Whats the deal?

[spike spiegal]

I want to thank you guys for buying into AMD's hype on the whole Opteron vs X2/64 issue. I have this visual in my head of Oompa Loompa's in clean suits "cherry picking" X2/64 chips with a big flashing sing that says "GOOD" and "BAD", and sorting accordingly.

While there are structural differences between the Opteron and X2/64, those differences do not manifest themselves on a typical workstation class machine. There are likely at most to only make a difference with 4/8 way machines, or, with cache sensitive benchmarks assuming the Opteron is running with more than 1meg.

To imply that Opteron's are superior than X2s/64's is implying that X2s/64s are defective in some way, which is frankly ludicrous. Processors are also don't give a sh^t if they are used in server vs workstation roles because the designation is vague at best. I've yet to see even the highest demand server I've deployed over the past 10 years peg a steady 100% processor utilization unless an application was broke while your typical game does this all the time. So, what system is getting higher CPU loads again? I guess that extra hyper transport link is going to make my RAID 5 faster.

I have a couple X2 based servers on the floor right now, and went that route over the Opteron because the VAR was giving me good deals on the X2. Given a choice between an X2 and dual core Opteron for the same price, hell yeah I'll take the Opteron. If the X2 is cheaper though, I'll take it because other than a cache difference it doesn't make a difference unless I have 4/8 of them talking to each other in a VMware box. The way you guys are talking if I use an X2 in a server my SQL and mail databases will soon get corrupted.

Also note that the P4 is quite popular and sold in "Server Class" machines all the time. But, according to you guys AMD's quality control isn't good enough and only the Opteron should be used in a server.

Yeah, you can overclock an Opteron more than an X2/64. Big deal. Since I don't drive a skateboard to my fast food job and put LEDs fans in my machines because I use them to perform actual work on, I don't over-clock.

AMD tried this same marketing junk with the Athlon MP and all it lead to was a lot of people hacking the L5 bridge on Athlon XP's and noting how they ran fine in dual processor rigs -vs- the over-priced MP. The Opteron vs X2/64 debate is almost just as political, but if it makes you feel cool saying you're running a 'server class processor (whatever the hell that is) in a gaming rig, then good for you.

 

[Bobthelost]

Originally posted by: spike spiegal
Yeah, you can overclock an Opteron more than an X2/64. Big deal. Since I don't drive a skateboard to my fast food job and put LEDs fans in my machines because I use them to perform actual work on, I don't over-clock.

You have completely missed the point of the replies given.

It has been consistently stated that the advantage of the Opteron ranges is the overclocking ability. The obvious implication is that they are no better if there is no overclocking in use. So if you don't overclock then don't bother getting an opteron.

Dual core CPUs are liked because of the performance increases you get with them, in applications like DVD transcoding, number crunching and heavy multitasking. In gaming the trend is for SMP games that can take of the additional CPU core. Not all, or even most will use both at this time.

Short short version:

Opterons are for overclocking and for servers.

Dual core is for heavy multitasking and some tasks which in the future will include gaming (i belive that oblivion will make use of mulitple core CPUs)

Oh and by the way:

I want to thank you guys for buying into AMD's hype on the whole Opteron vs X2/64 issue. I have this visual in my head of Oompa Loompa's in clean suits "cherry picking" X2/64 chips with a big flashing sing that says "GOOD" and "BAD", and sorting accordingly.

You're an idiot, please book a vasectomy as soon as possible to prevent further pollution of the gene pool.

 

[sammy1234]

spike spiegal is right to an extent. You think there is some sort of conspiracy surrounding chip manufacturing it seems though. All of you do. We'll I regret to inform you there is no conspiracy. The cherry picking procedure mentioned occurs, although not in the same way as the experts here like to believe.

First I'll start with the opteron. The opteron 939 socket processors can work in groups together to create a cohesive environment. The reason for this is the architecture. Opterons are not at all cherry picked A64's. The are made on a completely separate die. This is why the L2 cache is so big. You can?t make a processor change after the die has been cast, all of its instructions are hard wired, except power consumption and other external factors. On board memory and communication between systems is another piece of the pie entirely.

I thought all the brilliant hardware engineers on here would know that. Then again what hardware engineer would hang out with a band of geeks that think over clocking is a good idea.

Here is a link to the processor facts so you can read up.

http://www.amd.com/us-en/Processors/Pro...30_118_8825,00.html/40211A_Opteron.jpg

The A64 is actually the more powerful of the two processors. They are peak performance regulated because of the error threshold that is incurred during manufacturing. For example you may buy a processor that was intended to run at 2.8Ghz in a perfect world, but because of error prone manufacturing techniques the chip can only be rated to 2.2Ghz, this is the cherry picking that occurs, and maybe the reason you believe that opterons are A64's.

I'm pretty much done here you guy's should consider reading more often, or maybe take some hardware engineering or software classes. The mindless guesswork that happens here is ridiculous.

 

[sammy1234]

Oh yah and I forgot to mention, I live in the heart of Intel land, Beaverton Oregon. Most of my friends are hardware engineers at Intel, I work for a little company too maybe you've heard of it, INFOCUS I am a hardware engineer. What I would like to tell you all is this. Dual Core is not like a car engine dingbat, it is exactly what it sounds like two completely separate processors on the same mold. Dual North bridge chips due to be released soon will fully take advantage of both cores by pre-stripping code. As info comes down the pike it will be pre-processed then split, the split calcs will be performed by the main processors then the code will be reassembled and wham! You will be working with two processors, at equal speeds. Only vista can take advantage of this type of processing though sorry guy's.

 

[Matthias99]

Originally posted by: sammy1234
First I'll start with the opteron. The opteron 939 socket processors can work in groups together to create a cohesive environment. The reason for this is the architecture. Opterons are not at all cherry picked A64's. The are made on a completely separate die.

You mean the S940 Opterons, not the S939 ones. These are a different die/core, as S939 chips don't have the coherent HyperTransport links needed for multiprocessor synchronization.

The S939 Opterons are, for all intents and purposes, identical to the other S939 chips with 1MB caches.

This is why the L2 cache is so big. You can?t make a processor change after the die has been cast, all of its instructions are hard wired, except power consumption and other external factors. On board memory and communication between systems is another piece of the pie entirely.

Unless, of course, you have a way to disable parts of the cache at the hardware level -- for instance, if there is a defect in part of it.

I thought all the brilliant hardware engineers on here would know that. Then again what hardware engineer would hang out with a band of geeks that think over clocking is a good idea.

:roll:

I'm pretty much done here you guy's should consider reading more often, or maybe take some hardware engineering or software classes. The mindless guesswork that happens here is ridiculous.

Perhaps you should consider having a clue.

Most of my friends are hardware engineers at Intel,

Perhaps you should ask them to fill you in on some CPU basics.

I work for a little company too maybe you've heard of it, INFOCUS I am a hardware engineer

Then I would think you might know more about hardware.

Dual North bridge chips due to be released soon will fully take advantage of both cores by pre-stripping code. As info comes down the pike it will be pre-processed then split, the split calcs will be performed by the main processor's then the code will be reassembled and wham! you will be working with two processors, at equal speeds. only vista can take advantage of this type of processing though sorry guy's.

This will never be as efficient as true multithreaded code, which already works just fine on Win2K/XP-based kernels.

 

[sammy1234]

You mean the S940 Opterons

No I mean the 939 opteron processors. I am talking about multiprcessor chips. Try placing an 2 X2's on an opteron board an watch closely as nothing happens.

Unless, of course, you have a way to disable parts of the cache at the hardware level -- for instance, if there is a defect in part of it.

Its funny you think you know so much. In the chip manufacturing world when an unrecoverable error occurs in the manufacturing process the chips are tossed and recycled through the whole process again. Nobody repairs or edits chips after they have been produced, the cost is to high.

Perhaps you should consider having a clue.

what are you five...six maybe?

Perhaps you should ask them to fill you in on some CPU basics.

Quite often they do help me understand CPU baisics, but most of the time we just B.S. about other stuff.

Then I would think you might know more about hardware

It seems I do.

This will never be as efficient as true multithreaded code, which already works just fine on Win2K/XP-based kernels.

Not true. XP is 32 bit, this limmited code set does not allow more complex procedures such as true multithreaded processing, Remeber just because something is supported doesn't mean it works right.

 

[sammy1234]

AMD Athlon XP Barton 2500+ 1836MHz @ 2200MHz

This is your processor? How can you claim to have any expeirience when your sporting that.

 

[imported_michaelpatrick33]

LOL. This thread is hilarious. Mathias99 I bow to you in your infinite patience to answer these ... observations by the OP.

I am interested Sammy how will Vista take advantage of AMD's memory controller that is integrated along with the crossbar and hypertransport glueless interconnects on dual and multicore systems? Those dual FSB's will be interesting on AMD's ... err wait doesn't AMD have integrated memory controllers on their cpus? HMMM. Yes sarcastic.

You really think the AMD dualcores are just two cores sitting side by side like Intel's Prescotts dualcores?

I guess Dell/Alienware selling factory overclocked systems is also devoid of hardware and software engineers and designers because why would they sell factory warranteed overclocked parts :roll:.

Oh please Sammy explain to me how 32bit processors/Windows XP in any way limit the multi-threaded capabilities of the code being run. Are you talking about the 4gig memory limit due to the 32bit address space? My God, you can't be serious.

How do you explain the video encoding programs performance increasing 50-90% on dualcore systems running 32bit XP? Of course you won't get 100% increase in performance on dualcore but what you are saying doesn't make sense to me. Explain further please.

 

[Buck Naked]

hey Matthias99, just drop it. The question in the OP has been answered. Sammy is just trolling now.

 

[sammy1234]

AMD's memory controller that is integrated along with the crossbar and hypertransport glueless interconnects

Who's the idiot?

Oh I get it your trying to mull me into a saying something stupid, right? Go Mathias, Go Mathias, Go mathias, yeah! Wow. Glue? I don't think anybody is glueing processors together. And yes dual FSB isn't it amazing, oh wait you have no creativity, so if it's not physical you cant believe it. What the hell cross bar are you referring to?

 

[Luckyboy1]

I just got off the phone with wasting the time of Joe at www.magnum-pc.com and he says the Opteron is just an almost exact version of the other chip in question here, only with 1 MB of cache. AS a result, no, they're not run on the same die. The design with the exception of the cache thingy is however, remarkably the same and running with more cache makes for heating concerns, so quality of the final part is paid attention to, but not really any differently than the 64's, so in the end, cherry picking is a matter of production run luck and not a favoring of one over the other.

All his personal tinkering has confirmed the same/same nature in actual game performance and that quite often, the Opterons clocked farther than the 64's. He's also confirmed same/same with both the AMD and the Intel Reps.


Hhhhmmm, where have I heard all this before?

 

[sammy1234]

Good luck with your forum guy's. Looks like all the talent is to buisy using there computers for fun stuff.

 

[Buck Naked]

Originally posted by: sammy1234
Good luck with your forum guy's. Looks like all the talent is to buisy using there computers for fun stuff.

dont let the door hit ya...

 

[morkman100]

Originally posted by: sammy1234
morkman100 you are a jerk.

You seem to think you know everything.

You are the only one who thiks that the overclock is not possible. Just because you have no understanding of the english language and cant seem to read an article or thread fully, doesn't mean anything. you can take comments as they are offer advice, present factual information or statistics and give possitive comments or stay out of the frey. You just sound like an jerk who needs a friend. I feel sorry for you.

whhaaaaaaaa!

I guess you forgot to dispute anything I've said in the last post... oh well. have a good one.

 

[morkman100]

Its funny you think you know so much. In the chip manufacturing world when an unrecoverable error occurs in the manufacturing process the chips are tossed and recycled through the whole process again. Nobody repairs or edits chips after they have been produced, the cost is to high.

From Wikipedia...
"Fifth-generation Athlon Barton-core processors released in early 2003 featured PR ratings of 2500+, 2600+, 2800+, 3000+, and 3200+. While not faster than Thoroughbred-core processors in clock speed, they earned their higher PR-rating-per-clock by featuring an additional 256 KB of full-speed on-chip level 2 cache, for a total of 512 KB, and a faster FSB. The Thorton core was a variant of the Barton with half of the L2 cache disabled and thus functionally identical to the Thoroughbred B core. The disabled L2 cache on some Thortons was partially defective, but on others it could be re-enabled through bridge modifications."

Quite often they do help me understand CPU baisics, but most of the time we just B.S. about other stuff.

Let me guess... Dungeons & Dragons?

What the hell cross bar are you referring to?

From www.devx.com/amd/Article/28976...

"It's best to consider a dual-core processor as an upgrade for a single-core processor, because you get twice as many threads being processed. A single-processor system with a dual-core chip can handle two hardware threads?just like a dual-processor single-core chip. Also, the efficient design of the internal crossbar switch in AMD's Opteron design means that if two threads need to talk to each other, they can do so inside the silicon, therefore data can be exchanged very rapidly. "

Get over yourself. This is a discussion board. Not a "Sammy is right and everyone else is a moron" board. If you post info here that's questionable (i.e. a 3400+ 754 overclocked to 3.2), be prepared to have people question it.

Have a good one.

 

[Gagan]

I personally think people get opterons over the x2 to say "opteron"

"Hay dude wut u got in ur pc?"
"I got an AWPTERAWN"
"O RLY"
"YA RLY"
"RLY?"
"SRSLY"

That's my take on it.

 

[Matthias99]

Originally posted by: sammy1234

You mean the S940 Opterons

No I mean the 939 opteron processors. I am talking about multiprcessor chips. Try placing an 2 X2's on an opteron board an watch closely as nothing happens.

You can also watch closely as nothing happens if you do that with S939 Opterons, or S940 1XX Opterons. Only S940 Opteron 2XX and 8XX CPUs have the extra coherent HyperTransport links needed for multi-CPU operation.

S939 Opterons are the same as the S939 Athlon64 X2s with 1MB of L2 cache. The S940 Opteron 1XX CPUs are functionally identical except for the memory controller, which supports registered memory.

Unless, of course, you have a way to disable parts of the cache at the hardware level -- for instance, if there is a defect in part of it.

Its funny you think you know so much. In the chip manufacturing world when an unrecoverable error occurs in the manufacturing process the chips are tossed and recycled through the whole process again. Nobody repairs or edits chips after they have been produced, the cost is to high.

Both Intel and AMD have had this capability for years (at least wrt cache). Modern GPUs can have physical pipelines disabled through firmware.

Maybe your friends at Intel can tell you about it if you ask nicely?

Then I would think you might know more about hardware

It seems I do.

Uh... except you don't.

This will never be as efficient as true multithreaded code, which already works just fine on Win2K/XP-based kernels.

Not true. XP is 32 bit, this limmited code set does not allow more complex procedures such as true multithreaded processing, Remeber just because something is supported doesn't mean it works right.

WTF are you blathering on about? An OS being 32- or 64-bit has nothing to do with whether or not it supports running multiple processes or multiple threads within a process concurrently. WinNT-based kernels and their derivatives handle running multiple processes/threads on multiple physical or virtual CPUs just fine.

 

[morkman100]

Maybe your friends at Intel can tell you about it if you ask nicely?

That might be asking a lot..

 

[imported_michaelpatrick33]

Originally posted by: morkman100

Its funny you think you know so much. In the chip manufacturing world when an unrecoverable error occurs in the manufacturing process the chips are tossed and recycled through the whole process again. Nobody repairs or edits chips after they have been produced, the cost is to high.

From Wikipedia...
"Fifth-generation Athlon Barton-core processors released in early 2003 featured PR ratings of 2500+, 2600+, 2800+, 3000+, and 3200+. While not faster than Thoroughbred-core processors in clock speed, they earned their higher PR-rating-per-clock by featuring an additional 256 KB of full-speed on-chip level 2 cache, for a total of 512 KB, and a faster FSB. The Thorton core was a variant of the Barton with half of the L2 cache disabled and thus functionally identical to the Thoroughbred B core. The disabled L2 cache on some Thortons was partially defective, but on others it could be re-enabled through bridge modifications."

Quite often they do help me understand CPU baisics, but most of the time we just B.S. about other stuff.

Let me guess... Dungeons & Dragons?

What the hell cross bar are you referring to?

From www.devx.com/amd/Article/28976...

"It's best to consider a dual-core processor as an upgrade for a single-core processor, because you get twice as many threads being processed. A single-processor system with a dual-core chip can handle two hardware threads?just like a dual-processor single-core chip. Also, the efficient design of the internal crossbar switch in AMD's Opteron design means that if two threads need to talk to each other, they can do so inside the silicon, therefore data can be exchanged very rapidly. "

Get over yourself. This is a discussion board. Not a "Sammy is right and everyone else is a moron" board. If you post info here that's questionable (i.e. a 3400+ 754 overclocked to 3.2), be prepared to have people question it.

Have a good one.

:laugh: I dare mention the crossbar and glueless hypertransport of the AMD 64/X2/Opteron families and you mock me. I guess since you have to ask what the "h!ll a crossbar" is in a dualcore processor you really shouldn't be commenting on AMD X2 or Opteron dualcore processors at all. If I was thirteen I guess I could say you have been pawned :roll: :laugh:

For the record Sammy I have a X2 4400+ with 1024KB L2 cache per core core (amazingly similar to the 175 Opteron :roll: ) so I am not a forum member with an Opteron for overclocking. I hit 2640 with my X2. Not great, not bad but good enough for multithreaded applications.

 

[OSX]

Originally posted by: sammy1234

You mean the S940 Opterons

No I mean the 939 opteron processors. I am talking about multiprcessor chips. Try placing an 2 X2's on an opteron board an watch closely as nothing happens.

Nothing will happen, since dual opty boards are socket 940 and X2's are 939. Tool.

I love how sammy thinks he's some kind of tech god because he had a decent overclock for a whole hour before the computer burned up. By that standard, I can overclock a P4 to one million trillion billion gigahertz, even though it only lasts 10^-100 seconds, I win, right?

 

[morkman100]

Originally posted by: michaelpatrick33
:laugh: I dare mention the crossbar and glueless hypertransport of the AMD 64/X2/Opteron families and you mock me. I guess since you have to ask what the "h!ll a crossbar" is in a dualcore processor you really shouldn't be commenting on AMD X2 or Opteron dualcore processors at all. If I was thirteen I guess I could say you have been pawned :roll: :laugh:

For the record Sammy I have a X2 4400+ with 1024KB L2 cache per core core (amazingly similar to the 175 Opteron :roll: ) so I am not a forum member with an Opteron for overclocking. I hit 2640 with my X2. Not great, not bad but good enough for multithreaded applications.

This was posted to me, not Sammy, but your thoughts are still true. i don't have an Opteron either... just a lowly Athlon64 2800+ 754 oc'ed to 2.2GHz (darn crappy ECS MB).

morkman

 

[imported_michaelpatrick33]

Originally posted by: morkman100

Originally posted by: michaelpatrick33
:laugh: I dare mention the crossbar and glueless hypertransport of the AMD 64/X2/Opteron families and you mock me. I guess since you have to ask what the "h!ll a crossbar" is in a dualcore processor you really shouldn't be commenting on AMD X2 or Opteron dualcore processors at all. If I was thirteen I guess I could say you have been pawned :roll: :laugh:

For the record Sammy I have a X2 4400+ with 1024KB L2 cache per core core (amazingly similar to the 175 Opteron :roll: ) so I am not a forum member with an Opteron for overclocking. I hit 2640 with my X2. Not great, not bad but good enough for multithreaded applications.

This was posted to me, not Sammy, but your thoughts are still true. i don't have an Opterson either... just a lowly Athlon64 2800+ 754 oc'ed to 2.2GHz (darn crappy ECS MB).

morkman

Originally posted by: sammy1234

AMD's memory controller that is integrated along with the crossbar and hypertransport glueless interconnects

Who's the idiot?

Oh I get it your trying to mull me into a saying something stupid, right? Go Mathias, Go Mathias, Go mathias, yeah! Wow. Glue? I don't think anybody is glueing processors together. And yes dual FSB isn't it amazing, oh wait you have no creativity, so if it's not physical you cant believe it. What the hell cross bar are you referring to?

No, actually the quote you have about the crossbar is from Sammy's reply directly below my post about glueless hypertransport and crossbar technology. He directly quoted me. :laugh: Either way this has been a fun thread to read as Sammy posts stuff as a "hardware engineer" that makes me cringe.

 

[morkman100]

Either way this has been a fun thread to read as Sammy posts stuff as a "hardware engineer" that makes me cringe.

Maybe he means that he designs hammers and screwdrivers and stuff like that...

 

[Zebo]

Boy this troll is getting quite the audience .. Verdicts still out as to who's dumber at this point.

 

[daballard]

Originally posted by: Matthias99

Both Intel and AMD have had this capability for years (at least wrt cache). Modern GPUs can have physical pipelines disabled through firmware.

Maybe your friends at Intel can tell you about it if you ask nicely?

They probably just work in the fab