Penryn 40% faster on games, video...

[Shintai]

[p8ntslinger676]

Good news is they have 3.33GHz 1333FSB chips up and running already

Let's face it...most of us would be happy with identical clock-for-clock performance with only higher clocks...but we're getting both, even if it is only a small gain in some applications.

3.6ghz @1600mhz fsb

[kiwi]

First Xtreme edition penryn will be 3.63Ghz 11x333 according to unofficial source Later they will move to 1600fsb

[Vapor]

I could put them in a vise and give it a 20nm squeeze!

Or a girdle!

Hmmm, that DiVX is really, really impressive....hopefully h.264 encoders can utilize SSE4, DiVX is ugly Rest of the scores aren't too shabby either....would like a bit more clock-for-clock improvement in my main application of Photoshop, but I guess that would be asking too much

[savantu]

If you take out the 13% speed difference and the extra cache/FSB you end up with little IPC improvement if any.I guess you need SSE4 enabled SW to show Penryn's potential.

Anyway , Intel lifted the bar even higher for AMD.If the K10 top bin will be 40% faster on average* than QX6700 it will barely match the 45nm Intel generation.

[selectodude]

Well it seems that the same processor overloaded with L2 cache is faster then the old cores.

Surprise! L2 cache can only help so much. AMD's way of doing things is much more elegant.

[theteamaqua]

it will be close i think ... this is not like K8 vs presler where difference is still huge .

[savantu]

Well it seems that the same processor overloaded with L2 cache is faster then the old cores.

Surprise! L2 cache can only help so much. AMD's way of doing things is much more elegant.

That's why they are planning chips with up to 8MB of cache next year ?

[nn_step]

That's why they are planning chips with up to 8MB of cache next year ?

why don't we just take the leap and have 1GB of L2 and call it a day

[Carfax]

The DivX scores are impressive to say the least.

I'm sure that includes SSE4 optimizations.

However, the new Super Shuffle enhancement in the Penryn core should provide a boost on SSE optimized apps, even if it's not SSE4 optimized.

AMD won't get SSE4 until next year when they move to 45nm..

[madcho]

3.33-2.9 and higher bus.

15% for higher frequency.

about 5-10% for bus.

And you need use SSE4 for get the fake 40% ...
intel is really good for fake bench

And peryn don't have one L2 for the four cores ... that's always a MCM way.

caraf, AMD run SSE4A in K10 65nm too

[zerotol]

ill just wait for real benches instead of slides

[Carfax]

3.33-2.9 and higher bus.

15% for higher frequency.

about 5-10% for bus.

And you need use SSE4 for get the fake 40% ...
intel is really good for fake bench

And peryn don't have one L2 for the four cores ... that's always a MCM way.

caraf, AMD run SSE4A in K10 65nm too

SSE4A is NOT the same thing as SSE4.

Also, the reason for the inflated DivX scores is definitely due to the Alpha SSE4 optimized DivX..

The other programs have not been SSE4 optimized, hence the performance advantage is much lower.

[Shintai]

3.33-2.9 and higher bus.

15% for higher frequency.

about 5-10% for bus.

And you need use SSE4 for get the fake 40% ...
intel is really good for fake bench

And peryn don't have one L2 for the four cores ... that's always a MCM way.

caraf, AMD run SSE4A in K10 65nm too

Just like AMD needs specFP_rate. Its just as fake

[gOJDO]

3.33-2.9 and higher bus.

15% for higher frequency.

13.63%

about 5-10% for bus.

FSB frequency does not translates into CPU performance. It can help only a little(1%-3%) in certain cases, when the BUS is getting saturated and needs more than 8.5GB/s bandwidth, and such cases are very rare thanks to the aggressive prefetchers and large L2.

And you need use SSE4 for get the fake 40% ...
intel is really good for fake bench

WTF are you FUDing about?
SSE4 is the most significant update to the SSE instruction set and has more improvements over SSSE3, than SSE2 had over SSE1. Also Penryn's SSE engine has architectural improvements and Penryn will perform faster than same clocked Conroe in SSE3/SSE2(non SSE4) applications.
Do you have an idea what is SSE, how it works and what kind of software has benefit of it?

And peryn don't have one L2 for the four cores ... that's always a MCM way.

So what?

caraf, AMD run SSE4A in K10 65nm too

AMD's SSE4A has only 6 new instructions.
Intel's SSSE3 and SSE4(has the SSE4A instructions) have 82 new instructions.

[XS Janus]

This sounds great!
I guess Intel slides will be the guideline for AMD marketing on how to adjust their own slides not to look bad

[Ubermann]

Slides is ok for intel users but if amd does it then its far from ok.
Its a funny world =)

[alpha0ne]

Looks all good for the consumer, I dont give a toss if its AMD or Intel as long as the performance is ramped

[R101]

I want to see an AMD slide with Cinebench!

Or with HL2, for that matter..

[.Tret]

I want to see an AMD slide with Cinebench!

Or with HL2, for that matter..

I'll add to that; I want to see an AMD slide with clockspeed listed...

Slides suck, but this one slide is more useful than all of amd's slides so far.

[Metroid]

That is a great old news coming from the surface to the depths of Intel wafer's. I see a wishful thinking over and over about Penryn and I really think will be a great step for Intel as had been CD2. The truth is that Penryn will be faster and an improvement of 20% over the top current quad core QX6800 shows us how fast that will be. I wonder what AMD is going to do to fight it back. Penryn will be coming in the Q1 08 or earlier Q4 07.

[Shintai]

Slides is ok for intel users but if amd does it then its far from ok.
Its a funny world =)

There is a difference in real world applications. And then a specFP_rate benchmark. Also AMD compared Clovertown with a dualcore K8 and K10 quad. They left out dual dualcores on purpose.

K8 already beated Clovertown in that bench. Atleast here we have something REAL and something thats comparable to the previous.

Show me the AMD slide with a 2S dualcore K8 system vs a single socket K10. And you see K10 will lose or just equal the same specFP_rate bench.

[kniwor]

Looks all good for the consumer, I dont give a toss if its AMD or Intel as long as the performance is ramped

agree

conclusions on speculations are not the best indicators of performance, I cannot help but be silent.

[WillyWonka]

But we believe AMDs numbers out of the box. You are funny nn_step.

Word.

How bizarre that Penryn will ship AFTER Barcelona and yet we get Penryn benches now but no Barcelona benches. AMD seems afraid to expose their upcoming offering to scrutiny

PS: Even if K10 and Penryn are more or less the same speed, Intel will have the massive advantage of 45 nm vs AMDs 65 nm. Intel will be able to keep prices low forcing AMD to do the same. If AMD doesnt have a significant speed advantage they must compete on price ... which is an unhealthy position to be in vis a vis Intel.

[WillyWonka]

since when amd slides suck and intel slides rule ?

Since Intel gave a proper description of the chip being benched including the clockspeed.

AMD didnt bother giving a clockspeed. Their performance figures weren't a benchmark.