Will Thoroughbred overclock as well as Northwood?

[New2Computers]

I'm going off to college, and was going to wait until the fall to build a pc... but have decided that I would enjoy it more to build relatively (a month or so) soon.

I'm basically going to wait until Thoroughbred is released before I build I guess, so I can see if it will have the same overclockability (is that a word heheh) as the northwoods have had. Even if they can't overclock as well (it would be naive to expect near 50% speed overclocks) will it have significantly more headroom then the current XP models?

I don't really want to run the risk of getting a bogus northwood, when I could have just fronted a little more money for an XP 2000 or so for guaranteed speed. Then again if its going to save me money, and the overclock seems almost guaranteed to break the 2ghz barrier..

Noise is slightly important to me, although I've heard very good things about AX-7 and panaflo fans, and honestly if that is a little louder then an northwood set up would be, its definately not anywhere near the end of the world.

I'm kind of stumped.

Anyway, what do YOU think the new cooler running athlons will be like in the overclocking scene? Can we expect some headroom? I'm interested to hear what you guys believe. If they will overclock nicely, that would definately seal the deal on my decision.

On a side note, how do the overclocked 1.6A (anyone who has gotten 2.2ghz or so) compare to speeds of an XP 2000? Anyone had a chance to use both and see if there is any noticable difference, or benchmark to see those differences?

BTW, I'm currently running a 1.2ghz thunderbird, not overclocked on a MSI K-7 Master w/ a Geforce2 Ultra and 256mb DDR ram.

 

[BFG10K]

A 0.13 um process should give it some overclocking headroom but it depends on critical paths in the pipeline to see how well it overclocks compared to the P4.

So yes, it'll definitely be better than the XP models but you'll have to wait and see how well it does compared to P4s.

 

[ST4RCUTTER]

I would be greatly surprised if the Tbred has near the overclocking potential that the Northwood has. You have to realize that the 20-stage pipeline was pretty much designed for one thing...high Mhz. At .18um, the Willamette was really being held back by its manufacturing process. Moving to the .13um node has definitely turned the P4 into a market leading product. The .09um node should provide another large boost in speed towards the end of 03'.


I'm not saying that the Tbred won't overclock well...far from it. I imagine the P3 would be a good yardstick for those wondering how fast the Tbred will overclock to.

 

[Rand]

It's a virtual certainty they won't o/c anywhere near as well as the Northwood, even on .18u it still had potential to go beyond 2.0GHz without too many problems.... a full die shrink and re-layout of the core has only served to enhance it's head-room. The P4 core is designed to ramp in clockspeed like no X86 processor before it and we're seeing the impact of this.
In addition Intel's fabrication process is quite obviouly more advanced then AMD's and they spent considerable time gearing up for the re-tooling of their lithography equipment ahead of time.
The packaging is also better equipped to handle higher clockspeeds then that of AMD's processors. And Intel gained the befefit of copper interconnects along with their .13u process, AMD had already implemented copper previously.
As another added bonus the P4 still stands to gain considerably from improvements in process technology as their nowhere even remotely near the core's limits, and likely won't even hit the P4's 'sweet spot' in their process design until .09u or .08u.
On the .13u process I honestly expect the P4 to be limited by thermal characteristics before it reaches it's potential clockspeed.

This compared to a K7 core that while being rather nicely clockable compared to prior designs, it still merely a pale shadow of the of the almost exclusive attention Intel paid to pure clockspeed. While packaging constraints havent come into play yet for AMD and likely won't for a long time yet their packaging isnt as advanced as that of AMD's. They've had some mild electrical issues early on with their .13u process which has served to delay it from it's initial release expectations.

Amd's current .13u implementation seems to be little more then a base dumb shrink with few modifications, so it's likely not going to gain as much from the .13u process as it might otherwise.
And with the Palomino core they seemingly decided to focus on the 'Braniac' approaach of improving clock for clock performance rather then Intel's 'SpeedRacer' approach in sacrificing efficiency for sheer clockspeed.
Such leanings may have slightly inhibited the K7's potential clock ramping.
Beyond this the K7 core is past it's peak, and has likely reached the point of diminishing returns in each successive process improvements.

They've likely been pulling resourced away from the K7, and putting those into the K8's design which bodes well for the future but likely limits the K7 in the present.
In short I don't expect the K7 on .13u to drastically improve the clock ramping of the Athlon, it's still a relatively immature process for AMD. In the long term with future improvements to the process and minor core re-layouts it may have a decent impact but in the immediate future I don't expect their initial .13u processors to o/c much more then 100-200MHz further in the best case scenario then their clearly very mature .18 process and core layout that's likely heavily optimized for their present .18u process.
And then there is the perennial rumours that AMD already had more then the typical small few elements that were already .13u or lower in their .18u process... which would serve to reduce the benefits seen from a full shrink to .13u. A rumour I don't at all believe, but is worth mentioning.

I don't really want to run the risk of getting a bogus northwood, when I could have just fronted a little more money for an XP 2000 or so for guaranteed speed. Then again if its going to save me money, and the overclock seems almost guaranteed to break the 2ghz barrier..

On a side note, how do the overclocked 1.6A (anyone who has gotten 2.2ghz or so) compare to speeds of an XP 2000? Anyone had a chance to use both and see if there is any noticable difference, or benchmark to see those differences?

Typically I put an AthlonXP 2000+ as being on-par with a Northwood 2.2GHz + DRDRAM, but the higher FSB of an overclocked 1.6a would probably give the 1.6 @ 2.2 a small but tangible performance advantage.
Then again, the 2000+ is guaranteed at that speed and can usually hit 2100+ speeds... and while the 1.6A is a ery nice overclocker there are still a reasonable amount that don't want to do any better then 2.13GHz and sometimes not even that.

The Athlon is the 'safe' bet as you know your getting that level of performance and likely a small bit more from overclocking.
The P4 is a more risky choice, as you may get a poor 1.6A that won't even do 2.0GHz, though the overwhelming majority do at least 2.13. But that risky choice could pay dividends as a decent number of 1.6A's hit 2.4GHz and a small number even a bit beyond that. The 2000+ almostc ertainly won't overclock far enough to reach such performance levels.

In the end both offer excellent levels of performance, so it'll likely come down to extraneous matters such as cost, whether your willing to take a risk or what the 'safer' option, and possibly future room for processor upgradeability without changing the motherboard.

If it was me I'd probably take the chance and go with the P4 1.6A and overclock.
UNLESS I already had a DDR SocketA motherboard, in which case I'd just buy the 2000+ as it would be easier, less problematic as all you'd be doing is replacing the processor rather then processor + motherboard with the Intel option. Also it'd possibly be a bit cheaper depending on which P4 motherboard you chose, while offering performance that shouldnt fall TOO far beyond the P4 even if you get one that overclocks very well.

 

[Rand]

At .18um, the Willamette was really being held back by its manufacturing process.

I disagree, the 2.0GHz Willamette's were generally overclocking to 2.2GHz with regularity, and Intel was clearly getting very good yields on 1.7GHz+ Willamette parts. I think the very high thermal requirements of the 2.0GHz Willamette and the readiness of Intel's .13u process which had already had a decent amount of time to mature with the .13u Tualatin was the main reason the Willamette didnt go beyond 2.0GHz. That and the fact that at 2.0GHz Intel was already well ahead of AMD in the eyes of most people to whom MHz equated to performance, so they had little competitive reason to push the Willamette core and further.

 

[ST4RCUTTER]

I think the very high thermal requirements of the 2.0GHz Willamette...was the main reason the Willamette didnt go beyond 2.0GHz.


I agree with the gist of that statement 100%. I don't believe the .18um process was preventing the P4 from ramping above 2.0Ghz. I just don't think Intel wanted to sell 85W processors that doubled as space heaters. AMD was already in that market.

 

[Rand]

Originally posted by: ST4RCUTTER
I think the very high thermal requirements of the 2.0GHz Willamette...was the main reason the Willamette didnt go beyond 2.0GHz.


I agree with the gist of that statement 100%. I don't believe the .18um process was preventing the P4 from ramping above 2.0Ghz. I just don't think Intel wanted to sell 85W processors that doubled as space heaters. AMD was already in that market.

Well when you consider the Willamette 2.0 was pumping out a good 75W regularly, and over 100W peak I'd say it was actually worse then anything AMD's released.
While the heatspreader probably helped that's still a good 10W typically over anything from AMD, and more then 30W peak maximum consumption over that of AMD's hottest processors. While due to the ability to thermally down-clock it wasnt likely to ever reach 100W it would still at times produce way more heat then the widely regarded extremely hot running T-Bird 1.4GHz which is the hottest running processor AMD's yet pulled out.

Especially when you consider the 1MB L3 Xeon MP at 1.6GHz was already hitting close to similar thermal levels as the 2.0GHz Willamette I'd hate to imagine what an even higher clocked Xeon MP would be like.

I expect the Northwood to be able to pull up to 3.4GHz in the future before thermal limitations come into play.... that's assuming no voltage hikes are seen of course. I think we will probably see a voltage hike somewhere between 2.8-3.2 though which will probably drop that number a bit.

 

[Innoka]

It would be trash to try to predict the clockspeed potential of the Thoroughbred. The Thoroughbred will cost more until it has completely replaced the older XP, as it will be in shorter supply and only at the higher speed grades at first.
If you knew what chips used to cost, you'd see the Athlon now is a very good deal. Why don't you worry about the performance for your own needs rather than looking over other people's shoulders at their stats.

 

[merlocka]

It would be trash to try to predict the clockspeed potential of the Thoroughbred

LOL, Rand's analysis it pretty reasonable... far from trash. He's probably right on the money. A thread about overclocking and process headroom and you crap all the decent reasoning and talk about how much the Thoroughbred will cost.