2200+ over 2100+?

[BetaBoy]

Hi I am building a new system and confused on the whole 2200+ fiasco. With new system (not upgrade) which is a better choice to go with? 2100+ for $170 or 2200+ for $230? Is the performance difference that noticable? Also on the motherboard choice, what mobos support 2200+ now? I was thinking of going with Asus A7V333, but not sure if it supports 2200 or not.

Thanks for all your help.

 

[tritium4ever]

Hmmm, the 2200+ costs 35% more ($230 vs $170) but is only 3.9% faster (1.8GHz vs 1.73GHz). You do the math...

 

[Mingon]

The 2200 is based on the newer 0.13 micron die, it will run cooler and might o/c further

 

[Mikewarrior2]

Mingon,

That's not necessarily true... The smaller die size + very minimal wattage drop, coupled with most heatsinks, would yield probably equal if not higher temps for the Tbred.

As several reviews have said, high end tbreds are very hard to cool.



Mike

 

[PowerMacG5]

The TBred will probably run hotter than the Palomino with most HSF's. The TBred has a smaller contact patch, and even though it is .13u, it is still running at a high voltage. The reason P4's run so much cooler than Athlon's is for two reasons, 1) It has the IHS, and 2) Intel has a lower voltage on their processors than AMD. The P4 runs at 1.5V. The TBred has nothing but a die shrink to offer the CPU community right now. TBred is nearly identical to the Palomino except for it being made on .13u. It has no extra cache. Intel was smart, and so when they released Northwood, they released it on .13u with twice as much cache.

 

[Mingon]

OK I will rephrase what I said to the AMD on 0.13 micron produces less heat. although as you have said many times the contact area has reduced the t'bred should still run cooler clock for clock, if it doesnt due to incorrectly seated heatsinks that another issue.

 

[Duvie]

Does the actual .13 micron have anyhting to do with it necessarily??? Or is more the connectors and the type of material used that has more to do with it....I have heard the smaller the size (ie .13 vs .18) the more sensitive it is to higher voltages...

I am sure I got parts of this all messed up....bottom line amd needs to incorporate a IHS like intel as well as use it SOI technology....

I heard speculation that intel switched to copper with the northwoods while it used aluminum in the willamette...did I get that right??? And furthermore the athlon was already using copper and hence the vcore drop from the palomino was not as much then the northwoods (who had a material change) and thus the temps aren't as much a decrease from the palominos....Throw in that they released an increased speed of 2200+ producing more watts then the 2100+, then the smaller contact area, and its spell not much of a thermal benefit...

Doesn't copper connectors insulate and conduct better??? Hence maybe why less vcore is required, as you lessen the dampering of the power, as well as less heat being admitted in critical areas???


I would like to be more technically aware of this, so anybody feel free to correct me, tell me I am smoking crack, whatever....Just inform me....

 

[Mikewarrior2]

Mingon,

The issue is beyond improperly mounted heatsinks. Heatsinks with questionable base quality, and most of the all aluminum units will run hotter clock for clock with the XP2000+ and higher tbred versus palomino chips. That is the end-result of the high heat output + small contact surface. High end heatsinks should have no issue with this, but not everyone uses a high end heatsink.


Duvie,

.13 micron cannot take as much voltage as .18 micron chips... .I do not know the specifics, but running 2.0V on a .13 chip is not recommended my many people at all .

Yes, intel went to copper interconnects with Northwood, with Aluminum Interconnects with Williamate. This, among other enhancements, creates teh northwood heat drop versus williamette heat drop despite smaller die size etc. For example, a northwood at Williamette Vcore would not run as hot as a true Williamate, and a Williamette with as low a vcore as a northwood would not run as cool as a Northwood (phew, need a breath or air ). That information i've gotten from several sources.

If you look at the Tbred, there's no sign of improvements to the core besides being a (simple) (or not so simple) .13 micron process. If you took a tbred and ran it at palomino vcore, it would undoubtedly run hotter. If you ran a palomino at tbred vcore, it would run cooler than the tbred. So in that sense, there wasn't enough improvement to the tbred core to really see northwood type cooling and overclocking.

hoep that helps a little, and I hope that someone with more knowledge about copper interconnects, etc also posts.


Mike

 

[BetaBoy]

So back to my original question, what is the recommendation about 2200 vs 2100? And what are the mobos out there that support 2200? Thanks

 

[BD231]

Originally posted by: tritium4ever
Hmmm, the 2200+ costs 35% more ($230 vs $170) but is only 3.9% faster (1.8GHz vs 1.73GHz). You do the math...

I'll ad this, if a 35% cost increase is worth the 3.9%, knock yourself out.

 

[Cruze8]

BetaBoy, here is my opinion on your situation, Personally I'd go for the 2100+ for $170 rather than the 2200+, the performance between the two is almost identical, also tbred's are new and as of yet are having quite a few issues with heat and overclocking. Now for a motherboard I Highly reccomend the Asus A7V333 (Non Raid Version) It will support the thouroughbred processors with a simple bios upgrade, the board is very stable and has quite good overclocking options, also it reads the temps right from the AthlonXP or Thouroughbred Internal temp diode, and it also features Overheating protection, a standard which amd is about to set for mobo's while running the thouroughbred cpu, for instance if you have a board that doesn't have overheating protection and you put your cpu in there you void the warranty. But if you don't care for the Asus Mobo any of the other kt333 boards from the major manufacturers will serve you just as well, hope this helps ya

 

[BetaBoy]

Cruze thank you very much for your feedback. I actually would like the raid controller, what would you recommend for the raid controller?

 

[Cruze8]

head on over to anandtech's mobo section, they have a great kt333 mobo shootout, it will give you the best info about your mobo choice, I only knock asus for thier badly implemented raid support, also I've heard people having probs with asus's raid, that's why I got the non raid ver.

 

[Mingon]

The issue is beyond improperly mounted heatsinks. Heatsinks with questionable base quality, and most of the all aluminum units will run hotter clock for clock with the XP2000+ and higher tbred versus palomino chips. That is the end-result of the high heat output + small contact surface. High end heatsinks should have no issue with this, but not everyone uses a high end heatsink.

Sorry i dont agree, given the same heatsink a 0.13 micron xp will run cooler than a 0.18 xp unless it is improperly mounted. whilst the contact area is smaller and the transfer of heat is more focused on an area I dont think it will be an issue unless their is a problem with the contact area. The question of base quality is difficult, the majority of computers that are supplied with an xp chip will have decent cooling system, so the issue only effects people building their own machines and choosing poorly rated heatsinks.

 

[Cruze8]

I disagree, it's harder for a heatsink to dissapate heat with a small surface area, hence the reason the tbred is having problems I think. if you look at some reviews that have the heat wattages thrown off by a palomino 2100+ and a tbred 2200+ the diffrence is only a few watts diffrence in favor of the tbred, why do you think the p4 can stay so cool? 1)it runs lower voltages than the tbred and 2)the ihs on the p4 allows the heatsink to dissapate the heat more quickly, think of the bed of needles trick, if you lay on one or 2 needles it will poke right through you, however if you have a whole lot of pins it won't hurt at all.

 

[Mikewarrior2]

Sorry mingon,

You're wrong.
Frosty Tech Heatsink reviews

Regardless of heatsink, the smaller contact surface tests result in higher temps than the large contact area's. NO, the frostytech die sizes aren't exact analogs, and the temperature differences are probably a bit exagerrated than will be between palomino and Tbred, but the fact remains: Small die size hinders heatsink performance.

I mean, even look at the respectable Swiftech MCX370 and it still shows a significant difference between the two die sizes.

And to further show some data, let's apply 80W of heat to the small contact surface, and 100W to teh large contact surface.

arkua 628: Large contact surface = 57.5C. Small contact surface = 50.88 C. So this heatsink has good large/small die size performance.

swiftech mcxc370: Large contact surface = 53.5C. Small contact surface = 52.56. Similiar performance between teh two die sizes, but the small die size is 20W cooler.

Taisol CGK760 (heatsink inlcuded in tbred reviews): Large contact surface = 61.2C. Small contact surface = 62C. Again, comparable temps but the small die size is 20W cooler.

Now, let's examine the difference between an XP2100+ palomino and an XP2200+ Tbred. How many watts diference? ~4C.

And do I really need to go back and quote several (in fact, most) Tbred reviews from monday that said the same thing i'm saying? And for you to say that system integrator's use quality materials is extremely laughable. I've seen many computers built with subpar, junk OEM heatsinks for the purpose of saving $3 over a coolermaster.

When you shrink the die size of a CPU, and shrink the corresponding contact surface, you concentrate the heat transfer of the primary heat pathway. Depending on the heatsink, some will choke up when trying to spread out 65W of heat over an 80sq mm contact surface. That same heatsink might not have an issue doing the same wattage with a 128 sq mm concact surface. WE're not talking insignificant die size change. its a drastic drop in die size.


Mike

Again, higher end heatsinks should have no issue with tbred 2200+'s running hotter than XP2100 palomino's. I'd say everything from Alpha pal8045s, swiftech models, thermalright models, higher end copper models will be fine. But i've seen hundreds of cheap heatsinks with very poor base qualities, noticeable machining marks, etc. On these heatsinks, including thermaltake units, the contact surface could potentially be further reduced due to poor base quality. With the heavy machining, what if hte effective contact surface is now 70 or 60 sq mm. Even with a good grease, this would be a huge heat issue. You're not alking aobut 70W off of 70sq mm. Again, it will make a difference on those heatsinks.

 

[Akaz1976]

Are these issues with t'bred with the 2200 only or do they extend to lower clocked vesions as well? i have been considering buying 2000XP tbred but this discussion has made me cautious

Akaz

 

[dunkster]

I've been thinking about upgrading my crappy AGKGA-stepping XP1600, and I don't like either the XP2100 or XP2200.

The xp2200 is a kluge, for all the reasons expressed above. The only value I see in the XP2200 is that, hopefully, it will lead to reduced prices for Palominos.

The xp2100 is a kluge also, with respect to unlocking the multiplier (cut bridges?).

That leaves the XP2000 as my preferred candidate - to tide me over 'til Hammer time.

Here's a question:
I assume that all xp2000s have later steppings, and buying an XP2000 should be less of a crap-shoot than than the early Palominos. True? False?

Thanks for your thoughts!

 

[Bink]

Originally posted by: tritium4ever
Hmmm, the 2200+ costs 35% more ($230 vs $170) but is only 3.9% faster (1.8GHz vs 1.73GHz). You do the math...

I don't think this is the appropriate way of framing the decision (i.e. in terms of percentages). Your implication is that it is not good value (which may be true if you assume you want the cheapest MHz/$), but ignores the productivity (or other) benefit of the speed increase. A better way of framing the question is, "Is a 3.9% speed increase worth $60 to you?", then the answer depends on the application. If a PC is used continuously for number crunching (or CAD, etc), then $60 over even a 1 year life may save considerable time. Is it worth it? Sure. If the application is logging on over a 56K modem once a week to check email, the answer is no.