What do you think of the X-bit Labs review of the opteron 144 w/ nforce3 pro???

[Duvie]

To start out here is the conclusion: (quoted from X-bit labs):

" Conclusion
So, we have just taken a close look at a uniprocessor workstation based on the new AMD Opteron processor and ASUS SK8N mainboard on NVIDIA nForce3 Pro 150 chipset. So far, we would like to refrain from wild enthusiasm. We have to admit that despite all the innovations, this workstation still has a lot of drawbacks, which come not only from the CPU, but also from the mainboard and the chipset. For instance, I was very upset to see no AGP Pro slot or 64bit PCI slot on the mainboard. As for the CPU, even the today’s latest Opteron models are not ideal. The new processor has a few indisputable advantages, such as large L2 cache, high performance memory subsystem and SSE2 instructions support. However, the other scale contains relatively low core clock frequency. As a result, we either see a performance boost or a performance drop, depending on the factor each particular application is critical to. That is why even though Opteron performs very well in server applications, it is still unable to prove up to the mark in workstations.

We should also complain about the fact that Opteron based workstations appeared a little bit too early. Unlike the server market, where 64bit software has already spread pretty widely, the workstation field still doesn’t have any and unfortunately, you will have to use 32bit applications only, because the tasks of the kind do not support AMD64 technology yet. As a result, we can say that uniprocessor Opteron workstations can be demanded only by some enthusiastic users, dedicated fans of AMD platforms.

However, the situation can change any minute. The changes can be provoked either by the release of Windows OS optimized for AMD64 technology, or by the launch of AMD Opteron processors working at higher core clock.

At the same time, we have to admit that these tests give us some hope that the upcoming Athlon 64 processor turns out a success. Look, we have just seen that Opteron 144 working at 1.8GHz proved simply outstanding in gaming applications. That is why Athlon 64 with higher working frequencies may become an ideal gamer’s platform. So, have patience: there is not too much waiting left :) "


OPteron Review: Lame mp3 encoding.... (http://www.xbitlabs.com/articles/cpu/display/opteron-1_11.html)



Opteron Review: MOre testing that deosn't look so hot!!! (http://www.xbitlabs.com/articles/cpu/display/opteron-1_13.html)


I have to say that this shows me 1.8ghz wont cut it...The athlon64 will need to be 2.2ghz or prepare to have its but handed to it by the 3200+ Barton....The opteron was tested in uniprocessor mode with bith single and dual channel, and I am not sure if the athlon64 has ramped up to be dcddr as well...

If intel releases a 3.4ghz even with current model specs it could give a 2.2ghz athlon64 a dead heat....I don't think the 64bit os will be out to the masses to tout amd's strengths...The author of above thread states that the workstation market has yet to have adopted 64bit as well....

My one disagreement is that athlon64 will be a success if it turns out to be a good gaming rig....C'mon....That is such a small market and AMD needs more then that...

My conclusion workstation results in apps I run everyday on my INtel rig below are not up to par with what I have and at the deficits I saw in a majority of the test I don't see anything short of 2.2ghz getting close....Price is going to be a factor and the athln64 will carry a premium...is it worth it unless it at least lays the hammer down on its predecessor????


Not a flame war!! Discuss it with me...if I am wrong in my interpretations tell me why you think so and maybe I can learn something....I am not too proud to admit it....

[Jeff7181]

You seem to have selected a few benchmarks where the Opteron didn't do well... what about the ones where it did? Like Science Mark: Cipher, Blas x87, and Blas SSE2 Scalar? SSE2 is Intel's instruction set... yet the Opteron does better than the P4 in that SSE2 tests... what does that mean?

I don't think the results are as cut and dry as you're trying to make them out to be.

AND... you don't seem to be considering that the Opteron is just a few months old, and the nForce3 is even newer... how many core revisions have the Athlon XP and P4 gone through to get where they are now? Look at where the P4 started... it was an absolute joke... even you can admit that. But a couple core revisions put it back in the lead.

The Athlon-64 will probably be released at 1.6 Ghz at the low end... who knows.

You can look at AMD and Intel's approach to 64 bit processors many different ways. AMD is getting a head start on Intel... Intel already has a 64 bit CPU, but it's not backward compatible with 32 bit apps. So AMD has that going for them, you can have the best of both worlds. But maybe Intel doesn't care about that... maybe they'll wait for AMD to "clear a path" so to speak, and get more 64 bit mainstream apps, then once everyone is ready to switch to 64 bit completely, Intel will make a 64 bit processor.

I think you're right in saying that the Opteron isn't by any means a nail in Intel's coffin... but it is paving the way for the Athlon-64. Granted the Opteron isn't the cheapest processor you can buy... but neither is a $170 2.4 Ghz P4 =)

*EDIT* A note about gaming performance... it seems as though when the video card isn't being challanged, the P4 can take the lead and drive it harder and harder... but it seems as though when the video card is mainly the limiting factor, the Opteron is better at pushing it harder. And personally, I'd sacrifice 10% at 400 frames per second to gain 10% at 60 frames per second =)

[Duvie]

Those sciencemark test do not negate all the test it does lose in and considerably, and even to its own Barton 3200+....The SSE2 scalar test was also mentioned as very rare instance in real world....

the argument on it being new is valid....BUt it has been awhile now since I last brought up the topic and yet the scores are still quite similar. Before the opteron was mainly being tested in dual mode (which is likely more its intended use) and not in workstation type uses.

I agree the athlon64 will be quite nice when you figure where it will score with a sizeable disadvantage in speed, however can you argue that it needs to be at least 2 to 2.2ghz to beat the 3.2ghz p4 in majority of apps in this current still dominated 32bit environment???

Also I hardly picked a few test here or there....These were just the ones that showed the overall lack of speed and alu that hurts the performance...IE like what has been ailing amd's pr rating of late...lower speed but raise pr rating do to increased cache or fsb where their advantage is not always cut and dry in all apps...Read the review page by page and then tell me I picked just a couple instances of bad performance....

If this was a p4 3.2 versus opteron 144 (1.8ghz) then it is palin and simple a 3.2 majority victory with very sizeable leads in multimedia and some rendering, gaming tight but goes to p4 by an edge....It would have been closer with a 2.0 ghz processor but no guarantee it would have won then and if the athlon64 will be SCDDR which means it would perform on the lower range like the opteron 144 did with scddr...

[Jeff7181]

To be honest... for me the price/performance is more important than which one can get 500 FPS in Quake3 first.

The Opteron isn't cheap... but neither is a 3.2 Ghz P4.

And of course the performance of the Opteron hasn't changed the past 3 months. It's a server and workstation chip... how often does Intel release a new Itanium or a new Xeon? The Opteron is still in it's infancy... I think we need to see at least one core revision before it's deemed a looser. Otherwise, the P4 would still be consider trash, even by Intel guru's.

[Duvie]

Originally posted by Jeff7181
To be honest... for me the price/performance is more important than which one can get 500 FPS in Quake3 first.

The Opteron isn't cheap... but neither is a 3.2 Ghz P4.

And of course the performance of the Opteron hasn't changed the past 3 months. It's a server and workstation chip... how often does Intel release a new Itanium or a new Xeon? The Opteron is still in it's infancy... I think we need to see at least one core revision before it's deemed a looser. Otherwise, the P4 would still be consider trash, even by Intel guru's.

You got me wrong...I am not calling it a looser...All I am trying to get at is that it needs to be ramped higher sooner then later or it may be looked as not a loser but not the coup AMD had planned....Many sites have stated we can look at the opteron in uniprocessor mode to try to gain some sort of knowledge on how the athlon64 will place when it arrives in the marketplace.....I am assumming and hopefully not ASSumming...:wave:

NOt a loser just may be aiming alittle low by AMD to make the splash that I think they had always intended or built the hype up to be.....I don't think there will be enough 64bit stuff available at the time so calling it a loser off the bat would be grossly immature. It will be like the time it took to make the SSE2 a viable asset...

[Jeff7181]

Of course it needs to be clocked higher to compete with the P4 3.2 Ghz... but seeing as how it's not designed to compete with the P4 line, I don't see that as being of much importance.

As far as an indication of Athlon-64 performance, I think it provides a look at what the Athlon-64 at those clock speeds will be capable of... but AMD is stock piling Athlon-64's in their warehouses to prepare for the launch... that means they already have working models ready to sell... that also means they could be working on ramping clock speeds... and to me, THAT means 2.2 and 2.4 Ghz and possibly higher isn't that far ahead in the future.

I've read about AMD's plants to make a "Thorton" using SOI... the Thorton will be the value line, 256k L2 SOI processors, and I'm assuming they'd be in the same clock speeds that we're seeing Athlon XP's in right now.

In short, I would not be surprised at all if there were 2.4 Ghz Athlon-64's by christmas... but I think that has more to do with how Intel does with the 100 watt Prescott =)

[Duvie]

Originally posted by Jeff7181
Of course it needs to be clocked higher to compete with the P4 3.2 Ghz... but seeing as how it's not designed to compete with the P4 line, I don't see that as being of much importance.

As far as an indication of Athlon-64 performance, I think it provides a look at what the Athlon-64 at those clock speeds will be capable of... but AMD is stock piling Athlon-64's in their warehouses to prepare for the launch... that means they already have working models ready to sell... that also means they could be working on ramping clock speeds... and to me, THAT means 2.2 and 2.4 Ghz and possibly higher isn't that far ahead in the future.

I've read about AMD's plants to make a "Thorton" using SOI... the Thorton will be the value line, 256k L2 SOI processors, and I'm assuming they'd be in the same clock speeds that we're seeing Athlon XP's in right now.

In short, I would not be surprised at all if there were 2.4 Ghz Athlon-64's by christmas... but I think that has more to do with how Intel does with the 100 watt Prescott =)


Well since I said I was looking at this as a comparative to the future athlon64 then ofcourse it is being compared to the p4 line.....

2.2ghz at launch should give them the crown albeit maybe by small margins...Then if it is to be believed the 3.4ghz p4 prescott should arrive shortly thereafter and I believe will take it back. Will amd and intel play hop scotch for the est of the way I don't know. I have heard from Intel sources from some time that itel has had working prescotts and ramped up chips tested in the 3.6-3.8ghz range...If the prescott has advantages over the 3.2c then it could be quite the jump for maybe the 2.4ghz athlon64 to retake...

Ofcourse I am guessing on releases and such, but this info was valuable to gain some insight in possible performance comparisons....

[Jeff7181]

I'm just waiting for Cyrix to stick their head back in the game and say "we have a 5 Ghz processor" LOL

[DAPUNISHER]

I mostly agree with Duvie that there are some glaring flaws but I would like to make a few observations and speculations/comments of my own.

The feature of the Opteron that intrigues me the most is Longer 12-step integer pipeline and 17-step floating-point pipeline (by Athlon XP these pipelines are 10 and 15 steps long, respectively). This improvement primarily serves better architecture scalability.

It leads me to speculate that this processor has quite a bit more to offer than the present clockspeed suggests. Jeff touched on this but I actually think he may be a bit conservative in his estimates. I postulate that it will indeed scale higher than the T-bred or Barton is capable of.

The financial considerations are another matter the reviewer fails to integrate into his summations of the Opteron's viability as the basis workstation platform. Being able to work with all the present 32bit software a company uses is very important not only because of the cost of upgrading the software, but also because of the X-factor that almost no one but the CFO ever considers; The cost of training the workers to use new software.

Where the Opteron becomes more attractive than the P4 as a workstation for large corporations is that while it isn't as fast as the P4 at the moment in many uses, especially those that are MHZ dependent, it offers the ability to later make the transition to 64bit OS and apps without upgrading the hardware again. That will factor in to the cost analysis for some and be a major plus. Also, it's important to remember that the Opteron is excellent in Business Winstone which reflects a large area (obviously not exclusively) of what workstations will do in a busniess enviroment.

My last point is that there will be further performance gains derived from bios and driver revisions along with coupling it with a faster DDR or even DDRII since right now it's only using DDR2700. I realize Intel and other companies will be making improvements as well but the Prescott will not be 64bit and that can give the Opteron some attractiveness as a workstation platform if future supporting boards integrate the missing AGP pro and 64bit PCI.

Anywho, I don't think that the failure to indisputably outperform other CPUs at the moment is cause to believe this will be AMD's swansong. ;)

[Jeff7181]

My main point is that the Opteron is a completely new core and is only months old. There's LOTS of time to improve, and lots of research to be done by AMD and nVidia to figure out how to optimize performance.
As far as I know, the origional Athlon was the only new core that really made people's jaws drop since it was such an improvement over the K6-3.

[DAPUNISHER]

Originally posted by Jeff7181
My main point is that the Opteron is a completely new core and is only months old. There's LOTS of time to improve, and lots of research to be done by AMD and nVidia to figure out how to optimize performance.
As far as I know, the origional Athlon was the only new core that really made people's jaws drop since it was such an improvement over the K6-3. My sentiments exactly, and I just wanted to remind people that there are some compelling reasons why for some large corporations the Opteron can be the most fiscally responsible of the platforms they have to choose from.

[XJ.]

Originally posted by DAPUNISHER

The feature of the Opteron that intrigues me the most is Longer 12-step integer pipeline and 17-step floating-point pipeline (by Athlon XP these pipelines are 10 and 15 steps long, respectively). This improvement primarily serves better architecture scalability.


No much of an increase really, 2 more stages per pipeline, come on! That just won't allow much of an increase in clock rate on it's own. Some serious process technology is needed to boost this puppy. I really don't think the Opteron/Hammer CPUS will show any real muscle till they are built on 90nm process tech.


I agree with Jeff7181, with 64 bit apps and OSes, there should be furthur gains to be seen - and that'll take time. I just don't know how much time AMD really has - they can't just keep losing money. I think AMD needs a miracle right about now...

[Jeff7181]

The longer pipeline combined with SOI is what will give the Athlon-64 scalability. SOI chips are supposed to run on 1.2 volts as well... so now it looks as though the P4 is gonna be the mother of all heaters, and the Athlon-64 will be the cooler running chip =) Speaking of heat... I'm gonna see if I can find specs on the Opteron like I've found on P4's and Athlon XP's as far as maximum heat dissipation.

[XJ.]

Well so far SOI hasn't seemed to help the clock rate much...or if it has, then that's pretty sad!

[Jeff7181]

Originally posted by XJ.
Well so far SOI hasn't seemed to help the clock rate much...or if it has, then that's pretty sad! So far the Opteron has been the only SOI chips... and I think it's safe to say those were a test batch... I'm willing to bet you can expect A LOT more from the Athlon-64 now that they're producing them in large quantities and able to work out some manufacturing flaws... with IBM helping I think we'll see results a lot sooner than if AMD was doing it on their own.

[Chuck232]

The extra pipeline length probably won't even offset the extremely complicated manufacturing process. I would think at the same process levels, even a Barton would scale better.

[DAPUNISHER]

Originally posted by Chuck232
The extra pipeline length probably won't even offset the extremely complicated manufacturing process. I would think at the same process levels, even a Barton would scale better. You may well be right, but I wouldn't bet the farm on it brudda ;)

[Jeff7181]

I seriously doubt AMD would have switched to a process that has less potential than their current one.

[Chuck232]

No it does indeed perform faster per clock, but I seriously doubt it'll ram very high. It's probably more of a stop gap between their AXP (which can't scale much higher) and A64 .09u. They couldn't leave it a year without launching a new proc cause they Bartons won't go any further.

[Jeff7181]

Originally posted by Chuck232
No it does indeed perform faster per clock, but I seriously doubt it'll ram very high. It's probably more of a stop gap between their AXP (which can't scale much higher) and A64 .09u. They couldn't leave it a year without launching a new proc cause they Bartons won't go any further. Mine boots at 2.5 Ghz... with some core improvements they could get it there. Look what did with the thoroughbred... 1.8 was a barrier for a while... overclocked to 2.0... now they're hitting 2.4 on a pretty regular basis.

[Chuck232]

But AMD's getting to the land of diminishing returns. Yes, they could scale it probably 100-200MHz further, but the heat dissapated would mean they'd have to bundle some mean ass HSF with the chips. Also, getting the A64 out will give them the oppurtunity to still be competitive with Intel once they get their Prescott out.

Think about it. Prescott would almost beat the AXP clock for clock....

[XJ.]

Prescott does have some serious optimization, if all that intel intending to improve made it to tape out - I haven't heard many details about Prescott lately - AFAIK there are engineering samples out there...

[Chuck232]

I'm looking at PNI and 1MB L2 cache and 0.09u process. A wicked proc probably both at stock and overclocked speeds.

[Jeff7181]

Originally posted by Chuck232
But AMD's getting to the land of diminishing returns. Yes, they could scale it probably 100-200MHz further, but the heat dissapated would mean they'd have to bundle some mean ass HSF with the chips. Also, getting the A64 out will give them the oppurtunity to still be competitive with Intel once they get their Prescott out.

Think about it. Prescott would almost beat the AXP clock for clock.... And you don't think the 100 watt Prescott is going to create some heat? LOL... you've fallen victem to the "AMD's create more heat" myth. That's absolutely and completely false. AMD processors create less heat, but they run warmer because the core is a lot smaller than a P4's so there's less area to dissipate heat through. Believe me... there is absolutely zero problems with heat and AMD right now. My processor running at 2.2 Ghz never reaches 50 degrees C. And the heatsinks they bundle with Retail XP2500's is very good... the newest ones have a copper center, and the rest is aluminum, and the origional ones, I'm not sure what the hell they were, but they were as heavy as my SK-7, and I bet they did a great job too.

[Chuck232]

Originally posted by Jeff7181
And you don't think the 100 watt Prescott is going to create some heat? LOL... you've fallen victem to the "AMD's create more heat" myth. That's absolutely and completely false. AMD processors create less heat, but they run warmer because the core is a lot smaller than a P4's so there's less area to dissipate heat through. Believe me... there is absolutely zero problems with heat and AMD right now. My processor running at 2.2 Ghz never reaches 50 degrees C. And the heatsinks they bundle with Retail XP2500's is very good... the newest ones have a copper center, and the rest is aluminum, and the origional ones, I'm not sure what the hell they were, but they were as heavy as my SK-7, and I bet they did a great job too.

Ok, sorry about that. What I meant to say was that temps are higher... I guess dissipated heat doesn't really mean anything. Temps do. The thing about the Prescott dissipating 100W is probably cause Intel messed up. I'm sure they'll get that down or this is going to be crazy. Think about it, 100W at 3.4GHz.. What are you going to have at 5GHz... I'm positive they'll lower that.

Sure your CPU runs cool at 2.2GHz, but I bet you're not running stock cooling. How about you try even that HSF with the copper core... 2.2GHz is near the latest Barton 3200. Getting it to 2.4 or 2.5GHz would be no small feat. And even then, a 3200+ can't even beat a P4 3.0C much less the 3.2C. I'd say a Barton XP3400+ or even 3600+ would be needed to equal/beat the 3.2C. And have fun trying to get good yeilds of those.

EDIT: So in the end, I'm trying to say that the AXP core is getting outdated. There's limit to what a design can do. Intel I believe stretched it's NW core cause AMD wasn't really pushing out the new designs either. With A64, at least AMD will have a new platform where they can start the process again.

[XJ.]

I just have a hard time believing Prescott @ 90nm will be pushing 100W @ 3.4Ghz - I think something is amiss here. Although, Prescott will have more extensive support for HT, and that boost power consumption quite a bit (when it's being used).

[Jeff7181]

May I suggest this link. (ftp://download.intel.com/design/Pentium4/datashts/29864310.pdf)
And make sure you read the notes under the chart for Thermal Design Power... in case you don't... it says the 3.2 Ghz P4C Thermal Design Power is 82 watts... and the note says, and I quote...
"The numbers in this column reflect Intel's recommended design point and are not indicative of the maximum power the processor can dissipate under worst case conditions."

Then you go take a look at AMD's specs... shown here (http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/26237.PDF) under section 7.
We find that the Athlon XP3200 typically dissipates 60.4 watts, and it's maximum is 76.8 watts.

So, I think based on Intel's statements that their specifications reflect the recommended point, it's safe to say that the specs given are what it typically dissipates, and they don't have a spec listed for the maximum.
Assuming that's true, I don't find it at all hard to believe that the Prescott 3.4 Ghz CPU will dissipate 100 watts. Maybe not typically, but maybe under the "worst case condition" such as running SETI@Home or another DC program like that... or even a game that's CPU intense.

I'd also like to ad that I don't think switching to a new core with significantly effect Intel's heat dissipation, as you can see, the 2.4 Ghz P4C produces up to 10 more watts than the 2.4 Ghz P4B. And that's the SAME core, which shows that if the P4 is doing more work, it gets hotter. Which is exactly what the P4C does... more work because of more available memory bandwidth.
Granted, the Prescott will be on a new manufacturing process, and will operate on lower voltage (1.2 I think I read somewhere)... but will tha treally make a difference? I mean, the 3.4 Ghz one will be the low end, so I'm assuming the architecture will at least ramp to 4.0 Ghz eventually... what will the heat dissipation be like then?!? I'm no engineer, but when you've got something running that fast, it's doing more work, creating more heat, so even at less voltage, it could still be creating the same amount of heat... and if Intel has optimized the Prescott so that it's doing even more work per clock cycle than the P4 is now, that equates to even MORE heat.

[XJ.]

Originally posted by Jeff7181

I'd also like to ad that I don't think switching to a new core with significantly effect Intel's heat dissipation, as you can see, the 2.4 Ghz P4C produces up to 10 more watts than the 2.4 Ghz P4B. And that's the SAME core, which shows that if the P4 is doing more work, it gets hotter. Which is exactly what the P4C does... more work because of more available memory bandwidth.
Granted, the Prescott will be on a new manufacturing process, and will operate on lower voltage (1.2 I think I read somewhere)... but will tha treally make a difference? I mean, the 3.4 Ghz one will be the low end, so I'm assuming the architecture will at least ramp to 4.0 Ghz eventually... what will the heat dissipation be like then?!? I'm no engineer, but when you've got something running that fast, it's doing more work, creating more heat, so even at less voltage, it could still be creating the same amount of heat... and if Intel has optimized the Prescott so that it's doing even more work per clock cycle than the P4 is now, that equates to even MORE heat.

Yes, I would think that @1.2v vCore and with a 90nm process, we should be seeing the Prescott 3.4Ghz using LESS power (typical) than the 3.2Ghz P4C :!:

Unless Intel's 90nm process or the Prescott design is FUBARed.

[Jeff7181]

Does anyone know how many operations per clock cycle the Prescott core will be doing?

Oh... and sorry Duvie... this is getting way off topic, didn't mean to hijack your thread, lol.

[pointreyes]

Originally posted by Jeff7181
Does anyone know how many operations per clock cycle the Prescott core will be doing?

Oh... and sorry Duvie... this is getting way off topic, didn't mean to hijack your thread, lol.

May be OT but I sure would like to know this to because the P4 HT is about the P4 proc that has impressed me compared to AMD XP procs.

[Duvie]

Originally posted by Jeff7181
Does anyone know how many operations per clock cycle the Prescott core will be doing?

Oh... and sorry Duvie... this is getting way off topic, didn't mean to hijack your thread, lol.

No apologies necessary!!! I am enjoying the conversation and learning a bit.....Whatever the way it goes we are keeping it civil and putting out valuable information for others who may read this.....

[Chuck232]

Take a read here.. They suspect Opteron/A64 won't scale very far on the current process. And I still believe 100W is a mistake on Intel's part. I mean, just think about it on the .09u process at 4.4GHz.

http://www.overclockers.com/articles779/:eek:

[Jeff7181]

I just can't help feeling that sooner or later Intel is going to run face first into a wall. They can't keep increasing clock speed forever... and there will be a point at which increasing the clock speed requires so much power, and such extravagent manufacturing processes that it won't be cost effective. I like AMD's approach for gaining performance... "Do more work at once." And I have to say, I like Apple's even more... "You want it faster? Here's another CPU!"

I think AMD must have something up their sleeve. If they planned on keeping the Athlon-64 running at the same speeds the AthlonXP's are right now, why would they have switched to SOI? Current technology is perfectly capable of producing 1.6-2.2 and even 2.4 Ghz processors.

It looks to me like we're going back to the day of the K5 and Cyrix processors where only an Intel Pentium 133 actually ran at a true 133 Mhz. And to be honest, I'm not sure I like that... I really enjoyed the days of the AMD Athlon 800 pounding the crap out of a Pentium 3 800. =)

I hope those days are not over... I hope it's just taking AMD some time to work out the kinks in SOI... I'd really like to see an Athlon-64 at 2.5 Ghz or somethin... that would be impressive. Hell... a Barton at 2.5 Ghz would be impressive =)

Maybe AMD made a bad choice... maybe they should have stuck with the Barton core and worked on increasing it's clock speed rather than creating an entirely new chip?

[XJ.]

For the short term I think Intel's method will work out. They can afford larger die sizes because the produce so many more dice than AMD. So Intel can keep upping the number of pipeline stages, add HT, double the width of the SSE pipeline (I've forgotten not if its the SSE or ALU pipeline that was going to be doubled in Prescott :confused: ), etc...

[Jeff7181]

If Intel would drop their damn prices I might be inclined to try one of their CPU's. But at the moment, $170 for a processor, and $130 for a motherboard is a little out of my reach... at least... when I can get the same performance for $200 from an AMD setup.

[Duvie]

Originally posted by Jeff7181
I just can't help feeling that sooner or later Intel is going to run face first into a wall. They can't keep increasing clock speed forever... and there will be a point at which increasing the clock speed requires so much power, and such extravagent manufacturing processes that it won't be cost effective. I like AMD's approach for gaining performance... "Do more work at once." And I have to say, I like Apple's even more... "You want it faster? Here's another CPU!"

I think AMD must have something up their sleeve. If they planned on keeping the Athlon-64 running at the same speeds the AthlonXP's are right now, why would they have switched to SOI? Current technology is perfectly capable of producing 1.6-2.2 and even 2.4 Ghz processors.

It looks to me like we're going back to the day of the K5 and Cyrix processors where only an Intel Pentium 133 actually ran at a true 133 Mhz. And to be honest, I'm not sure I like that... I really enjoyed the days of the AMD Athlon 800 pounding the crap out of a Pentium 3 800. =)

I hope those days are not over... I hope it's just taking AMD some time to work out the kinks in SOI... I'd really like to see an Athlon-64 at 2.5 Ghz or somethin... that would be impressive. Hell... a Barton at 2.5 Ghz would be impressive =)

Maybe AMD made a bad choice... maybe they should have stuck with the Barton core and worked on increasing it's clock speed rather than creating an entirely new chip?


I agree efficiency in computing is the way to go and you can tell INtel has been looking into that with more optimization in code with a newer advanced sse2 and HT. i think this will all be mute as I think we will all be going to multiple cpus in different arrays...

[XJ.]

Originally posted by Duvie
I agree efficiency in computing is the way to go and you can tell INtel has been looking into that with more optimization in code with a newer advanced sse2 and HT. i think this will all be mute as I think we will all be going to multiple cpus in different arrays...

Yeah, but multi-cpu systems are harder to code for (I know this from experience). Many desktop applications just don't lend themselves to mutithreading, so the gains are not at all uniform across the application spectrum.

[Jeff7181]

Originally posted by XJ.
Yeah, but multi-cpu systems are harder to code for (I know this from experience). Many desktop applications just don't lend themselves to mutithreading, so the gains are not at all uniform across the application spectrum. Unless Microsoft could create an OS that would multi-thread the applications whether they're written for it or not... for example... Windows could direct say, a divx encoding application to one CPU while you're playing Doom3 on the other CPU. Each program would be unaware that the other program is even running.

LOL... what am I saying... Microsoft create an OS that benefits it's consumers? LOL

[DAPUNISHER]

This topic is being discussed in EVERY forum I belong to, yet this forum has by far the most intelligent and productive of them :cool: :beer:

IMO, Intel makes great microprocessors but AMD's microprocessors are great!:D Can you say win/win boys and girls? Good! I knew you could! :rotflmao:

Use whichever you prefer but don't make it seem as though either is a bad choice or will leave you disgruntled because it just isn't the case ;)

[Jeff7181]

Originally posted by DAPUNISHER
This topic is being discussed in EVERY forum I belong to, yet this forum has by far the most intelligent and productive of them :cool: :beer:

IMO, Intel makes great microprocessors but AMD's microprocessors are great!:D Can you say win/win boys and girls? Good! I knew you could! :rotflmao:

Use whichever you prefer but don't make it seem as though either is a bad choice or will leave you disgruntled because it just isn't the case ;) That's very true... they both have their strengths and weaknesses... right now, AMD seems to be able to provide more performance for the price... and Intel holds the absolute performance crown.
However, no matter which you choose, you should be able to do most anything you want to without a problem.

[pointreyes]

Originally posted by DAPUNISHER
IMO, Intel makes great microprocessors but AMD's microprocessors are great!:D Can you say win/win boys and girls? Good! I knew you could! :rotflmao:
Yes. :cool:

Use whichever you prefer but don't make it seem as though either is a bad choice or will leave you disgruntled because it just isn't the case ;)
Very true. But I have to admit that I was sorely disappointed with the P4 line until the P4 HT procs came out. And that just happens to go inline with the type of stuff I'm doing-Oracle development in conjunction with Java. For multi-media then by all means P4 is the better of the two (at this time) but I'm throughly sick of the people that make the P4 the god of superior procs. Good grief, I'm in the realm of corporate computing with multi-threading and I'm tired of 32-bit due to what I need (and use) at work. Oh yeah, Linux works absolutely fantastic on a P4 HT-better than XP Pro does with the type of work I'm doing. I'm so glad that I have choices. :)

[Duvie]

Originally posted by Jeff7181
Unless Microsoft could create an OS that would multi-thread the applications whether they're written for it or not... for example... Windows could direct say, a divx encoding application to one CPU while you're playing Doom3 on the other CPU. Each program would be unaware that the other program is even running.

LOL... what am I saying... Microsoft create an OS that benefits it's consumers? LOL


I am guessing here cause I am not that technical, but when we talk about parallel processing is that done at the os level or is this something that is coded into the mobos or the cpus by themselves??? I mean to the OS it is still single threaded but at the level of the first cpu or through say like the northbridge or some other controller chip the single threaded process is analysed and then split up in smaller chunks to have each cpu work on a bit of it then reassemble upon completion. The point being that even with the hit in time to analyse the operation the cpus get it done faster cause they theoretically reduced the size of the operation by the number of chips working in the parallel process...Lets say 4...It now takes 25% less time to do with a hit in time for analysing at the front door and reassembly at the back door. Nevertheless the work can say be done at 35% of the time.


I have heard this idea for some time....The chips are getting smaller and smaller yet the number of transistors increase and at some point we need to start looking at the continue cost benefit of doing these transitions and look at multiple arrays as being the most efficient.

[Jeff7181]

That's what I'm sayin... multi-threading right now is handled by instructions in the CPU right? Instead, make the OS able to divide up applications between the CPU's.

[Duvie]

So therefore if the OS does the actual separating of the task to be disseminated down to the cpus then it wouldn't matter if the application itself was multithreaded, right???

I think parallel processing is the future...For reasons of cost as we can't keep continuing to ramp down in size .09 micron to .05 micron, etc while heat continues to increase (eventhough it drops a bit relatively due to the drop in volatge each time we do drop in size)....Maybe this doesn't happen until we get into the .05 range but can you imagine say a mobo with 4 cpus with hsf's covering an area not much bigger then the area the p4 retention clip covers now. The chips wouldn't have to be as big if each chip did not need as much cache as there is on one chip now. My understanding that takes up quite a bit of space. 4 cpus about 3/4 to 2/3 the size of current cpu with one large hsf spreading on top of all of them in one large square pattern. I can see it now!!!:-)

the real question is will most of us need that power??? Not likely but it seems in this day and age the cost it takes to keep retooling these fabs for smaller and smaller components is starting to to reach its limits. Fabs are starting to become obsolete in a fraction of the time they use to be yet the fabs are more and more expensive to build. I think parallel processing offers that break from this rat race and to utilize existing technology.

[Jeff7181]

So therefore if the OS does the actual separating of the task to be disseminated down to the cpus then it wouldn't matter if the application itself was multithreaded, right??? Yes, that's what I'm thinking.

I think you're right about the parallel processing... although, it would seem to me Intel will keep ramping clock speeds... and once another company overtakes them with parallel processing, Intel will just switch to parallel processing with it's insanely high clock speeds and take back the lead.

It does make sense though... super computeres, even the newest ones aren't made up of Pentium 4 3.2 Ghz processors... they're made of a lot more slower processors.

What would be very interesting is if either AMD or Intel made a system with say, 4 or 8 processors all running at a low enough speed and low enough voltage that they could be passively cooled... all you'd need then is good case ventilation.

[Duvie]

Originally posted by Jeff7181
Yes, that's what I'm thinking.

I think you're right about the parallel processing... although, it would seem to me Intel will keep ramping clock speeds... and once another company overtakes them with parallel processing, Intel will just switch to parallel processing with it's insanely high clock speeds and take back the lead.

It does make sense though... super computeres, even the newest ones aren't made up of Pentium 4 3.2 Ghz processors... they're made of a lot more slower processors.

What would be very interesting is if either AMD or Intel made a system with say, 4 or 8 processors all running at a low enough speed and low enough voltage that they could be passively cooled... all you'd need then is good case ventilation.

That last part is exactly what I was thinking....Say intel goes up to 5ghz before they need to start looking at this.....Then say they could run 8 cpus in a paralllel process array. If they could each of those chips running 1ghz or less could we get passive cooling or could we cool all eight of the cpus with a very low speed low audible fan....6cpus running 1ghz or 4 cpus running 1.4ghz. The heat dissipation off of 1.4ghz with likely vcore ranges of 1v or less in the .05 micron process could make the heat factor reasonable.

[DAPUNISHER]

The Opteron is proof AMD is not "sleeping on the job" when it comes to developing arrays either. The 4-way Opteron they did the SETI benchies on showed some good distribution among all 4 WUs. Times weren't as impressve as I had hoped but there isn't a 64bit client for it as there is with Itanium so that held it back.

[XJ.]

Currently the OS (though it can be overridden by programmers) handles the distribution of app workload over multiple cpus. Multithreading must be done by the application programmer, then the OS handles the scheduling. HT on P4s uses hardware multithreading that is not under the control of the programmer and appears to be a seperate logical processor to the OS (Talking Windows here). There are some new instructions being added to HT in Prescott which will allow the OS to HT "aware" - whether applications will have access to the "knowledge" or not is beyond what I know.

Mutliple parallel processors generally would not improve the peformance of most desktop applications. Period. Over time as more applications are being used in a semi-simulaneous fashion, and applications become more deeply mutli-threaded, then dualies would be helpfull. Quads may even be make sense to those who make heavy use of mutli-media applications - but I can't see much use for multi-core or parallel processing in the near future.

I'm talking about desktop computing only - servers are a different story and have been for quite some time...

[Jeff7181]

Originally posted by XJ.
Currently the OS (though it can be overridden by programmers) handles the distribution of app workload over multiple cpus. Multithreading must be done by the application programmer, then the OS handles the scheduling. HT on P4s uses hardware multithreading that is not under the control of the programmer and appears to be a seperate logical processor to the OS (Talking Windows here). There are some new instructions being added to HT in Prescott which will allow the OS to HT "aware" - whether applications will have access to the "knowledge" or not is beyond what I know.

Mutliple parallel processors generally would not improve the peformance of most desktop applications. Period. Over time as more applications are being used in a semi-simulaneous fashion, and applications become more deeply mutli-threaded, then dualies would be helpfull. Quads may even be make sense to those who make heavy use of mutli-media applications - but I can't see much use for multi-core or parallel processing in the near future.

I'm talking about desktop computing only - servers are a different story and have been for quite some time... That's true, the average person who goes to CompUSA and buy's a $600 computer with a 1.6 Ghz Celeron and 256 MB of RAM because it "would look pretty on the desk in the office" wouldn't be the target market for this parallel rigs. The target would be gamers, and multi-media enthusiasts, and programms, and 3D designers, and web designers, and server administrators, etc etc.

Imagine for a minute, a rig with 8 processors, each running at 1 Ghz with a bus speed of 250 Mhz, each with it's own 1 MB of L2 cache, maybe a 128k or 256k L1 cache, each with it's own 512 MB of ultra low latency DDR RAM... and of course they'd have a high speed link between all of them for when acting as a single CPU in "compatability mode"
Then imagine the Operating System could take an application that isn't designed to run on parallel processors, and make those 8 processors look like one. So each CPU would be processing a portion of the application. Or maybe, one CPU would have to be dedicated to splitting up tasks evenly between processors when the OS is in "compatability mode" for applications that don't support it.
Then imagine Quake VIII (lol) is designed to run on a parallel rig... and you tell it how many CPU's you have in the in-game settings, and it uses one processor to process the physics engine, between 2 and 4 for floating point calculations, uses another for AI, uses another to process sound, and uses the 8th one in multi-player mode to keep track of other players.

Or imagine SETI@Home running on an 8-way rig... you'd be cranking out WU's every 20-30 minutes, lol.

:yummy: We can dream can't we? =)

[DAPUNISHER]

I watched a streaming video from last X-Mas where Anand Lal Shimpi addressed a bunch of questions at an event Kyle from [H] put together with Nvidia, Intel, and some other sponsors and he made some excellent points about the near future of computing. One of the things he spoke about was that CPUs are soon going to need a completely new packaging technology to overcome the limitations of heat and power requirements.

If that packaging technology isn't developed inexpensively enough due to the need for exotic materials and processes like this ->
Motorola is pursuing a different approach. The company has developed a new manufacturing process that combines gallium arsenide with silicon. Gallium arsenide has long been considered an attractive alternative to silicon because its crystal lattice structure offers less resistance. However, whereas silicon chips are durable and cheap, gallium arsenide chips can be fragile and expensive. By layering gallium arsenide onto silicon, Motorola has created a semiconductor that combines the durability and economy of silicon with the high speed of crystal compounds normally reserved for lasers and fiber optics. The result: the potential to produce processors that run at speeds up to 70 gigahertz. NASA has conducted experiments on Gallium Arsenide Crystal Growth in Microgravity link (http://microgravity.grc.nasa.gov/expr/gaas.htm)

There is a wealth of information about using Gallium arsenide and growing gadolinium oxide on a gallium arsenide semiconductor available. Bell labs made a significant breakthrough for instance good stuff! (http://www.lucent.com/press/0399/990319.bla.html)

Problem is how soon will such technologies be viable for mass production if ever? DNA computing appears to be far too limited in what it will be capable of, although it's too early to really know what it's true potential is they say.


My point is, as Anand lectured about, we need a new packaging technology to emerge that is viable for replacement of contemporary ones to keep up with Moore's prediction (why they call it law is beyond me) or As others here have asserted we need processor arrays combined with the OS and software that can both multithread and mutitask well to effectively increase computing power on our desktops. Of course I could be completely out of my mind but the jury is still out on that one :look:

[Jeff7181]

Very interesting.