got a copy of each 03 version from ms via vouchers their reps gave us when i was over at my buddys old job at a college(he was the IT manager effectivly, his boss sighned off on whatever he wanted done, hell the boss told him to throw a couple extra benq flatpanils on the order as a bonous!!!!!)

You can have him send 1 or 2 of those panels my way. lol.

Off Topic: How does 03 run with games and video encoding programs? How does it compare in resource usage(mostly RAM) vs XP?

 

[Deleted member 3]

http://www.sgi.com/pdfs/3905.pdf
Opteron vs the inferior Itanium 2 by SGI.

 

[AshenSugar]

You can have him send 1 or 2 of those panels my way. lol.

Off Topic: How does 03 run with games and video encoding programs? How does it compare in resource usage(mostly RAM) vs XP?

i will tell you what i tell everybody who asks, its everthing xp isnt, light, fast, stable, low on bloat, its just a joy to use, snappyer then xp, never crashes unless I CAUSE IT by installing something like a driver i know may be problematic(beta apps that use drivers and such, only happened 2times so far, and i fixed the cause myself)

on install 2003 has allmost nothing excess installed/enabled, if ur gonna use it as a server you set it up as a server using eather their wizzard/guided method or manualy.

if you want to use it as a workstation, well then just follow any one of the guides to setting windows 2003 up as a workstation(google windows 2003 workstation guide) or do what i started doing use nlite to costomize your install so it has ur drivers and prefered default settings for windows pre-enabled on install(saves you about 5-15min depending on how fast you can change all the settings manualy to dissable stuff like shutdown event tracker.)

for gaming u just gotta enable video acceleration(set to low by default for best server stab) takes a few seconds, my total tweak to gaming time after install is like 30seconds, if you want help trying it out gime a holler i have a few things that speed the prosess up

as to SGI, last i heard hey where in trouble, possably going under, and i said intium not intium2, eather way tho, they have sold far less units then server opteron's, and i would guess they even have sold less then sparc/ultra sparc in the same ammount of time!!!!

sparc is a good design, RISC multicore, VERY responcive, very powerfull, oh yeah and the ultrasparc is true 64bit, as it should be!!!
http://en.wikipedia.org/wiki/SPARC
incase you dont know what sparc is, after all unless u have delt with it personaly you probbly dont know about it, i had an old sparc system(it died to an ef gf who couldnt understand not to drink around my computer stuff, she got drunk and fell on it and crused the case(no joke) never could get it to work after that )

i was playing with an ultra sparc laptop a that a friend has(used for developing software and scripts for sparc servers) the damn thing for 6gb ram in a laptop, running solaris10 with java desktop, it just FLYS!!!!!

as to your "ia64 can run x86 emulated" yeah at what 1/1000th real time speed?, the demo i saw of intium running x86 software was so slow it was like being back on a 486sx25mhz with 4mb ram and windows 95(HELLLL)

and to the top500, 2nd is an opteron, the cray redstorm
"The new No. 2 systems is Sandia National Laboratories’ Cray Red Storm supercomputer, only the second system ever to be recorded to exceed the 100 Tflops/s mark with 101.4 Tflops/s." most are Power Architecture based but your vaunted Itanic2....i mean Itanium 2's best showing is 7th place, behind even the xeon!!!!

"The NovaScale 5160 system built by the French company Bull and installed at France’s Commissariat a l'Energie Atomique (CEA) slipped to No. 7, despite a new Linpack measurement of 52.84 Tflops/s. The original system debuted in the No. 5 position in June."
thats your vaunted Itanium 2 in action hehe humm 52.84 tflops/s vs 101.4 tflops/s, i will take the opteron

i seem to see more opteron super computers on the top 10 then itanium2's, strange,since according to you its the better chip.........
4 Itanium2 in the top 100.....
16 opteron in he top 100........

4 vs 16 why do you think that is, if as you say ia64 and Itanium2 are better then opteron/amd64 why are there only 4 super computers using them on the top 100 vs 16 opteron systems?

http://www.top500.org/list/2006/11/100
count for your self if you like

 

[AshenSugar]

 

[Deleted member 3]

First of all, if you're such a Sparc lover you must have noticed how I mentioned the Niagara? Obviously I know what a Sparc is.

Back to the Itanium issue the 486 performance is an old story, the code has been updated years ago and it performs just as fast as a Xeon at the same clock. The Itanium has only existed for a year or so, it was replaced by the Itanium 2 really fast. Plus if I can't name Itanium 2 as part of the Itanium discussion I would say you should compare Itanium with AMDs server platform in 2001. As AMD "fanboi" you obviously know the MP and MPX were a big joke.



I as naming the supercomputer list to explain that there are plenty of Itanium clusters on this planet. I wasn't saying they are #1 performance wise, in fact you can't even compare the systems in that list since they are all totally different. When a system has 4096 processors it's not surprising that it outperforms a system with "just" 512 processors. That's how large the differences are between systems in that list.

I never said IA64 (Itanium IS IA64) is fast at everything, I clearly stated that it is a very specific chip, not some all rounder.
I only stated that IA64 is superior to X86, in fact X86 is about the crappest platform there is.

The reason there are more Opterons in the list is because Opterons are cheaper, readily available and there is more software for them.
If you check http://www.top500.org/stats/28 you will see that there are only 3× as much Opteron systems, however Opteron sales vs Itanium sales shows a much larger difference. In fact, guess which platform stands above the Opteron? Exactly. Intel's outdated platform, where are the Athlon MPs? Half the list is dominated by Intel, must hurt for your pro AMD attitude.