Penryn only 0-10% over Conroe?

[Idontcare]

Originally posted by: Acanthus

Originally posted by: bryanW1995

Originally posted by: Acanthus

Originally posted by: HopJokey
One thing people are overlooking is that in the preview, Anand was able to push the A0 chip to 3.22Ghz on 1.15v!

Dont forget, however, that they wont tolerate "conroe like" voltages either.

The ceiling for safe voltage will be around 1.3v

why would the safe voltage ceiling be so low?

Because the smaller you make the gates, the easier it is to slow roast them with voltage

They also require less voltage to run, so its a non issue as long as you know not to set it too high.

And the "safe" voltage bias you can place across adjacent copper lines scales linearly with node shrinkage so as to maintain same electric-field on the dielectric from node to node and maintain 10-year lifetime expecation for time-dependent dielectric breakdown (tddb) strength.

Expect minimum metal half-pitch at 65nm node to be 90nm and the minimum half-pitch at 45nm node to be 70nm and you can quickly determine that you must scale the voltage by approximately 70nm/90nm = 0.778x or else you can expect a reduced tddb (reduced chip lifetime from e-field induced failure mechanisms).

 

[SickBeast]

Originally posted by: HopJokey

Originally posted by: TuxDave

Originally posted by: SickBeast

Originally posted by: Acanthus

Originally posted by: SickBeast
Barcelona should be cooler and more efficient than Penryn, but who knows until it's out.

Err... 45nm vs 65nm?

I would take an A64's cool-n-quiet feature over a C2D on a smaller process, just as an example. My Opteron 165 is dead silent while browsing and at idle (90% of the time) because of that feature.

AMD has awesome power efficiency; they always have. The manufacturing technology only helps to a degree. Barcelona *will* be more efficient than anything intel will have.

Since when did processors make noise? And if you're talking about fans, then what's stopping an Intel user from buying the same HSF?

Hey, electrons make noise too!

?

I'm referring to the fact that the processor lowers its clockspeed/voltage enough to the point that it doesn't require the fan.

 

[Regs]

Originally posted by: coldpower27

Originally posted by: Regs
Wasn't "barc" suppose to be out this month?

Barcelona was supposed to be Mid 2007, but has been pushed out to Q3 2007.

Sunnyvale, Calif. -- June 29, 2007 --Continuing to lead the shift of mainstream enterprise computing to energy-efficient processors, AMD (NYSE:AMD) announced today that Quad-Core AMD Opteron? processors, code-named ?Barcelona,? are planned for shipment in both standard and low power versions at launch later this summer.


This coming from a news statement right from amd.com .


Labor day is only a week away. So I'm guessing by the middle of Sept.

 

[coldpower27]

Originally posted by: Regs

Originally posted by: coldpower27

Originally posted by: Regs
Wasn't "barc" suppose to be out this month?

Barcelona was supposed to be Mid 2007, but has been pushed out to Q3 2007.

Sunnyvale, Calif. -- June 29, 2007 --Continuing to lead the shift of mainstream enterprise computing to energy-efficient processors, AMD (NYSE:AMD) announced today that Quad-Core AMD Opteron? processors, code-named ?Barcelona,? are planned for shipment in both standard and low power versions at launch later this summer.


This coming from a news statement right from amd.com .


Labor day is only a week away. So I'm guessing by the middle of Sept.

So later this summer would be anytime before September 23rd 2007, what you suggested would be a reasonable estimate.

 

[Maluno]

Originally posted by: Rogue 2
Maybe Intel is getting lazy again now that they are on top...
Or maybe people just expect too much.
Either way, the bar's set.. time for AMD to answer.

Hmm... In my experience, the one on top usually does most of the work...

 

[bryanW1995]

Originally posted by: Idontcare

Originally posted by: Acanthus

Originally posted by: bryanW1995

Originally posted by: Acanthus

Originally posted by: HopJokey
One thing people are overlooking is that in the preview, Anand was able to push the A0 chip to 3.22Ghz on 1.15v!

Dont forget, however, that they wont tolerate "conroe like" voltages either.

The ceiling for safe voltage will be around 1.3v

why would the safe voltage ceiling be so low?

Because the smaller you make the gates, the easier it is to slow roast them with voltage

They also require less voltage to run, so its a non issue as long as you know not to set it too high.

And the "safe" voltage bias you can place across adjacent copper lines scales linearly with node shrinkage so as to maintain same electric-field on the dielectric from node to node and maintain 10-year lifetime expecation for time-dependent dielectric breakdown (tddb) strength.

Expect minimum metal half-pitch at 65nm node to be 90nm and the minimum half-pitch at 45nm node to be 70nm and you can quickly determine that you must scale the voltage by approximately 70nm/90nm = 0.778x or else you can expect a reduced tddb (reduced chip lifetime from e-field induced failure mechanisms).

gotcha. so 1.15 v is more like 1.40 on a conroe... wow, those thing should really run much cooler just do to lowered voltages if nothing else. Of course, the higher fsb's will make up for total system heat I suppose...

what about if we only care about the system lasting 3-4 years? Does that mean we can run 1.75v on air?

 

[bryanW1995]

Originally posted by: coldpower27

Originally posted by: Regs

Originally posted by: coldpower27

Originally posted by: Regs
Wasn't "barc" suppose to be out this month?

Barcelona was supposed to be Mid 2007, but has been pushed out to Q3 2007.

Sunnyvale, Calif. -- June 29, 2007 --Continuing to lead the shift of mainstream enterprise computing to energy-efficient processors, AMD (NYSE:AMD) announced today that Quad-Core AMD Opteron? processors, code-named ?Barcelona,? are planned for shipment in both standard and low power versions at launch later this summer.


This coming from a news statement right from amd.com .


Labor day is only a week away. So I'm guessing by the middle of Sept.

So later this summer would be anytime before September 23rd 2007, what you suggested would be a reasonable estimate.

I believe that their most recent target date is 9/11 with 2.0 ghz max clock speed.

 

[Griswold]

Originally posted by: Acanthus

2. the benchmarks did not include SSE4, which is an enormous performance gain in certain applications.

Right, SSE4 - exactly how many applications supporting it do you think we'll at launch / 6 months later / 12 months later? If we consider how slowly SSE3 received widespread adoption, you cant use this as an argument. Some divx6 codewhackjob probably done by intel for divx for benchmarking purpose doesnt really cut it at this point in time.

 

[Griswold]

Originally posted by: SickBeast
According to Fudzilla, Barcelona will have a golden opportunity seeing as Penryn is slower than expected.

I really do expect Barcelona to be much better than people are saying.

Well, it wont be hard for K10 to top the opinion of these people if it was the same people who spread their uneducated guesstimates about penryn, expecting monumental performance increases and what not. Penryn is evolution, not revolution.

 

[Chronoshock]

Some preliminary Yorkfield benches are out:
http://www.dailytech.com/article.aspx?newsid=8585

Seems like similarly small performance gains

 

[Nemesis 1]

Don't you guys think that to determin the performance gains of penryn over conroe we should wait till . We have other than A0 steppings.

I see a lot of guessta mating here. The advantages everyone are pointing at are @ 2.33 ghz . On M/Bs that are not optimized for penryn .

I think we should wait Till we see the results of A penryn running at 3.0 ghz on a fully optimized X38 chipset with the memory Intel intended to be used with penryn and the X38 chipset which is in fact DDR3. Cost be dammed . Thats not the issue . Performance is . I think when its all said and done On the top of the line hardware we well see a 15-20% increase in performance.

That my belief.

Now on SSE4 . People keep saying programmers won't optimize for it . Than they point at SSE3 and use that as an example . How fast was SSE2 implamented?

If SSE4 does for other apps . what ii did for div6+ . Programmers will be all over it and it will be implamented very rapidly Intel has the tools already available to do the optimizations and as I understand it its easy to do. So this 5-10% increase in performance realy is nonsense.

 

[nyker96]

Originally posted by: Frackal
http://www.anandtech.com/cpuch...howdoc.aspx?i=3069&p=3



Bummer! I was looking for 10%-20%+

that's plenty fast!