AnandtechAMD Carrizo ExcavatorReview

[Shivansps]

Carrizo-L is Mullins, which supports cTDP between 12-25W. Each and every Carrizo has TDP of 15W, but they all support cTDP between 12-35W.

It is up to the ODM if they wish to implement cTDP or not. So far Lenovo is the only ODM who has implemented cTDP for Carrizo (Y700-15ACZ).

But this is nonsence, Carrizo-L E1-7010 (Dual, 1.5Ghz) is 10W, no cTDP, Carrizo-L A6-8500P is 15W (1600-3000,256SP), also no cTDP... if Carrizo-L could need up to 25W the smaller Carrizo need 15W, why allow OEM to run 2 module Carrizos with 384 and 512SP igp at something as low as 12W? this is BS to me...

This cTDP thing in both Intel and AMD is getting way out of control, you never know what are you getting.
(And even Intel does not offer Skylake U 15W with lower cTDP than Skylake Y or Skylake U 28W with lower cTDP than U 15W...)

 

[The Stilt]

I think you´re somewhat confused...

 

[Shivansps]

I think you´re somewhat confused...

I dont think so, OEMs can sell a FX-8800P at 12W and we have no way to know it, thats true or not?

And even at 15W is highly limited, thats the TDP that a A6-8500P should be running.

In contrast, look at what intel does,

SKY-Y = 3.5W to 7W cTDP
SKY-U(15W) = 7.5W to 15W cTDP

Translating this last thing to AMD Carrizo

SKY-Y = 3.5W to 7W cTDP
SKY-U(15W) = 3.5W to 15W cTDP


In short why AMD allows its top Carrizo to run at TDP that is already insufficient for Carrizo-L and we have no way to know it?

I dont like this cTDP thing, but i like AMD implementation even less, to a point im even wondering why Carrizo-L exist if its sharing the same TDP floor.

 

[Abwx]

In contrast, look at what intel does,

SKY-Y = 3.5W to 7W cTDP
.

The figure are much higher, at least 12W permanently in the exemple below :

http://www.notebookcheck.net/HP-Spectre-x2-12-a003ng-K3D42EA-Notebook-Review.158679.0.html

And roughly 18W in the comparison they are quoting :

We measure 20 watts here, while the Core M-5Y71 in the Helix 2 and Z20t consistently drain 30 watts.

Same with the Toshiba 12"5, given the screens sizes thoses items CPUs are well over 15W :

http://www.notebookcheck.net/Toshiba-Portege-Z20t-B-Convertible-Review.140840.0.html

 

[The Stilt]

I dont think so, OEMs can sell a FX-8800P at 12W and we have no way to know it, thats true or not?

And even at 15W is highly limited, thats the TDP that a A6-8500P should be running.

In contrast, look at what intel does,

SKY-Y = 3.5W to 7W cTDP
SKY-U(15W) = 7.5W to 15W cTDP

Translating this last thing to AMD Carrizo

SKY-Y = 3.5W to 7W cTDP
SKY-U(15W) = 3.5W to 15W cTDP


In short why AMD allows its top Carrizo to run at TDP that is already insufficient for Carrizo-L and we have no way to know it?

I dont like this cTDP thing, but i like AMD implementation even less, to a point im even wondering why Carrizo-L exist if its sharing the same TDP floor.

All current Carrizo models (A6-8500P/B, A8-8600P/B, A10-8700P/B and FX-8800P/A12-8800B) have 15W TDP. Unless the manufacturers specifically alter the default configuration, these chips will SUSTAIN the 15W TDP and be able to boost to 25W for 200 seconds.

Thats the base line.

The ODMs can configure the cTDP to any combination between 12/20W to 35/42W if they like, on any of these SKUs or keep it at stock settings like all the vendors but Lenovo have done. There is zero evidence of any ODM actually lowering the TDP from the default 15/25W value.

 

[coercitiv]

I dont like this cTDP thing, but i like AMD implementation even less, to a point im even wondering why Carrizo-L exist if its sharing the same TDP floor.

Cheaper to make, for the time being.

I did not follow on this subject, but last time I read about Bristol Ridge & Stoney Ridge, available info pointed towards the retirement of cat cores in favor of XV cores (1 module vs. 2-4 cats) The info was scarce to say the least, so we need more sources before painting a picture.

For me the configurable TDP is not a problem, I've seen very few designs using this feature: hybrid units that functioned with cTDP down in tablet mode, normal TDP in laptop mode. Granted it gives more room for the OEM to screw up, but here's the kicker: they would have ****** it up anyway, with even worse consequences, by enforcing some power profile that would bring the device on it's knees (say hello to max 1200Mhz, user patience overclock&#8482.

 

[Shivansps]

All current Carrizo models (A6-8500P/B, A8-8600P/B, A10-8700P/B and FX-8800P/A12-8800B) have 15W TDP. Unless the manufacturers specifically alter the default configuration, these chips will SUSTAIN the 15W TDP and be able to boost to 25W for 200 seconds.

Thats the base line.

The ODMs can configure the cTDP to any combination between 12/20W to 35/42W if they like, on any of these SKUs or keep it at stock settings like all the vendors but Lenovo have done. There is zero evidence of any ODM actually lowering the TDP from the default 15/25W value.

a A6-8500P and a FX-8800P both at 15W TDP kinda confirms what im saying, FX-8800P at 15W is like a SKL-Y...

 

[Flash831]

Hawaii's performance is hardly bad at higher resolutions in DX12, particularly considering its age.

Don't get me wrong, I think Hawaii is a great chip performance wise. Anyone who bought Hawaii (or Tahiti!) at release bought a very futureproof GPU.

However, in terms of efficiency (performance/watt) I think it has passed its time. R9 Nano is already 2x the performance per watt. The 14nm GPU's will be even better.

This is a thread about Carrizo, so let us not get derailed.

 

[mikk]

And roughly 18W in the comparison they are quoting :



Same with the Toshiba 12"5, given the screens sizes thoses items CPUs are well over 15W :

http://www.notebookcheck.net/Toshiba-Portege-Z20t-B-Convertible-Review.140840.0.html

Average is 5W for the whole package. This is a 5W Soc, although it can exceed 5W for a few seconds. But it's clearly SKL-Y with default TDP of 5W.

The other shows 8W average for the whole package, it's possibly configured for 7W TDP.

 

[cbn]

I dont like this cTDP thing, but i like AMD implementation even less, to a point im even wondering why Carrizo-L exist if its sharing the same TDP floor.

Cheaper to make, for the time being.

I did not follow on this subject, but last time I read about Bristol Ridge & Stoney Ridge, available info pointed towards the retirement of cat cores in favor of XV cores (1 module vs. 2-4 cats) The info was scarce to say the least, so we need more sources before painting a picture.

Yep, Carrizo-L is smaller at 102mm2 compared to 245mm2 for Carrizo-L

With that mentioned, I am concerned about the performance of Excavator quad core at 15W compared to cat core quad at 15W.

For example, while digging through the FX-8800P passmark scores I saw there was a mix of laptops and developer boards, etc (and thus CTDPs) for the overall score reporting of 4230.

However, this score of 2569 CPU marks from a FX-8800P ASUS X555DG laptop is actually slightly lower than A8-7410 (Carrizo-L) score of 2583 CPU marks.

With that mentioned, the ASUS X555DG does have only 1 DIMM socket. This may or may not be limiting performance in this CPU benchmark (although A8-7410 is also single channel, of course).

P.S. A8-6410 (another 15W cat quad core chip) has passmark score of 2521.

 

[cbn]

If the passmark score of FX-8800P @ 15W is actually lower than Carrizo-L A8-7410 (see above post) and the laptop makers (in this case ASUS) are building boards with one DIMM slot (thus limiting the iGPU advantage of FX-8800P over A8-7410) it makes me wonder how much OEMs are paying for these FX-8800P chips?

My guess is not much.

 

[Abwx]

Average is 5W for the whole package. This is a 5W Soc, although it can exceed 5W for a few seconds. But it's clearly SKL-Y with default TDP of 5W.

The other shows 8W average for the whole package, it's possibly configured for 7W TDP.

Numbers are numbers :

We measure 20 watts here, while the Core M-5Y71 in the Helix 2 and Z20t consistently drain 30 watts.

Consistently mean permanently, i dont know how you manage to reconcile 30W at the main (with 12" screens items..) permanently with your "7W TDP", indeed the scores couldnt be achieved with so low powers...

Even the device that get to 20W is well over 7W for the Soc..

 

[mikk]

Numbers are numbers :



Consistently mean permanently, i dont know how you manage to reconcile 30W at the main (with 12" screens items..) permanently with your "7W TDP", indeed the scores couldnt be achieved with so low powers...

Even the device that get to 20W is well over 7W for the Soc..

I (we) refer to the CPU package, the entire device power consumption isn't relevant for CPU package TDP/consumption.

 

[Shivansps]

If the passmark score of FX-8800P @ 15W is actually lower than Carrizo-L A8-7410 (see above post) and the laptop makers (in this case ASUS) are building boards with one DIMM slot (thus limiting the iGPU advantage of FX-8800P over A8-7410) it makes me wonder how much OEMs are paying for these FX-8800P chips?

My guess is not much.

Yes thats exactly what i meant if Carrizo-L runs at 15W and can scale up to 25W, and the Low End Carrizo also runs at 15W, there is no way a FX8800P could run at 15W even remotely OK, and we have no way to know what we are buying because TDP depends on OEM.

My best guess is 15W Carrizo devices are actually designed for Carrizo-L, like that Asus, and OEM are allowed to cram a FX8800P on it and charge a premium for it and everybody wins, except the ones that actually go and buy them. Carrizo just dont belong in sub-25W, Intel kinda forced their hand because they killed 35W M series with Skylake.

thats why I HATE cTDP on both AMD and Intel, give the OEMs a small chance to go cheap and they will take it.

 

[USER8000]

The whole confusion just shows how AMD marketing is not helping things and the Anandtech article has actually stated now that two of the laptops can run dual channel memory. AMD needs to handle its communications with reviewers and the public better.

 

[Shivansps]

The whole confusion just shows how AMD marketing is not helping things and the Anandtech article has actually stated now that two of the laptops can run dual channel memory. AMD needs to handle its communications with reviewers and the public better.

AMD needs to stop doing stupid things and start to show up products that actually perform OK, this whole thing about dual and single channel whould have never happen if Carrizo-L was dual channel, there is just NO reason no keep Carrizo-L single channel, this was something that was already bad with Bobcat because it was a step back from the Neo and Neo 2 platforms, but OK it was its first product, it was hard to understand why with Kabini/Temash, but now with Carrizo-L is just unacceptable.

 

[Abwx]

I (we) refer to the CPU package, the entire device power consumption isn't relevant for CPU package TDP/consumption.

It is relevant for whom has an idea of the other parts power comsumption, among other screens, wifi, RAM and the HDD, and i can tell ou that those items SoCs are well above their rated cTDP, as said 30W in a 12"5 devices induce that the SoC is at least at 18W max TDP.

The whole confusion just shows how AMD marketing is not helping things and the Anandtech article has actually stated now that two of the laptops can run dual channel memory. AMD needs to handle its communications with reviewers and the public better.

These are enginering samples, actualy all the models he tested run dual channel in their commercial versions, so this article has about no relevancy at all even in respect of the very tested laptops..

And AMD marketing has nothing to do about it since the info is public for several months and can be easily found by googling a bit..

 

[cbn]

If the passmark score of FX-8800P @ 15W is actually lower than Carrizo-L A8-7410 (see above post) and the laptop makers (in this case ASUS) are building boards with one DIMM slot (thus limiting the iGPU advantage of FX-8800P over A8-7410) it makes me wonder how much OEMs are paying for these FX-8800P chips?

My guess is not much.

Yes thats exactly what i meant if Carrizo-L runs at 15W and can scale up to 25W, and the Low End Carrizo also runs at 15W, there is no way a FX8800P could run at 15W even remotely OK, and we have no way to know what we are buying because TDP depends on OEM.

My best guess is 15W Carrizo devices are actually designed for Carrizo-L, like that Asus, and OEM are allowed to cram a FX8800P on it and charge a premium for it and everybody wins, except the ones that actually go and buy them. Carrizo just dont belong in sub-25W, Intel kinda forced their hand because they killed 35W M series with Skylake.

thats why I HATE cTDP on both AMD and Intel, give the OEMs a small chance to go cheap and they will take it.

1.) Carrizo-L is the first cat core chip (besides socket AM1) that I have seen listed up to 25W. With Beema and Kabini, the mobile chips were 15W max.

However, looking at the passmark score of 2586 for A8-7410 (Carrizo-L) compared to the passmark score of 2521 for A8-6410 (Beema) it appears that 15W is still commonly be used for Carrizo-L.

2.) It definitely appears that some Carrizo laptops (like the ASUS X555DG which has only one dimm slot) may simply be Carrizo-L designs adapted to run Carrizo. However, as others have pointed out some Carrizo laptops can run dual channel.

 

[mikk]

It is relevant for whom has an idea of the other parts power comsumption, among other screens, wifi, RAM and the HDD

20W for a wall socket measurement on a complete device doesn't look off for device with assumingly configured CPU for 7W, you must take into account a bigger measurement inaccuracy for a wall socket measurement, we can't even be sure if this is remotely accurate from Notebookcheck. Efficiency rate of the power supply might play a role too in this test.

On the other side the reported power for CPU package is most likely pretty much accurate and much more reliable.

CPU TDP is clearly 7-8W. This device consumes average 8W in idle where CPU package power consumption in basically zero on a Skylake ULV. This is coming from LCD, RAM or something else. DRAM is measured with 4W alone under full load.

20W for the entire device is in line with 7-8W CPU, definitely not more. Confirmed by both device power and CPU package power measurement in this test.

 

[USER8000]

These are enginering samples, actualy all the models he tested run dual channel in their commercial versions, so this article has about no relevancy at all even in respect of the very tested laptops..

And AMD marketing has nothing to do about it since the info is public for several months and can be easily found by googling a bit..

The problem is AMD supplied the engineering samples so there lies the problem. I just hope Bristol Ridge,Polaris and Zen are handled better with regards to review samples.

 

[Shivansps]

1.) Carrizo-L is the first cat core chip (besides socket AM1) that I have seen listed up to 25W. With Beema and Kabini, the mobile chips were 15W max.

However, looking at the passmark score of 2586 for A8-7410 (Carrizo-L) compared to the passmark score of 2521 for A8-6410 (Beema) it appears that 15W is still commonly be used for Carrizo-L.

2.) It definitely appears that some Carrizo laptops (like the ASUS X555DG which has only one dimm slot) may simply be Carrizo-L designs adapted to run Carrizo. However, as others have pointed out some Carrizo laptops can run dual channel.

Dual channel is just one thing, cooling is one big factor on a notebook design, if somebody makes a 15W Carrizo-L its likely that Carrizo will also be a 15W product.
Thats probably why now Carrizo-L can scale up to 25W, probably AMD idea was that OEM to design 25W notebooks that can use both.

Anyway, if we take Dell as an example, the A8-7410 and A10-8700P notebook are at the same price, both single channel and probably both are 15W...
And i have a hard time figuring out which one will perform best, and when something like that happens you know you screw up big time.

 

[cbn]

Anyway, if we take Dell as an example, the A8-7410 and A10-8700P notebook are at the same price, both single channel and probably both are 15W...
And i have a hard time figuring out whitch one will perform best, and when something like that happens you know you screw up big time.

Which Dell are you referring to?

EDIT: You must mean this one:

http://www.dell.com/us/p/inspiron-1...cd=123092202532786738&ven3=263503210632344468

Yes, that is interesting that A10-8700p and A8-7410 are the same price.

However, looking back at the Passmark scores I posted here, it doesn't really surprise me. And it pretty confirms what I wrote here.

 

[Abwx]

And it pretty confirms what I wrote here.

It confirm nothing, Carrizo based laptops are signifiantly more expensive than their Carrizo-L counterparts at equal features, to get significative comparisons looking within a manufaturer offering is not enough, one should look at the retail market.

https://geizhals.de/?cat=nb&xf=6749_13#xf_top

https://geizhals.de/?cat=nb&xf=6749_14#xf_top

These are Carrizo and Carrizo L respectively, the lower priced 7410 in a Lenovo is 180€ cheaper than a 8700P, let say 130€ to account for the screens.

At HP the bottom of the barrel 8700P is 100€ more expensive than its 7410 counterpart, from here and accounting for VAT, margins and all intermediary cost those 100€ should yield 40-45€/$ difference for the chip.

 

[cbn]

It confirm nothing, Carrizo based laptops are signifiantly more expensive than their Carrizo-L counterparts at equal features, to get significative comparisons looking within a manufaturer offering is not enough, one should look at the retail market.

https://geizhals.de/?cat=nb&xf=6749_13#xf_top

https://geizhals.de/?cat=nb&xf=6749_14#xf_top

These are Carrizo and Carrizo L respectively, the lower priced 7410 in a Lenovo is 180€ cheaper than a 8700P, let say 130€ to account for the screens.

At HP the bottom of the barrel 8700P is 100€ more expensive than its 7410 counterpart, from here and accounting for VAT, margins and all intermediary cost those 100€ should yield 40-45€/$ difference for the chip.

Just because the Carrizo laptop is more expensive than a Carrizo-L laptop doesn't mean the Carrizo chips cost more.

Example: Logic follows If the production run of Carrizo specific laptops (with dual channel motherboard, etc) is small enough the per unit cost of the laptop could be higher than a larger run of Carrizo-L laptops even if the chip price were the same.

 

[The Stilt]

Just because the Carrizo laptop is more expensive than a Carrizo-L laptop doesn't mean the Carrizo chips cost more.

So you recon it costs the same to manufacture a 107mm² and a 246mm² chip? :sneaky: