3 Ghz Athlon 64 Q1 2006

[Jeff7181]

Originally posted by: Duvie

Originally posted by: Jeff7181

Originally posted by: Sentential

Originally posted by: CheesePoofs

Originally posted by: Sentential
While this appears to look all well and good there is something you arent seeing. The reason why Venice and others are scaling so nicely, and why there is a 3ghz A64 on the horizon, is because they are lengthening the pipeline much like Intel did with prescott.

Granted not drastically but be aware, you are not getting what you think you really are. They will not be as fast clock-per-clock as the orginal design would have been. If anything they will be the same speed with SSE3 to make up for the shortcomings.

If you want proof go read the A64 article on the front page. AMD rep says such clear as day: "AMD *will* be increasing the pipeline stage to increase CPU speed"

How would you know?

Yeah, amd said they will, but that doesn't mean they have (seeing as rev E is already out in the form of the opteron 252s.)

Ive long suspected that E0 use a relaxed pipe. All signs from what I have seen point to it (this was before I went to intel, all fanboism aside). Here is what we know:

#1) Almost no difference in speed on SuperPI and in benches between them.. like 0%
#2) Higher heat. 67W --> 84W (whincester vs venice)

No one has come up with an explination to this... I cant either other than they legthened the pipeline for 3ghz which was recentally announced

That's pure speculation for now. Yes they said they will, but I think they were referring to the next major model change... the successor to the Athlon-64.

The "higher heat" could be the specification for the Venice core, not a specific model. If Venice scales to 3 GHz, of course it'll put out more heat than a 2.4 GHz Winchester. That doesn't mean that the Vencie core uses more power than a Winchester if clock speeds and voltage are the same.

Exactly....There is no logic in his thought!!!

There's also no reason to believe AMD would want to lengthen the pipeline at this point. FX-55's scaling to 3 GHz is proof that the pipeline is good up to 3 GHz and probably beyond before it HAS to be lengthened to go any further.

Also, the tests being done in that other thread show that the Venice performs slightly better than a Winchester at the same speed... the opposite should be true if the pipeline has been lengthened.

 

[Duvie]

Again exactly....Plus they should try to run the prescott SSE3 version of superpi to see if it has an increase there...superpi at 1mb takes quite a bit more speed to eclipse just one more second. Also in those reviews the guy is running a modded ersion that shows .xxx seconds and that one is confirmed to not be the patched version with SSE2 options let alone other things. Iam convinced it wil not show any added strengths of the venice core...

 

[n7]

It sounds like 3 GHz A64 (non-Opterons) are still a ways off sadly.

I will be upgrading my 754 system as soon as there is cheaper A64 that is either clocked at 3GHz, or there's lower ones that are nearly certain to hit 3 GHz.

 

[Acanthus]

Isnt the venice core strained silicon? That increases leakage and thermals.

 

[Jeff7181]

Originally posted by: Acanthus
Isnt the venice core strained silicon? That increases leakage and thermals.

LOL... what gives you that idea?

*EDIT* Oh... because of the Prescott? :laugh:

 

[Sentential]

Originally posted by: Acanthus
Isnt the venice core strained silicon? That increases leakage and thermals.

Yes its a form of Strained silicon called DSL (dual stress liners).

__________

@ Duvie

I dont see how my logic is flawed. AMD relaxed the pipeline during the move to A64. Why wouldnt they do that again? As fot the increased thermal data that is usually a sign of the CPU queing more cache as it does in Prescott's case. Longer pipes use more cache which drastically increases the wattage count

Also the information I had up to the point of making that post is incorrect. Now that AMD updated their spec sheets (taken from Xbit) the wattage is the same @ 67W. 84W is reserved only for the 3800+.

On a curious note the E0s appear (or atleast Xbit suggests) that the new E0s support some form of EIST. . Normally AMD cpus have one voltage set for idle and load but Xbit shows 2 (1.35 and 1.4). Not sure what to make of that at this point, but im asking around.

 

[Duvie]

Originally posted by: Sentential

Originally posted by: Acanthus
Isnt the venice core strained silicon? That increases leakage and thermals.

Yes its a form of Strained silicon called DSL (dual stress liners).

__________

@ Duvie

I dont see how my logic is flawed. AMD relaxed the pipeline during the move to A64. Why wouldnt they do that again? As fot the increased thermal data that is usually a sign of the CPU queing more cache as it does in Prescott's case. Longer pipes use more cache which drastically increases the wattage count

Also the information I had up to the point of making that post is incorrect. Now that AMD updated their spec sheets (taken from Xbit) the wattage is the same @ 67W. 84W is reserved only for the 3800+.

On a curious note the E0s appear (or atleast Xbit suggests) that the new E0s support some form of EIST. . Normally AMD cpus have one voltage set for idle and load but Xbit shows 2 (1.35 and 1.4). Not sure what to make of that at this point, but im asking around.

It is flawed to expect they would do this now and not during a more defined core revision. AMD long ago said SS/SOI would be enough to push the limits further. Also longer pipeline should reduce the IPC (ala Intel) and clock for clock the venice should have beem slower. IN non SSE3 apps that doesn't appear to have happened so I dont think there were any other core logic changes....

I am interested in that 1.35v as well....I like many on extreme undervolting mobos like MSI Neo2 know that many of our winchesters now can run 1.34v at stock speeds if not even a bit higher. If I can get to 2.5ghz with stock vcore chances are I may have been able to run near 1.3v at stock 1.8ghz...

Edit: as for the wattage count I believe Jeff gave some logic into what those reported specs were and what they mean. I tend to agree with it...