Thoroughbred @ 2Ghz 6-10-02

[CrazySaint]

The reason Intel's flagship still spanks AMD's is that until Monday AMD hadn't released a single processor in about 3 months. At the time the 2100+ came out, it was the performance leader, albeit by a small margin. Since that time, Intel has increased clockspeed by 533MHz on the Northwood core, increased clockspeed by 333MHz over the fastest Willie core, upped the FSB speed by 133MHz, and released several new chipsets. No way in hell a single 66MHz speed increase by AMD was gonna overcome all that. The TBred's only hope of re-capturing the enthusiast market was if it OC'd as well as we were hoping - it doesn't. So unless AMD significantly improves its process and OC goes up soon, Intel will hold the speed crown virtually unchallenged at least until the Barton comes out. Whether the Barton's 512k cache will be enough for AMD to catch back up again is anybody's guess, but right now even that looks doubtful unless the TBred and Barton somehow manage to scale really well. So I hate to say it, but right now, it looks like Intel may well hold the performance crown at least until the Hammer comes out.

 

[WarCon]

This is the whole article now.



There wasn?t much information here. Gigabyte had a relatively short presentation but there were 2 points they touched on. The first is that they?ve been expanding into networking in the last 2 years; they had several routers, switches and hubs on display. But the more interesting point is that with their overclocking utility, Easy Tune 4, they were able to get a Thoroughbred 2200+ to 2.6-2.7 GHz, quite the impressive feat. If this is the norm for these chips Pentium 4?s definitely have a run for their money.

EDIT: The first reviews of the Thoroughbred 2200+ are out and none of them are Overclocking to anywhere near 2.7 GHz however they are achieving what would be equivalent to 2500-2600+ so it appears Gigabyte was only able to get it up around 2.1 GHz, still an impressive feat but nothing like an Overclock of almost 50%, more like 20% Apparently this is a case of PR guys mistaking some information from engineers.

Posted for completeness sake alone. Even at 2500-2600+ speeds these processors will be more than plenty fast. I personally don't care which processor manufacturer takes the lead as long as they don't get sloppy trying to catch/stay up with the competition.

(I may get flamed for this prediction, but I see a lot of cores getting crushed because of the size and shape. They are similar to duron cores (in shape and size) and I have read about many duron cores getting chipped and crushed. Smaller means easier use as a fulcrum for a pivot. Pivoting heatsink = chipping. JMTC)


...........edit.......
To italicize the article.

 

[DivideBYZero]

MajinVegeta originally stated:

The truth is Hammer is in the same boat. deeply stacked pipeline only work whe they are short and al lower speeds then the TLB predictor dies out.

AMD needs a new core the problem is they havent designanything new just an add on AMD is going to have to 1: adopt the P4 cores basic design 1 signle long pipeline, or 2: totally overhaul the core and see what they get.

Hammer is on the same boat? Which boat is this exactly? Have you overclockled your Hammer? Nah, didn't think so.

Wake up, AMD have a new core, it is Hammer. I cannot stop laughing at your suggestion that AMD copy the P4 architecture...

 

[Duvie]

There was obviously some reason why the majority of the large sites remained silent on their thoroughbred reviews until sunday and monday...probably NDA's...meaning the likeliness of misinformation by earlier sites...

I also think the edited post of 2200+ ocing to 2400-2500+ seems to be a far stretch looking at these credible sites and there oc attempts. Right now it looks like fab results getting those maybe tough and I thought the 2300,2400 & 2500 are all to be released shortly here after the 2200+...

 

[MajinVegeta]

Hammer is a AXP with 64 bit op codes, 40 bit physcal 48 virtual memory adressing, a better TLB predictor, Intgraeted Memory Controlle, Extra FPU pipeline, and 2 extra stages in the ALU and FPU pipes.

Thats what boat it is in, ifs Half Itanic half P4. It's stacked very wide and deep so the TLB predictor is maxed out trying to deal with 5 different pipelines 2 Alus 3 Fpus and they are long 13 and 17 stages.

Plus here the other issue that no one has addressed

transistors per mm^2

P4 @ .18 w/ Alu 194,000
AXP @ .18 w/ copepr 291,000
P4 @ .13 w/ copper 420,000
AXP @ .13 w/ copper 460,000
K8 @ .13 copper w/ SOI 600,000+

thats a lot more heat per Sq mm before thats why it wont scale or OC to much thermal load on a small die space that cant disapate the heat.

 

[jbond04]

Originally posted by: MajinVegeta

thats a lot more heat per Sq mm before thats why it wont scale or OC to much thermal load on a small die space that cant disapate the heat.

Umm... There's a technology to fix that problem--it's called a heat spreader.

Don't get me wrong, though. I am very skeptical about AMD's vaunted Hammer. I don't think that there will be a giant leap in performance over what a P4 can offer. Besides, who wants to buy a 2GHz Hammer in June of 2003 when you can wait just a little longer for Prescott? (Just a little jab at AMD's release dates. Can you say Tbred in Q1 2002? )

 

[mechBgon]

Originally posted by: jbond04

Originally posted by: MajinVegeta

thats a lot more heat per Sq mm before thats why it wont scale or OC to much thermal load on a small die space that cant disapate the heat.

Umm... There's a technology to fix that problem--it's called a heat spreader.

To make a long story short, the heat spreader is there... it's the copper plate forged into the heatsink.

Whether the face of the die is touching an integrated heatspreader or the heatspreader in the bottom of a heatsink, it has the same physical area through which the heat must be moved. Adding more layers to the path that the heat takes from the die to the heatsink fin will inhibit thermal transfer, because the thermal-interface material joining the layers (namely, thermal grease or PCTC) has a higher resistance to heat transfer than the materials which it is connecting (copper or aluminum). If the thermal-interface material had a lower thermal resistance than the heatspreader (say, if you could magically fill the gaps completely at the atomic level with silver or monocrystaline carbon [diamond]), then it might be otherwise.

This is why some Alpha 7125 owners removed the heatplate from their old SlotA Athlons and ran them "naked," in order to cool the CPU die directly instead of by proxy.

Heatspreaders do protect the core from physical damage, and I'm guessing the smaller die will be easier to damage, as has been said by others.

 

[MajinVegeta]

Exacty mech I could see the space to disappate the heat is the same w/ or w/o the IHS. So the core size is still important to transfer the heat away. BBUL could help by allowing Intel to spread the chip out some inside the package and allow the to configure more speciality processors but then distance of the connections could become and issue.

But Imagine being able to call up Intel ya I want my P4 with 2 Megs L2 cache, and quadruple L1s.

 

[SSXeon5]

Originally posted by: MajinVegeta
Hammer is a AXP with 64 bit op codes, 40 bit physcal 48 virtual memory adressing, a better TLB predictor, Intgraeted Memory Controlle, Extra FPU pipeline, and 2 extra stages in the ALU and FPU pipes.

Thats what boat it is in, ifs Half Itanic half P4. It's stacked very wide and deep so the TLB predictor is maxed out trying to deal with 5 different pipelines 2 Alus 3 Fpus and they are long 13 and 17 stages.

Plus here the other issue that no one has addressed

transistors per mm^2

P4 @ .18 w/ Alu 194,000
AXP @ .18 w/ copepr 291,000
P4 @ .13 w/ copper 420,000
AXP @ .13 w/ copper 460,000
K8 @ .13 copper w/ SOI 600,000+

thats a lot more heat per Sq mm before thats why it wont scale or OC to much thermal load on a small die space that cant disapate the heat.

Your right ..... im impressed you know your stuff dude, some of that I didnt know

SSXeon

 

[mechBgon]

I agree, AMD seems to be pushing the limits on how much heat can be squeezed through such a small area.

 

[MajinVegeta]

SSXeon

Thanks I have only been around PC's sence 1996 my mother bougth a gateway i was 13 at the time so I used it to play games and learn more about computers from sites like this and over time i have pick up some stuff here and there,
but the is a lot I dont know I wish I knew it all like the other teens out there LOL.

 

[Viper5]

Originally posted by: christoph83

It'll be reassuring to see AMD take the performance crown back.

Thats not a garuntee. A 2.53ghz P4 paired with PC1066 beats even a overclocked athlon XP 2400+ by a good margin. We'll see though....prices arent too bad.

Oh yeah? And where have u seen a overclocked 2400+ processor?

 

[h2sammo]

Duke Chestnut, Chief AMD thread troll...

jest to make it clear peeps.

the T-Bred review on Gigabyte has been edited and it did not make 2.7Ghz, but 2.1 Ghz:

There wasn?t much information here. Gigabyte had a relatively short presentation but there were 2 points they touched on. The first is that they?ve been expanding into networking in the last 2 years; they had several routers, switches and hubs on display. But the more interesting point is that with their overclocking utility, Easy Tune 4, they were able to get a Thoroughbred 2200+ to 2.6-2.7 GHz, quite the impressive feat. If this is the norm for these chips Pentium 4?s definitely have a run for their money.

EDIT: The first reviews of the Thoroughbred 2200+ are out and none of them are Overclocking to anywhere near 2.7 GHz however they are achieving what would be equivalent to 2500-2600+ so it appears Gigabyte was only able to get it up around 2.1 GHz, still an impressive feat but nothing like an Overclock of almost 50%, more like 20% Apparently this is a case of PR guys mistaking some information from engineers.

Link, ...scroll towards bottom

 

[SteelCityFan]

Originally posted by: Viper5

Originally posted by: christoph83

It'll be reassuring to see AMD take the performance crown back.

Thats not a garuntee. A 2.53ghz P4 paired with PC1066 beats even a overclocked athlon XP 2400+ by a good margin. We'll see though....prices arent too bad.

Oh yeah? And where have u seen a overclocked 2400+ processor?

Best to do some research before you make a fool of yourself...



Tom's Hardware