[NEWS] - Intel invents overclock deterrent method

[Mercutio]

IMO, the "tiny percentage faster than the appropriate P4 in most desktop benchmarks" that AMD implies in the XPwhatever+ naming convention is pretty much the only reason I find that nomenclature palatable. By and large it holds, regardless of which 1:1 set of CPUs is being compared (I admit I haven't seen a XP3000+4 3.06GHz benchmark).

e_dawg: Heat/Noise is being rendered moot as well. The latest P4 chips dissipate more heat than any AMD chip, and need cooling measures just as extreme. That's not much of a selling point on the higher-end chips (2.6, 2.8, 3GHz), where the price/performance line finally becomes blurred.

 

[Tea]

The heat and power myth: some actual max power numbers

Older 0.18u parts:
54W: 0.18u Thunderbird 1000C (vastly faster than the early P-4)
75W: 0.18u Pentium 4 1.5 (Socket 423)
(There were at least three variants of the 1.5 - the other two are 76W and 79W.)

Now let's consider the next performance level. There was hot debate as to which of these two chips was, in fact, faster, the P4 1.7 or the Thunderbird 1400. Most experts - about two out of three - regarded the Athlon as the faster of the two, but some argued the other way, and everone agrees that they were very close.
72W: 0.18u Thunderbird 1400
87W: 0.18u Pentium 4 1.7 (Socket 423)

Now let's look at the next step again (and another hotly debated "which is the best of two rather similar chips" contest. The consensus is that the XP 1700 was comfortably faster than a P4 2.0 - remembering that this is the ''old'' 256k 1.8 micron P4, not the vastly improved 0.13u Northwood that came along in January 2002.
64W: 0.18u Athlon XP 1700+
96W: 0.18u Pentium 4 2.0 (Socket 423)
100W: 0.18u Pentium 4 2.0 (Socket 478)

Now let's throw in the Northwood as well - but we must remember that in doing this we are not comparing like with like, as the 2.0 Northwood is faster than an XP 1700 and made on a different and much cooler running process. I'll also include the 0.13 Athlon at 2000+ for comparison
69W: 0.13u Pentium 4 (Northwood)
62.8W: 0.13u XP 2000 (Thoroughbred)

Finally, let's take the two current-model 2600 parts:
68.3W 0.13u Athlon XP 2600+
81W: 0.13u Pentium 4 2.6 (Northwood)

I'm not entirely convinced that heat and power matter too much, or that there is a great difference when you compare Intel vs AMD, but let's at least get the numbers right, shall we?

Source: www.sandpile.org

Motto: And we had fud, fud, fud till my daddy took the T-bird away - ( the Beach Boys)

 

[honold]

PS: it's not from AMD's website, BTW. Not that it matters..

it was definitely sourced from there. it's their graph/font/etc

 

[honold]

(same font/graph/etc they used when they said the k6-2 300 was as fast as a pii-400...)

anyway, yes, amd chips use less power than intel. they are also much hotter across the board, and lack the anti-destruction mechanism the p4s have from overheating.

http://www.anandtech.com/cpu/showdoc.html?i=1718&p=5 . start there for a large list of real benchmarks from a good source.

 

[honold]

http://www.anandtech.com/cpu/showdoc.html?i=1718&p=3#Willamette shows different thermal output than you quoted as well

 

[honold]

with a lower ambient temp my retail 1900+ at work runs 5-7c hotter than my overclocked northwood 1.8a/400 (running at 2.4/533). both use retail hsfs, and the p4 fan is quieter as well. i had to manually apply thermal grease on all of the 1.4 tbirds we use to keep them from crashing overnight in the summer when the building doesn't use the ac as liberally.

1.4 tbirds are the hottest chips i have ever encountered.

 

[time]

Sysmark 2002 was comprehensively discredited last year. If you're going to quote Anandtech, at least use a more up to date article:

http://www.anandtech.com/cpu/showdoc.html?i=1746&p=11

The Athlon XP2800 appears to win slightly more benchmarks than the P4 2800, although the difference in many cases is less than 3%, and therefore outside the margin of error.

Now the Athlon XP has the lead; it just goes to show you that the days of absolute performance leaders are gone, you can only hope to pick the processor that does better overall.

 

[e_dawg]

AMD CPU's have historically run hotter than their Intel counterparts. It is not always true, like with some models of the Wilamette P4, but in general, AMD CPU's run hotter than Intel. This is just common knowledge, herr chimp. You must be the only one to refute this seemingly fundamental truth.

Intel Pentium 4 2.0AGHz (Northwood) 54.3W 0.13-micron
Intel Pentium 4 2.20GHz (Northwood) 57.1W 0.13-micron
Intel Pentium 4 2.26GHz (Northwood) 58.0W 0.13-micron
Intel Pentium 4 2.40GHz (Northwood) 59.8W 0.13-micron
Intel Pentium 4 2.50GHz (Northwood) 61.0W 0.13-micron
Intel Pentium 4 2.53GHz (Northwood) 61.5W 0.13-micron
Intel Pentium 4 2.60GHz (Northwood) 62.6W 0.13-micron

AMD Athlon XP 2000+ (Thoroughbred-A) 60.3W 0.13-micron
AMD Athlon XP 2100+ (Thoroughbred-A) 62.1W 0.13-micron
AMD Athlon XP 2200+ (Thoroughbred-A) 67.9W 0.13-micron
AMD Athlon XP 2000+ (Thoroughbred-B) 61.3W 0.13-micron
AMD Athlon XP 2200+ (Thoroughbred-B) 62.8W 0.13-micron
AMD Athlon XP 2400+ (Thoroughbred-B) 68.3W 0.13-micron
AMD Athlon XP 2600+ (Thoroughbred-B) 68.3W 0.13-micron

Source: http://www.anandtech.com/cpu/showdoc.html?i=1718&p=3#Willamette

 

[time]

Different temperatures on completely different chips means absolutely nothing. The smaller Athlon also has a smaller contact patch, so it's logical to expect a higher temperature if everything else was equal (which it isn't).

In actual fact, the P4 is more susceptible to overheating than the Athlon. This is why it has complex thermal management built-in, including clock throttling in extreme cases. The Athlon only needs to shut down if onchip temperatures exceed 100-110 degrees C - which it does (motherboards were required to support this more than a year ago, although almost every decent Athlon board has always had some kind of thermal shutdown).

The P4 retail fan is 70mm, the Athlon 60mm. To deliver the same airflow, the 60m fan has to spin at least 36% faster, which produces 85% more noise. On top of that, the 60mm fan is under greater load because of the closely spaced fins necessary on the Athlon's smaller heatsink.

In practise, the P4 needs more airflow, so things aren't quite as bad as this picture suggests. Ask Mercutio about the improved noise levels of coolers with 70mm fans, or anyone here about 80mm fans.

 

[time]

The table on Anandtech is flat wrong. That's not maximum dissipation, but the figure Intel provides as a design guide to a minimum thermal solution.

Until recently, the typical dissipation of Tbreds has been higher than it could be because Suspend during Idle (or whatever the hell it's called) has been disabled on motherboards.

Go check out ARS or Aces or something similar for real info on P4 thermal issues.

 

[e_dawg]

Oh well, I guess things have changed... AMD was always known as the fire breathing power hungry CPU's in the Athlon vs. P3 days...

 

[honold]

Sysmark 2002 was comprehensively discredited last year. If you're going to quote Anandtech, at least use a more up to date article:

http://www.anandtech.com/cpu/showdoc.html?i=1746&p=11

The Athlon XP2800 appears to win slightly more benchmarks than the P4 2800, although the difference in many cases is less than 3%, and therefore outside the margin of error.

i was using that article because it was the first athlon 333 review, and we were mostly talking about the low end

intel is winning in business & content creation, media encoding, archiving, and 3d rendering. amd wins in gaming, workstation 3d, and science. where is the slightly more?

 

[honold]

Different temperatures on completely different chips means absolutely nothing. The smaller Athlon also has a smaller contact patch, so it's logical to expect a higher temperature if everything else was equal (which it isn't).

pull off the integrated heat spreader on a p4 and you'll see a core just as small as the athlon. this a bad omission by amd.

In actual fact, the P4 is more susceptible to overheating than the Athlon. This is why it has complex thermal management built-in, including clock throttling in extreme cases. The Athlon only needs to shut down if onchip temperatures exceed 100-110 degrees C - which it does (motherboards were required to support this more than a year ago, although almost every decent Athlon board has always had some kind of thermal shutdown).

and the motherboards weren't required to read the temp off an internal diode instead of a thermistor? i'm not buying that, and nobody implemented the diode reading while it sat there useless. the auto-shutdown is a manufacturer feature and it isn't specific to amd or intel.

The P4 retail fan is 70mm, the Athlon 60mm. To deliver the same airflow, the 60m fan has to spin at least 36% faster, which produces 85% more noise. On top of that, the 60mm fan is under greater load because of the closely spaced fins necessary on the Athlon's smaller heatsink.

another bad omission by amd. should we celebrate the use of louder, less efficient products? you can fit a 70mm fan onto an athlon.

i don't understand why temps aren't commonly read off the diode, why they don't have an ihs, and why they use crappy retail hsfs.

 

[blakerwry]

i understand that the reason they use crappy HSFs is related to cost... the retail ones work... that's about it... a 70mm fan would be a good idea.. and I suspect that the K8 series probably comes with something better than what they're currently using.


As far as I know a majority of kt266a and above board *DO* use the on die thermal diode that is included in the athlon XP and Barton Cores.

However, some motherboards still include an in-socket thermister to "fall back on" when using a Duron or t-bird CPU without the thermal diode... they probably also use this in-socket thermister for overheat protection.

 

[Tea]

Honold sez that the P4 and Athlon cores are about the same size, and that the lack of a heat spreader on the Athlon is a "bad omission".

OK. I'll bite. Why?

CPU's always used to have a heat spreader, until Intel's Coppermine. This was the first "naked" chip. AMD followed suit with the Thunderbird and before too long, everything was naked - Duronz, Celeronz, the lot.

Then, with the final P-IIIs and the P-4 and the Celeronz, Intel switched back to CPUs fully and modestly dressed in metal again. AMD are doing the same with the 64-bit Athlonz.

So, if a heat spreader is good, why did everybody stop doing it? And if it's bad, why are they doing it again now? Reminds me of that ztupidest of all CPU mounting methods, Zlot 1. No. Sorry, make that "second ztupidist method of all time" - for the designers of Slot A had had every opportunity to see how clumsy and crappy Slot 1 was, and chose to do likewise anyway. Sometimez I think it's all just mindless fashion. Or zomething.

Go figure.

 

[blakerwry]

Reminds me of that ztupidest of all CPU mounting methods, Zlot 1. No. Sorry, make that "second ztupidist method of all time" - for the designers of Slot A had had every opportunity to see how clumsy and crappy Slot 1 was, and chose to do likewise anyway. Sometimez I think it's all just mindless fashion. Or zomething.

Go figure.

lol, did you ever realize that a slot A is physically just an upside down slot 1? you can litterally take an athlon and put it in a slot 1 motherboard and vise versa.... (No Tea! I did not mean you can put a slot 1 motherboard in a athlon, I meant that you can put a slot I CPU in a slot A board)


At least AMD was smart enough to steal their socket designs from Intel....(although now using an incompatible bus) as they had in the past with socket 7, socket 5, socket 3, and socket 2... (did any AMD chips actually work with socket1?)

 

[Tea]

Thankz, mate.

I tried it.

When I turned the power on ....


IT WENT PFFFHT!!!!


End of a perfectly good Athlon 750 and my favourite old Gigabyte BX board.


Lazt time I lizten to you, Blake.

 

[time]

intel is winning in business & content creation, media encoding, archiving, and 3d rendering. amd wins in gaming, workstation 3d, and science. where is the slightly more?

Okay, obviously you're reading a different link to the one I am. :-?

In my universe, the Anandtech benchmarks read as follows:

Test                  XP2800   P4 2800  AMD/Intel
Content Creation       44.8      45.4     -1.3%
Business               34.0      32.0     +6.3%
DiVX MPEG4             66.3      71.4     -7.1%
Lame MP3               88        87       +1.1%
WinRAR                372s      315s     -15.3%
UT2003 Flyby          194.2     191.3     +1.5%
UT2003 Botmatch        78.5      70.0    +12.1%
3DMark2001          14694     15089       -2.6%
Jedi Knight           163.4     168.1     -2.8%
Serious Sam           156.0     143.5     +8.7%
Comanche 4             52.4      55.5     -5.6%
3DSMAX SinglePipe     226s      216s      -4.4%
3DSMAX Underwate      301s      309s      +2.7%
3DSMAX Rays            21s       22s      +4.8%
3DSMAX Cballs          43s       43s       0.0%
3DSMAX Vol_light       21s       19s      -9.5%
Maya                   70s       73s      +4.3%
Lightwave Raytrace    136.3s    107.8s   -20.9%
Lightwave Radiosity    91.0s     52.2s   -42.6%
Lightwave Sunset       42.1s     56.4s   +34.0%
SPECviewperf 3dsmax_01 10.74     10.57    +1.6%
SPECviewperf drv_08    39.88     31.41   +27.0%
SPECviewperf dx_07     62.22     48.58   +28.1%
SPECviewperf light_05  15.48     12.06   +28.4%
SPECviewperf proe_01   13.36     11.67   +14.5%
SPECviewperf ugs_01     7.104     7.167   -0.9%
ScienceMark Molecular  80.8s    109.4s   +35.4%
ScienceMark Primordia 466.3     589.9    +26.5%
-----------------------------------------------
AMD wins                                  16
Intel wins                                11
Dead heats                                 1
TOTAL                                     28

Tea is correct in that Athlon is faster in general usage and Office performance (i.e. Business Winstone). Although Anandtech didn't explicitly include the Athlon 2400, you can calculate that it will perform very similarly to the P4 2800 in these common functions.

I find it astounding that even with one of the most extensively P4-optimized programs around (Lightwave), the Athlon still manages to sigificantly best its competitor in one of the three related tests.

 

[honold]

Honold sez that the P4 and Athlon cores are about the same size, and that the lack of a heat spreader on the Athlon is a "bad omission".

OK. I'll bite. Why?

because ihs cpus run cooler and don't easily chip the cores.

 

[honold]

intel is winning in business & content creation, media encoding, archiving, and 3d rendering. amd wins in gaming, workstation 3d, and science. where is the slightly more?

Okay, obviously you're reading a different link to the one I am. :-?

In my universe, the Anandtech benchmarks read as follows:

5 benchmarks are games, and 5 benchmarks are specview 3d.

should i claim the p4 is better by testing 680 different encoding apps and listing them? one only needs a large enough sampling on a category to claim superiority, and for my MONEY, the category is all that matters. p4 won more (and more relevant to me) categories.

i am a highly active, competitive gamer but the margins are irrelevant for me in the games that count - namely q3 promode and war3. if my game plays faster, it doesn't save me time. if my gigantic rar builds faster, or my divx/mp3 encodes faster, i have saved time.

 

[NRG = mc²]

because ihs cpus run cooler and don't easily chip the cores.

Although I'm sure it protects the core, I'm not sure it helps at all with the heat transfer. Its still the same area for the heat to pass through (core -> IHS), but with a heat spreader you have two separate interfaces for the heat to pass through.

 

[honold]

it's been confirmed by the overclocking crowd that it does work for heat reduction via removal and tests. many people now remove the ihs, apply thermal glue, and place it back on because they're dissatisfied with the default material they use to connect it.

 

[mubs]

If indeed the IHS is effective, one would think the mfrs would make the IHS out of metal with a really low thermal resistance??

 

[Tea]

OK, I'll go along with the info that it helps get rid of the heat, at least for now. So why did Intel and AMD both stop using them?

 

[honold]

probably cost savings, maybe the same reason they went to slot? the world may never know.

 

[Tea]

Cost savings makes sense. But going to slot was a reverse cost saving. IIRC, Slot processors cost around $15 more than socket ones. BTW, the core damage thing is a non-issue now: it only really applies to P-IIIs and early Thunderbirds. The current ones are no problem at all, except for real gorilla techs.

 

[honold]

i remember the initial cost being higher, but i recall they said after n months it would be more cost-efficient and blah blah

 

[CougTek]

The slot adventure was caused by the inability, at the time, to integrate large L2 cache into the same die as the microprocessor. As soon as they've been able to master (ie - improved their yield to decent levels) how to put large amount of fast L2 cache into the same die as the rest of the processor, slot no longer had reason to be.

 

[Tea]

Well, yes, Coug. But it was still a damnfool way of doing it. Would have been much better to mount the chip itself on a socket, same as usual, and have the cache chips on horizontal "wings".

 

[Buck]

horizontal "wings".

You lost me (which is easy).

 

[Jan Kivar]

The slot adventure was caused by the inability, at the time, to integrate large L2 cache into the same die as the microprocessor. As soon as they've been able to master (ie - improved their yield to decent levels) how to put large amount of fast L2 cache into the same die as the rest of the processor, slot no longer had reason to be.

Are You sure, Coug? I've always thought that Intel started using Slot-1 as they had the rights to it, similiar to renaming 586 to Pentium for rights. AMD bought their Slot-A technology (with the EV6 bus architechture along it) from Digital.

Cheers,

Jan

 

[Mercutio]

CougTek is correct. In fact, I remember my Intel Sales rep telling me BEFORE I EVEN SAW A SLOT 1 package that they'd be moving back to Socketed CPU at the first possible opportunity (of course, at the time I was told both would be options for the future, too), once the technical aspects of integrating the cache were solved.

I loved the Slot 1 interface. Waaaaay easy to work on.

 

[iGary]

Via motherboards seem OK to me. What do you do with your PCs that I don't?

One would have to suspect that -- by now -- Via has finally tweaked their KT chipset + driver + core BIOS code enough to gain stability for the masses of home and office computer users. However, you would be foolish to build a serious 24x7 server based on Via's existing marginal chipsets.

Someday, maybe someday, Via. Just not now. For 24x7 servers, the present is owned by Intel and Serverworks.

...Ye Gods! They've discovered processor remarking already!...

Monsieur Owlsley and crowd, let's agree to spell it as: re-marking

 

[Howell]

At least AMD was smart enough to steal their socket designs from Intel....(although now using an incompatible bus) as they had in the past with socket 7, socket 5, socket 3, and socket 2... (did any AMD chips actually work with socket1?)

At some point it becomes deciding how many pins or pads you need to make the proc work and then designing the physical interface. AMD didn't design the slot-A interface for AMD or you or me but for the motherboard manufacturers. They already have the stock on hand and don't have to keep two kinds of stock.