Sign in with
Sign up | Sign in
Your question

AMDs response to Northwood

Last response: in CPUs
Share
January 5, 2002 4:38:09 AM

I've been following the AMD and INtel wars from a distance and recently been wondering what would AMD's response in terms of technological breakthroughs/anything to combat the Northwood chip release... ANy wild speculations? Does Northwood really look to be the PowerProcessor its hyped to be? will normal people with normal budgets like me get to use it in the near future?

amd_dma
January 5, 2002 5:02:53 AM

Quote:
will normal people with normal budgets like me get to use it in the near future?


Are you the chosen one?

<font color=red>God</font color=red> <font color=blue>Bless</font color=blue> <font color=red>America!</font color=red>
January 5, 2002 5:06:27 AM

AMD will release the Thoroughbred, which will basically be an Athlon XP built on the .13 micron process (it MIGHT have more cache).
As for the Northwood, on it's initial release I'd expect it to perform no more than 10% faster (that's my personal guess) than the current Pentium 4 (clock for clock). However, it will ramp up the clock speeds quickly and gain a 533Mhz FSB (4 x 133Mhz) and Hyperthreading as it goes along.
I'm sure Ray could give you more details on what the Northwood will and will not include, but he probably wont. ;) 

"Ignorance is bliss, but I tend to get screwed over."
Related resources
January 5, 2002 5:27:33 AM

533mhz fsb! wow. whats the barton fsb going to be?

also, what is hyperthreading?
January 5, 2002 5:37:32 AM

now that AMD and Intel are gonna make these turbocharged chips, what about their heat solutions? is there any way i can enjoy all that power without allthe air(and dust)going around and all the noise existing heat solutions provide? watercoolers sound nice though but really, do we all want to go that route? thanx for all da info!


amd_dma
January 5, 2002 5:38:18 AM

I'm wondering how much of a performance boost we should expect from SOI on the Barton and Clawhammer. 10-20%?
January 5, 2002 5:42:18 AM

Beats me.
That 533Mhz FSB is with RD RAM, and would offer around 4.3GB/second of bandwidth.
I've head that the Hammer will have a 333Mhz FSB (2.7GB/second), so I suppose the Barton would be no faster than that.

Hyperthreading is something like SMP-on-chip, where the CPU does multiple things at once. At least a 10% performance increase is expected, but that's really just a guess.

"Ignorance is bliss, but I tend to get screwed over."
January 5, 2002 5:45:16 AM

hmmm... 4.3gb/second... will we even be able to use all that up?
January 5, 2002 5:50:16 AM

With the die shrinks the heat produced will go down initially and then build up again as the processors get faster and faster...until the next die shrink. Who knows, maybe they'll get cooler overall? (wishful thinking I'm sure)

"Ignorance is bliss, but I tend to get screwed over."
January 5, 2002 5:57:49 AM

Can we use 4.3 GB/s? Yes.
Will we? That depends on the application.
The important thing is that both the CPU and BUS will support it. The reason the NForce board doesn't create a huge increase in performance is because Althons CPU's can only use 2.1GB/s, despite the 4.2GB/s available to it.

"Ignorance is bliss, but I tend to get screwed over."
January 5, 2002 7:25:20 AM

As I see it Intel could rip AMD apart until Throughbred arrives, if they immediately started releasing
Northwoods in versions near the cores frequency limits, maybe around 2.8GHz and adopted the following price structure for boxed processors for a starters.

2.8GHz Northwood A: 350$
2.6GHz Northwood A: 300$
2.4GHz Northwood A: 250$
2.2GHz Northwood A: 200$
2.0GHz Northwood A: 180$
2.0GHz Willamette: 150$
1.9GHz Willamette: 140$
1.8GHz Willamette: 130$
1.7GHz Willamette: 120$
1.6GHz Willamette: 110$
1.5GHz Willamette: 100$

The P4's strong side is its ability to reach high frequencies, so Intel better start capitalising on that right now, being the only ones to master a fullblown 0.13 micron technology.


/Copenhagen - P4 Willamette 1700MHz@2109 MHz, Vcore 1.75V@2.20V on Abit TH7II-RAID.
January 5, 2002 12:39:57 PM

<i>... how much of a performance boost we should expect from SOI on the Barton and Clawhammer. 10-20%? </i>

IBM (who pioneered SOI) have had a lot of success with it on the Power chips seeing 20-25% more headroom with it.

<font color=blue> Smoke me a Chip'er ... I'll be back in the Morgan </font color=blue> :eek: 
January 5, 2002 4:10:48 PM

Untill Intel increases their IPC's then they will always be slower than AMD clock for clock. Thats why my 1.33 is just as fast as others 1.7

Crap, all the good ones are already taken.
January 5, 2002 6:37:32 PM

Intel should go straight to the high frequencys, AND lower all of their prices dramatically? You may as well say AMD should go straight to the Hammer, and give their processors away for free. :p 

"Ignorance is bliss, but I tend to get screwed over."
January 5, 2002 7:23:37 PM

thats the same old arguement...
who cares which has the higher IPC? doesn't the end result of faster applications and such matter more than which does more per clock cycle?
right now the P4 is not impressive, I think once there is a bout 1ghz between the P4 and the Athlon it may be, but the FPU performance still sucks.
Regardless, the P4 will outlive the Athlon, once the speed gets rediculous, hopefully so will the performance.

----------------------
Independant thought is good.
It won't hurt for long.
January 5, 2002 7:25:47 PM

I wouldn't mind buying an Athlon for free, hehe.

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
January 5, 2002 8:22:52 PM

You wouldn't rather have the free Hammer? :wink:

"Ignorance is bliss, but I tend to get screwed over."
January 5, 2002 9:04:30 PM

Yup you're right. P4 will outlive Athlon. Athlon is to be phased out in 2003 with the Hammer taking it's place. Hammer will be out in a mobile form factor supposedly early in 2003 and of course, a workstation chip sometime late this year.

I have few concerns about the P4 overtaking Athlon--to any great degree. Preview benchmarks I have seen seem to indicate that the Hyperthreaded P4 only offers around 10% performance boost over current P4 in real world benches.

The beauty of the Athlon AND Hammer is these chip designs don't NEED to move to astronomical clockspeeds to retain performance parity. So, while P4 could reach 3 ghz possibly in the next 12-18 months, Athlon should be able to offer similar performance in the low 2ghz clockspeed..and I believe that may be in the cards for Athlon this year. So, while some people may buy the latest and greatest P4 because it's the *fastest*, other folks will buy Athlon because it offers the most bang for the buck.



When all else fails, throw your computer out the window!!!
January 5, 2002 9:32:28 PM

Quote:
The beauty of the Athlon AND Hammer is these chip designs don't NEED to move to astronomical clockspeeds to retain performance parity. So, while P4 could reach 3 ghz possibly in the next 12-18 months, Athlon should be able to offer similar performance in the low 2ghz clockspeed..and I believe that may be in the cards for Athlon this year. So, while some people may buy the latest and greatest P4 because it's the *fastest*, other folks will buy Athlon because it offers the most bang for the buck.

If that's not a hypothetical scenario, I'd sure like to know what you're sources are. :lol: 
I'd particularly like to know how you know that:
- a 2Ghz Hammer will outperform a 3Ghz Northwood with Hyperthreading and a 533Mhz FSB.
- a Hammer processor is going to be cheaper than a Northwood processor of similar performance.

:wink:

"Ignorance is bliss, but I tend to get screwed over."
January 5, 2002 9:55:17 PM

well if you watch Intel and Amd it seems to be Leap frog. 5 to 6 months Amd faster then its Intel turn to be faster. I hope it keeps this way. It make the price Lower. I have a feeling Intel chips are going to be about $300 Lowend to $600 highend for the new chips. But that watching the price from the last hype. When Intel took over Amd with the 2.0 ghz And Then Amd chips comming out.
January 5, 2002 9:58:00 PM

To the subject of memory bandwidth, when the 128bit DDR ram is out and cheap to buy, that will double DDR's bandwith as it is now? Because right now the DDR ram is 64bit correct?

if(GetSystemMetrics(SM_PROCESSOR) != AMD_PROCESSOR)
{
SendMessage(hwnd,WM_CLOSE,0,0);
}
January 5, 2002 10:06:57 PM

DDR RAM currently has a 64 bit data path. If you had a 128 bit data path, or two 64 bit data paths (such as the NForce has) you will get double the bandwidth.
However it's not good enough for the BUS and RAM to offer that amount of bandwidth, the CPU has to be able to as well. So unless your Athlon CPU is modified to accept 128 bits of data (or maybe 64 bits at twice the clock speed), it will make very little difference.

"Ignorance is bliss, but I tend to get screwed over."
January 5, 2002 10:19:35 PM

I agree I don't think AMD is just going to lie down and become old news, their *ss will be on the line.

Complicated Nit Picker
January 6, 2002 1:10:08 AM

"I'd particularly like to know how you know that:
- a 2Ghz Hammer will outperform a 3Ghz Northwood with Hyperthreading and a 533Mhz FSB.

I never said that. I said a 2Ghz Athlon *should* be able to. This is all suppositions based on currently available benchmarks and extrapolating upward.

- a Hammer processor is going to be cheaper than a Northwood processor of similar performance."

Again I made NO reference to Hammer price. I specifically mentioned ATHLON pricing.

Just wondering...do you know something about HAMMER speed/performance pricing that I don't? Right now the only performance info I can go on is the comments AMD reps have made and none have mentioned pricing yet. But maybe you know something about this and are trying to determine if I know the same thing??

When all else fails, throw your computer out the window!!!
January 6, 2002 4:49:13 AM

I honestly believe that Intel will make the gap between the ALU and FPU performance even greater than it is now.
Most apps I use are rarely limited by the ALU performance, FPU is holding them back.
the P4 is the K6 2 of the Intel lineup. If you played a game that was 3Dnow enabled, it did pretty damn good, otherwise, you'd have to run less detail or a full color depth or resolution lower than an equally matched PII.
The difference here is Intel is the one releasing the sub par FPU and they are still the industry leader, so I still think the P4 will become the "standard" CPU, much as the Intel P6 is now. before AMD fans start bagging on that last comment, prove to me that the P6 core is not THE most popular processor of current times. PII, Celeron, PIII, Tualatin, its everywhere in one form or another.
Then, the other side of the coin, eventually the P4 speeds will exceed everything else(should anyway) to the point that is doesn't matter how strong the FPU is.
an analogy, same car, same weight, roughly the same horsepower curve, just 4000 rpms higher than its competition, which one reaches the end of the track first?

----------------------
Independant thought is good.
It won't hurt for long.
January 6, 2002 4:53:46 AM

AMD expects to release the Clawhammer somewhere around the 3400+ mark, using its current Performance Rating system. The Clawhammer will have an integrated memory controller, drastically reducing memory latencies. It will also use Hypertransport to offer 6.4GB of bandwidth per CPU initially, with the ability to quickly and cheaply ramp that up as memory technologies advance. In a multiple-CPU configuration, each CPU can have up to 3 hypertransport connections, resulting in 19.2GB of bandwidth between CPU's. If I remember correctly, when processing 32bit code it operates similar to a dual processor system, only transparently. At this point, nothing indicates that the Clawhammer won't be able to meet AMD's projected performance rating.

The Clawhammer will be introduced at .13mu using SOI. Its projected die size (I can't find any links right now, it was either 80mm^2 or 64mm^2, with the .13mu Athlon being the other) will actually be smaller than current Athlons. This should allow AMD to price them very near current Athlon prices. The integrated memory controller should result in cheaper motherboards as well, though the cost of hypertransport-everything may counter this.

Of course everything here is simple conjecture, but given AMD's excellent execution with the Athlon, I am looking forward to the first benchmark results to say the least.

-= This is our wading pool.
Stop pissing in it. =-
January 6, 2002 8:00:08 AM

<blockquote><font size=1>Svar på:</font><hr><p>Intel should go straight to the high frequencys, AND lower all of their prices dramatically? You may as well say AMD should go straight to the Hammer, and give their processors away for free. :p <p><hr></blockquote><p>My point is: Intel could do just that, while AMD is not in a position where they could start releasing the Hammer; they're simply not ready yet.



/Copenhagen - P4 Willamette 1700MHz@2109 MHz, Vcore 1.75V@2.20V on Abit TH7II-RAID.
January 6, 2002 11:17:31 AM

Quote:

You wouldn't rather have the free Hammer?

Oh yeah! I'd go out an get a dozen or so of them!

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
January 6, 2002 11:17:40 AM

Quote:

You wouldn't rather have the free Hammer?

Oh yeah! I'd go out and get a dozen or so of them!

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor
January 6, 2002 11:32:25 AM

It would be Barton most likey. I mean once Northwood B comes out with 533 FSB and higher speeds, the Athlon would see some serious competition. Of course AMD is going to respond to this no doubt.
Woah and imagine SMT enabled, ohh [-peep-] damn Northwood would spank. Seeing around 20-30 % performance boost. Jackson Technology also can be called Hyperthreading.
SMT is already enabled in Prestonia. What SMT is that it acts like TWO processors but its really just one.

Im not sure about the AMD's roadmap yet.
Also of course Northwood and Thoroughbred would be overclocking beast due .13 micron.

GamerzCitadel.com
January 6, 2002 11:42:22 AM

Hyperthreading is not as big as everyone hypes it out to be.

Got THAT from the mouth of an intel software engineer in the family!

"The Cash Left In My Pocket,The BEST Benchmark"
No Overclock+stock hsf=GOOD!
January 6, 2002 12:23:58 PM

Mat, I agree. I've read several previews with benchmarks. While these could be wrong, I don't think so.Apparently, a lot is wasted in communicating between the threads..to the point that some benches came out slower than a like speed Northwood P4.

So, Hyperthreading has potential. I really think it does. But there are a lot of issues to work out before it will start exhibiting that potential. I don't expect that potential to really be seen for at least another year, if not longer, based on the similar ramping of the P4 before IT finally has started to show some real power.

Mark-

When all else fails, throw your computer out the window!!!
Anonymous
a b à CPUs
January 6, 2002 12:34:04 PM

I wont argue on HT performance b4 I see any good benchmarks.. but it sure as hell is a smart idea, and a cool technology. Wondering if AMD will implement it in Hammer or some other cpu, as it doesnt appear to be so hard to implement (at least, it doesnt require a whole lot of die space)

= The views stated herein are my personal views, and not necessarily the views of my wife. =
January 6, 2002 12:58:26 PM

Im not debating ht's performance either, the guy I talked to gave from 5-30%, but he said its closer to 5 than 30.

We will see however.

"The Cash Left In My Pocket,The BEST Benchmark"
No Overclock+stock hsf=GOOD!
January 6, 2002 5:47:33 PM

I don't know how HT will perform, but word is Northwood will not have it.
as that goes, it probably will, but near the end of the lifespan of the P4...

----------------------
Independant thought is good.
It won't hurt for long.
January 6, 2002 6:15:34 PM

Apparently Microsoft aren't too pleased about Hyperthreading, due to complex licensing issues. Should Microsoft treat a single hyperthreaded P4 the same as a single threaded system ? Software lisencing and providers will also need to work out how to bill customers with hyperthreaded enableded software/systems.

<font color=purple>~* K6-2 @ 333MHz *~
I don't need a 'Gigahertz' chip to surf the web just yet ;-)</font color=purple>
January 6, 2002 7:27:44 PM

I don´t agree you can´t compare the 2 situations. Today Amd is much stronger and could be prepared to go their own ways when it comes to cpu achitecture.
January 6, 2002 7:36:08 PM

I like AMD, I use AMD, I recommended the mass purchase of AMD systems, but they are NOT an industry leader.
Fact is, crappy as it may sound, AMD has better technology, imho, but Intel still leads the market. Until AMD has at least 50% of the market, don't expect differently.
anything else would be lying to yourself for the sake of being a fan of a company that exists only to make money.

----------------------
Independant thought is good.
It won't hurt for long.
January 6, 2002 10:13:44 PM

I can assure you that AMD's "performance rating" does not accurately reflect real world performance: A 3400+ processor will not be exactly 70% faster than a 2000+ processor. The Hammer looks great on paper, but so have a lot of things that have dissapointed.
Don't get me wrong, I'm sure the Hammer will be a great processor, but I won't expect it to leave Intel bleeding on the ground. The Pentium 4 is a processor with a planned future, it's got a whole lot of room to grow.

I think the Hammer was going to be 64mm^2 at .13mu, so it could easily be sold at a low price. On top of that, the intergrated northbridge will reduce the complexity of both motherboards and chipsets. The price performance of the Hammer could be very nice indeed.

I'm also looking forward to the first beanchmark result.

"Ignorance is bliss, but I tend to get screwed over."
January 7, 2002 12:45:25 AM

Quote:
I can assure you that AMD's "performance rating" does not accurately reflect real world performance: A 3400+ processor will not be exactly 70% faster than a 2000+ processor.

AMD's performance rating does reflect real world performance, that is its whole point, rather than relying on now arbitrary GHz ratings to specify speed. If rated at 3400+ at release, it should perform close to the same as an Athlon rated at 3400+. As the performance rating currently seems to use the P4 for comparison, albeit humbly, you should expect a 3400+ Clawhammer CPU to perform at least as well as a 3.4GHz P4.

-= This is our wading pool.
Stop pissing in it. =-
January 7, 2002 2:50:59 AM

First, the performance rating supposedly compares the AthlonXP to the Thunderbird. It is not supposed to compare AMD processors to Intel processors.

Second, AMD's performance rating is chosen by AMD to suit their purposes. You can't expect a benchmark suite chosen and executed by AMD to be a truly accurate representation of performance.

Third, the AthlonXP 2000+ (1.66Ghz) has a clock speed 25% higher than the Athlon1500+ (1.33Ghz), yet it's "performance rating" is 33% higher. Therefore it's rating can't be accurate unless the 2000+ has other improvments over the 1500+...and it doesn't.

I'm afraid that AMD's "True Performance Initative" is just another way for marketing to dumb down the masses.


"Ignorance is bliss, but I tend to get screwed over."
January 7, 2002 6:37:05 AM

Quote:
Second, AMD's performance rating is chosen by AMD to suit their purposes. You can't expect a benchmark suite chosen and executed by AMD to be a truly accurate representation of performance.



Which is independantly verified by a third party, there is no cheating going on here.


Quote:
Third, the AthlonXP 2000+ (1.66Ghz) has a clock speed 25% higher than the Athlon1500+ (1.33Ghz), yet it's "performance rating" is 33% higher.



Each mhz on a palomino is worth a certain amount of pr points, this is determened in relation to a tbird's clock speed. As cpus get faster, they tend to gain performance in a non linear fasion.( a 1100mhz cpu is never exactly 10% faster than a 1ghz cpu in benchmarks). And if a cpu's design is good, it will gain more from having more clock speed.(this is just one possible explination for the increased pr rating)



"The Cash Left In My Pocket,The BEST Benchmark"
No Overclock+stock hsf=GOOD!
!