Official Desktop Penryn Discussion Thread

[Shintai]

Hehe, tread is going rampant fast. Anyway, lets relax abit. Its just mainly certain SSE stuff that gets boosted. The extra cache will add 0-3% I guess. And it would just be a series of reviews quite abit from launch I guess.
But we already seen some benchmarks, unless they were another PR spin.

Anyway, I wait and see for Nehalem.

So...alot of hot air mainly until some time later.

The only thing I hope is that they keep lowering the TDP...

[savantu]

Hehe, tread is going rampant fast. Anyway, lets relax abit. Its just mainly certain SSE stuff that gets boosted. The extra cache will add 0-3% I guess. And it would just be a series of reviews quite abit from launch I guess.
.....

Even without SSE4 , Wolfdale will bring tangible perf improvements , they lowered the latency for a lot of instructions and this will help.

[BitpowerPM]

Only the radix-16 divider and SSE4 instructions (including super shuffle engine) coming with Penryn will boost some multimedia/math applications by 40 - 100%.

So - if necessary - Intel could offer 4 GHz parts of Wolfdale/Yorkfield and thats a large gap compared to 2,5-2,7 GHz K10...

cu, BitpowerPM

[GoThr3k]

Only the radix-16 divider and SSE4 instructions (including super shuffle engine) coming with Penryn will boost some multimedia/math applications by 40 - 100%.

So - if necessary - Intel could offer 4 GHz parts of Wolfdale/Yorkfield and thats a large gap compared to 2,5-2,7 GHz K10...

cu, BitpowerPM

4GHz in a 65w TDP
keep on dreaming

AMD B0 stepping is doing very well,so expect a 2.9 ghz K10 QC

[nibble]

Penyrn is a Die shrink and minor enhancements, not exactly a revolution.

I don't see why thats necessarily a bad thing, Intels current revolution then evolution strategy seems very good.

[onewingedangel]

4GHz in a 65w TDP
keep on dreaming

AMD B0 stepping is doing very well,so expect a 2.9 ghz K10 QC

Higher end conroes, and kentsfields already have higher tdp's than the mainstream parts.

K10 parts too will consume more than 65w at the high end.

At least compare like for like.

[death metal]

Penyrn is a Die shrink and minor enhancements, not exactly a revolution.

Heh, welcome to Tick/Tock approach...

[GoThr3k]

Higher end conroes, and kentsfields already have higher tdp's than the mainstream parts.

K10 parts too will consume more than 65w at the high end.

At least compare like for like.

i was talking dualcore
intel has a 65W TDP (in theory ) for dualcores
i dont see a 4GHz DC possible at that TDP...

i dont think Kuma will consume more than 65W,possible speeds of ~ 3ghz should be possible in a 65W TDP for amd
Wolfdale may squeeze around 3.3Ghz or a bit more in a 65W (intel) TDP
so we will end up with a 3GHz Kuma vs a 3.3/3.5ghz Wolfdale
i don't think there will be much of performance difference

[nn_step]

I don't see why thats necessarily a bad thing, Intels current revolution then evolution strategy seems very good.

then why are people thinking it is going to be a revolution when even Intel says it is just going to be an evolutionary step?

[savantu]

I meant MARKET volume. The QC market is much lower volume than the DC market, and has higher ASPs, so you use your DC dies to make MCM-based QCs first, then sell DCs when you've ramped more volume.

I was not bashing MCM yields vs "native" QC yields. Anyone who isn't a moron realizes that the MCM approach has a yield advantage.

To have an ideea how "much lower volume" the QC market is vs. DC , let just say that in Q1 2007 Intel shipped more QC DP Xeons than the entire DP Opteron production ( SC+DC ) .

75% of Xeon will be QC by Q4 2007.Intel is talking volume already.

[rozzyroz]

then why are people thinking it is going to be a revolution when even Intel says it is just going to be an evolutionary step?

because of the 5-10% increase in clock to clock performance vs current conroe chips, and the possibility to clock to around 4.5ghz on air. that will probably be the biggest evolutionary step i have ever seen. i bet we will see 4ghz oc's on quads be a common thing with penryn.

so i would have to say that it should offer a pretty hefty performance boost for an evolutionary step, and thats what is causing the excitement.

no big thing really, the better intel performs, the better amd has to perform. we all win.

[GoThr3k]

stupid question, with 45nm how worse will the cold bug be?

[Shintai]

intel has a 65W TDP (in theory ) for dualcores

So true, they are so much below their spec. So its only in theory they will ever reach 65W.

Note its measured before the VRM, so its actually lower than on the graph.

[GoThr3k]

So true, they are so much below their spec. So its only in theory they will ever reach 65W.

Note its measured before the VRM, so its actually lower than on the graph.

doesnt prove anything...
i dont think that are brisbanes either...
and which program was run? superpi 32M? thats not a 'burn' condition if you ask me

edit// S&M burn,still that doesnt change the fact these arent brisbanes...

and what i meant is that there is a difference between intel TDP and AMD TDP

[cky2k6]

there used to be... conroes live up to their tdp quite well, no tiptoing around it...

[iterations]

doesnt prove anything...
i dont think that are brisbanes either...
and which program was run? superpi 32M? thats not a 'burn' condition if you ask me

edit// S&M burn,still that doesnt change the fact these arent brisbanes...

and what i meant is that there is a difference between intel TDP and AMD TDP

What did Brisbanes have to do with this?

You implied that Intel's claimed TDP of 65W were false (insinuating that they sink more power). Then Shintai provided you with real data, measured at the VRM by a respected independant lab (Xbit Labs), showing that even at the VRM the power usage is just hitting 65W at the top end, therefore the thermal sink must be below that (since power at the CPU VRM is the ceiling for the amount of power actually flowing through the CPU).

Trapped by real data you quickly try to change the subject.

Ignore the longer bars for the AMD stuff as that clearly has your panties in a knot.

Even at the 65nm node Intel has been able to stay within a 65W TDP, I see no evidence that they won't be able to do this at the 45nm node, especially given the supposed reductions in source/drain leakage from the new High-K and metal gate materials.

[GoThr3k]

What did Brisbanes have to do with this?

You implied that Intel's claimed TDP of 65W were false (insinuating that they sink more power). Then Shintai provided you with real data, measured at the VRM by a respected independant lab (Xbit Labs), showing that even at the VRM the power usage is just hitting 65W at the top end, therefore the thermal sink must be below that (since power at the CPU VRM is the ceiling for the amount of power actually flowing through the CPU).

Trapped by real data you quickly try to change the subject.

Ignore the longer bars for the AMD stuff as that clearly has your panties in a knot.

Even at the 65nm node Intel has been able to stay within a 65W TDP, I see no evidence that they won't be able to do this at the 45nm node, especially given the supposed reductions in source/drain leakage from the new High-K and metal gate materials.

i don not try to change the subject...
brisbanes just suck less power
at that time the tdp of X2 was 89/109 and 125 for FX i think
the tdp of intel is not the same as the tdp of amd
you cannot compare,that test in xbitlabs is a nice test
we should see more of those tests

[Qkjhfhaiguihfma]

Penyrn is a Die shrink and minor enhancements, not exactly a revolution.

you can call it whatever you like, but it will still be the fastest cpu when it's released.

[cky2k6]

i don not try to change the subject...
brisbanes just suck less power
at that time the tdp of X2 was 89/109 and 125 for FX i think
the tdp of intel is not the same as the tdp of amd
you cannot compare,that test in xbitlabs is a nice test
we should see more of those tests

netburst, not core... core2 tdp is the exact same thing as k8 tdp, maximum thermal envelope for a particular family of processor. netburst was expected average consumption. brisbanes might suck around the same amount of power, but they suck alot more balls than core 2...

[Neon Biker]

On the other hand the IMC also adds a bit to those longer bars for the AMD CPUs.

[XS Janus]

I think Paul Otellini is playing on his dualie yorkfield system as we speek, and you guys call it what you will

[Nedjo]

there used to be... conroes live up to their tdp quite well, no tiptoing around it...

speaking about consumption this is how things look in reality:

http://www.tomshardware.com/2007/05/...pu/page11.html

[GoThr3k]

netburst, not core... core2 tdp is the exact same thing as k8 tdp

wrong...

and amd has the IMC included into the thermal envelope...
what would happen to intel if they do that
85W? 95W?

[xVeinx]

wrong...

and amd has the IMC included into the thermal envelope...
what would happen to intel if they do that
85W? 95W?

Only if intel were to include the TDP of the entire northbridge. Which, if AMD were to do the same, would be rather painful, especially if it were to be an nvidia NB (ATI wouldn't be far behind I'm sure).

[Shintai]

wrong...

and amd has the IMC included into the thermal envelope...
what would happen to intel if they do that
85W? 95W?

A memory controller dont use that much. Far from it to say it mildly.

Also remember theinq saying P965 would be ultra hot and a powerhog because of higher possible TDP? Yet it uses less than half of the 975X with lower TDP.

And you do know the worst powerhog chipsets is actually nVidias AMD chipsets?

Also if the northbridge used 20W, then we wouldn´t have the passive cooling we got today. The main thing on a northbridge is the PCIe etc I/O. Not the MCH.

My bet is an IMC would add 2-5W. And 5 being very high.

One thing is sure, you aint gonna remove 10 or 20W from a cooler like this on this board:
http://www.asus.com/prog_content/mid...spx?model=1595

So stop the FUD.

With an IMC the C2D would still be way under its current TDP.