New Carrizo Benchmarks

[ComradeHX]

Steam machine with Carizzo APU perhaps?

[Opcode]

So in comes a conundrum. According to this source the GPGPU gains of Carrizo is roughly only 11% over the desktop Kaveri.

[Opcode]

Sisoft is now reporting a 6 CU variant of Carrizo (likely Carrizo-L) validating at 428 Mpix/s with a clock frequency of 685 MHz.

 

 

That's still faster than the A10-7850k overclocked (389 Mpix/s).

[Dabombinable]

Sisoft is now reporting a 6 CU variant of Carrizo (likely Carrizo-L) validating at 428 Mpix/s with a clock frequency of 685 MHz.

 

 

That's still faster than the A10-7850k overclocked (389 Mpix/s).

What I want to know is the thermals that the bench markers are getting, because AMD can only make one line of inefficient processors, right? (And from experience-AMD's current APU and Ultrabooks don't mix, because they get quite hot)

[Shahnewaz]

Carrizo should be an awesome chip for laptops.

No clue about Carrizo-L though.

[Opcode]

What I want to know is the thermals that the bench markers are getting, because AMD can only make one line of inefficient processors, right? (And from experience-AMD's current APU and Ultrabooks don't mix, because they get quite hot)

I imagine not hot at all as Carrizo-L has been marketed with a 10-25w TDP according to AMD slides. In perspective the chip should put out no more heat than the Athlon 5350 (25w TDP).

 

According to LegitReviews the Athlon 5350 topped out at 43C in Prime95.

When Prime95 was fired up the CPU temperature topped out at 43C and the fan was now running at 2689 RPM. 

Not bad for such a small heatsink!

[Dabombinable]

I imagine not hot at all as Carrizo-L has been marketed with a 10-25w TDP according to AMD slides. In perspective the chip should put out no more heat than the Athlon 5350 (25w TDP).

Well if its able to score as high as it is with a TDP that low, maybe there is hope.

[patrickjp93]

I imagine not hot at all as Carrizo-L has been marketed with a 10-25w TDP according to AMD slides. In perspective the chip should put out no more heat than the Athlon 5350 (25w TDP).

 

According to LegitReviews the Athlon 5350 topped out at 43C in Prime95.

No no no, AMD clearly forgot the 5 in between the 1 and 0 and the 0 after the 25.

[Dabombinable]

No no no, AMD clearly forgot the 5 in between the 1 and 0 and the 0 after the 25.

IT wouldn't actually surprise me if they did end up doing that at some point, maybe a 'mis-communication' like Nvidia's.

[patrickjp93]

IT wouldn't actually surprise me if they did end up doing that at some point, maybe a 'mis-communication' like Nvidia's.

The goal has always been to ensure AMD laptops double as 3-hour emergency heaters.

[Dabombinable]

The goal has always been to ensure AMD laptops double as 3-hour emergency heaters.

They got that downpat years ago with the Phenom II's, so I don't think that's what they are looking to improve.

[Opcode]

The goal has always been to ensure AMD laptops double as 3-hour emergency heaters.

Carrizo TDP is not much higher than Broadwell-U on 14nm and that's saying something with Carrizo on 28nm bulk CMOS.

 

Carrizo on 14nm FinFET would be damn impressive.

[Dabombinable]

Carrizo TDP is not much higher than Broadwell-U on 14nm and that's saying something with Carrizo on 28nm bulk CMOS.

 

Carrizo on 14nm FinFET would be damn impressive.

What's the overall performance though......

[boboman342]

I know nothing is for certain, but given these leaks, do you think these cpu's will come to laptops within 6 months? 

 

I am currently favoring this 15.6in laptop with some pretty amazing spec/build quality. It's got intel hd 4400 graphics, which are good enough for the 1366x768 screen when playing dota2, tf2, csgo, etc, but this really seems like an apu you can really game on! 

 

Give me an aggressively priced/well built ultrabook with a touchscreen and Carrizo level graphics, and I will pre-order 10!

[patrickjp93]

Carrizo TDP is not much higher than Broadwell-U on 14nm and that's saying something with Carrizo on 28nm bulk CMOS.

 

Carrizo on 14nm FinFET would be damn impressive.

It's not bulk. It's SOI.

[Opcode]

I know nothing is for certain, but given these leaks, do you think these cpu's will come to laptops within 6 months? 

 

I am currently favoring this 15.6in laptop with some pretty amazing spec/build quality. It's got intel hd 4400 graphics, which are good enough for the 1366x768 screen when playing dota2, tf2, csgo, etc, but this really seems like an apu you can really game on! 

 

Give me an aggressively priced/well built ultrabook with a touchscreen and Carrizo level graphics, and I will pre-order 10!

One thing that has been debatable since the launch of Tonga was how much L2 does it have. If Volcanic Islands was designed with a larger L2 cache then Carrizo will see a decent performance jump in gaming compared to Sea Islands used in Kaveri. A larger L2 cache makes better use of memory bandwidth.

 

It's not bulk. It's SOI.

Do you have a source?

 

All of my sources (even AMD) are pointing to GF28A. Likely for sake of porting the architecture to GF20LPM which is what I'm betting Nolan (Carrizo-L cut down) will use.

[Shahnewaz]

I'm hoping this time OEMs give Carrizo (being a proper mobile chip) a proper treatment and put them onto premium designs with top notch specs and great battery life.

Unless, Intel.......... forces OEMs' hands.

[patrickjp93]

One thing that has been debatable since the launch of Tonga was how much L2 does it have. If Volcanic Islands was designed with a larger L2 cache then Carrizo will see a decent performance jump in gaming compared to Sea Islands used in Kaveri. A larger L2 cache makes better use of memory bandwidth.

Do you have a source?

All of my sources (even AMD) are pointing to GF28A. Likely for sake of porting the architecture to GF20LPM which is what I'm betting Nolan (Carrizo-L cut down) will use.

The last bulk silicon was Vishera on 32nm Bulk. Since then the industry wised up to using SOI, a compromise without incurring the full costs of FDSOI or research time into FinFET designs or the insanity of IBM's combination of the two. While still being planar, SOI does clear up some of the leakage problems seen with 32nm bulk silicon, but that breaks down at 20nm , especially for high performance chips, something Intel and IBM saw but TSMC was arrogant enough to think it could get around which is why 20nm is being reused for 16nm FinFET.

But yes, the TSMC 28nm process is not bulk. It's SOI.

[patrickjp93]

I'm hoping this time OEMs give Carrizo (being a proper mobile chip) a proper treatment and put them onto premium designs with top notch specs and great battery life.

Unless, Intel.......... forces OEMs' hands.

Oh please more Intel blaming? Why doesn't AMD just market its own name-brand tablet which it has built by Dell/Lenovo/Asus and prove it actually has a winning design instead of being a crybaby over the fact Mullins got trampled by Baytrail for reasons far more important than iGPU performance?

Maybe if AMD would act as a mature player in this game it might actually come out with a few wins?

[Shotgun]

Looks nice.

[Opcode]

The last bulk silicon was Vishera on 32nm Bulk. Since then the industry wised up to using SOI, a compromise without incurring the full costs of FDSOI or research time into FinFET designs or the insanity of IBM's combination of the two. While still being planar, SOI does clear up some of the leakage problems seen with 32nm bulk silicon, but that breaks down at 20nm , especially for high performance chips, something Intel and IBM saw but TSMC was arrogant enough to think it could get around which is why 20nm is being reused for 16nm FinFET.

But yes, the TSMC 28nm process is not bulk. It's SOI.

I asked for a source not an irrelevant lecture. AMD is using GloFo for manufacturing Carrizo not TSMC just like they used GloFo 28nm bulk SHP for manufacturing Kaveri.

[Shahnewaz]

Oh please more Intel blaming? Why doesn't AMD just market its own name-brand tablet which it has built by Dell/Lenovo/Asus and prove it actually has a winning design instead of being a crybaby over the fact Mullins got trampled by Baytrail for reasons far more important than iGPU performance?

Maybe if AMD would act as a mature player in this game it might actually come out with a few wins?

Are you mad I blamed Intel for the notorious (possibly illegal) practices they do?

Making an own-brand tablet is just stupid. It's shooting yourself on the foot. That will literally ruin relationships with OEMs just like Microsoft did with their Surface.

As for the matter of Mullins vs. BayTrail, we've had this conversation before. I showed you Mullins was clearly better (just look up any review) and Intel Contra Revenue speaks volumes why Mullins weren't used in all the new devices that sports Bay-Trail cpus.

 

Yeah. If Intel can't compete with better chips, they compete with money power. It's upto you to decide whether that's "mature" or not.

[patrickjp93]

I asked for a source not an irrelevant lecture. AMD is using GloFo for manufacturing Carrizo not TSMC just like they used GloFo 28nm bulk SHP for manufacturing Kaveri.

It's not an irrelevant lecture and it's common knowledge. The GloFo SHP process is SOI.

[patrickjp93]

Are you mad I blamed Intel for the notorious (possibly illegal) practices they do?

Making an own-brand tablet is just stupid. It's shooting yourself on the foot. That will literally ruin relationships with OEMs just like Microsoft did with their Surface.

As for the matter of Mullins vs. BayTrail, we've had this conversation before. I showed you Mullins was clearly better (just look up any review) and Intel Contra Revenue speaks volumes why Mullins weren't used in all the new devices that sports Bay-Trail cpus.

Yeah. If Intel can't compete with better chips, they compete with money power. It's upto you to decide whether that's "mature" or not.

If AMD has no relationship in the first place, it's irrelevant, and as it's shown they have success with their own branded tablet others will join on. Also, Mullins was only better for heavily threaded benchmarks and graphics, two things tablets aren't intense on anyway. Next inane point?

Intel has done nothing wrong using contra revenue to break ARM's stranglehold on the tablet market, and that strategy is entering its sunset stage because Intel stopped givingn Atom chips away for free. They're sold at production cost (not covering fab R&D obviously) in the tablet sector, though in phones Intel will have to give chips away and wait on developers' hands and feet until enough of them convert for real competition.

[Opcode]

It's not an irrelevant lecture and it's common knowledge. The GloFo SHP process is SOI.

AMD left SOI behind with 32nm looking for density. The GF 28nm SHP process used on Kaveri is also bulk.