Vishera spotted running at 5GHz

[pelov]

If im not mistaken same 28nm GloFo for Excavator.

So Broadwell/Excavator on 22nm Intel 28nm GloFo bulk, respectively?

Where are they going after that? 20nm? 14nm FinFET?

 

[AtenRa]

Question: Is the 7% improvement measured from a 3.6Ghz Bulldozer to a 4 Ghz PildeDriver or did they run the CPUs at the same clockspeed?

Same Clock Speed(Clock to Clock) = 7%

+

7% from higher frequency 4GHz vs 3.6GHz

The combination gives you 15% more performance at the same TDP(125W)

We just dont know yet if that 7% + 7% is the highest, average or lowest it can get.

 

[AtenRa]

So Broadwell/Excavator on 22nm Intel 28nm GloFo bulk, respectively?

Where are they going after that? 20nm? 14nm FinFET?

Broadwell is on Intels 14nm FinFet

After 28nm Glofo (2013) will go to 20nm (2014/15?) and then FinFet at 14nm(20..?)

 

[pelov]

Broadwell is on Intels 14nm FinFet

Right. Brain fart

I'm wondering how quickly GloFo can get 14nm FinFET to viability. They're bragging about their 14nm accomplishments with expected tape-outs next year. Granted, that's "14nm-XM" It sounds to me they're catering more towards ARM SoCs than AMD's x86 ambitions.

 

[Idontcare]

If im not mistaken same 28nm GloFo for Excavator.

Excavator will be 28nm? Bummer then, I was completely thinking Excavator would be on 20nm. A 28nm Excavator competing against a 14nm Broadwell is not the kind of battle AMD wants to be having.

 

[AtenRa]

Excavator will be 28nm? Bummer then, I was completely thinking Excavator would be on 20nm. A 28nm Excavator competing against a 14nm Broadwell is not the kind of battle AMD wants to be having.

Perhaps they will only use 28nm for a single year(2013) and jump to 20nm in 2014 (Excavator). New York's Fab 8 was supposed to start producing 22/20nm in 2014.

 

[pelov]

Excavator will be 28nm? Bummer then, I was completely thinking Excavator would be on 20nm. A 28nm Excavator competing against a 14nm Broadwell is not the kind of battle AMD wants to be having.

That was my thinking as well, which is why I was curious.

Perhaps they will only use 28nm for a single year(2013) and jump to 20nm in 2014 (Excavator). New York's Fab 8 was supposed to start producing 22/20nm in 2014.

I'm hoping that will be how it pans out.

I think what's interesting here is that under Rory Read, AMD actually makes 2 processors a year, not including the 17W-and-under variants. The only delay they've had was due to fab issues, cancelling the 28nm Brazos. The other delay was due to oversupply of Llano, forcing Trinity desktop back a couple of months. In the same time frame, Intel was also forced to push back their Ivy Bridge release due to supposed yield issues (I remember they pushed the dates back but I don't think a reason was ever given). It's kind of shocking to see AMD being on time, for the most part.

I'm kind of hoping they've had an SoC brewing. Battling Intel with fewer chips per wafer, lower margins and lower volume is a path to destruction. Maybe streamlining their product line into HPC/mobile with APUs (if they can get openCL penetration at the HPC level. Right now it's mostly a CUDA ordeal), and an SoC targeted at the ultramobile range, they would cut some costs. I'm sure supporting Brazos, APUs, and a separate CPU design for desktops/server isn't cheap.

 

[guskline]

Same Clock Speed(Clock to Clock) = 7%

+

7% from higher frequency 4GHz vs 3.6GHz

The combination gives you 15% more performance at the same TDP(125W)

We just dont know yet if that 7% + 7% is the highest, average or lowest it can get.

It won't be that long to find out. I hope your right about the improvement.

 

[inf64]

It won't be that long to find out. I hope your right about the improvement.

I guess there might be cases where improvement will be miniscule. Think single/poorly threaded fp heavy benchmark. So both will run at 4.2Ghz and fp perf. improvement is less than integer with Piledriver(and you end up with same performance). And then there will be cases of 10-15% IPC increase AND 5% sustained clock jump( 3.9Ghz full core Turbo on 8150 and 4-4.1Ghz on 8350). It all averages at 15% total perf. increase which is not bad. We don't know the power draw yet,but if it's lower than 8150's then it will be an upgrade in every way.

 

[AtenRa]

http://hardforum.com/showthread.php?t=1722207

@4.5GHz

Catching the Core i7 2600K at Cinebench 11.5 MT is not bad at all. In other MT tests it should be close to 3770K and faster than 2600K.

 

[Mallibu]

http://hardforum.com/showthread.php?t=1722207

@4.5GHz

Catching the Core i7 2600K at Cinebench 11.5 MT is not bad at all. In other MT tests it should be close to 3770K and faster than 2600K.

Sorry but I still think it's crap.
i7 3770k which is the current competitor scores 7.6 @ 3.5 GHz.
If it's slower than him in a test that uses 8 threads, then it's gonna be pretty fail in the rest 95% applications that today exist and make use of 1-4 threads.

As we move closer to launch date my magical sphere predicts waves of cherry picking incoming

 

[skipsneeky2]

Nice CineBench score for the FX8350 (4.8GHz or 5GHz), faster than Core i7 2600K at 4.5GHz.

FX8350

FX8150@4.7GHz

Core i7 2600K @ 4.5GHz

FX8150 and Core i7 2600K source :
http://forums.anandtech.com/showthread.php?t=2241712&highlight=

I kept thinking i was crazy but the 4.7ghz 8350 is getting 7.72,but the stock clocked i7 2600k is getting 8.78....tell me if i am missing something,higher is better correct?

LOL,oh thats for the freaking 8150,i just looked more closely at the model of the chip in cinebench.