AnandtechAMD Carrizo ExcavatorReview

[jhu]

Jhu, how do the results fits there?

They're not comparable since different Blender versions produce different render times and the older file was used in that list. However, I did run the newer file (BMW27.blend) on a Core i5 2400S with a time of 9 minutes 14.69 seconds. This gives us:

Core i5 2400S:   35945 samples/s/GHz/core
15/25W/DDR-1600: 34475 samples/s/GHz/module
35/42W/DDR-2133: 34030 samples/s/GHz/module

Looks like Excavator IPC still isn't great. I'll run this on my FX8350 and put results up later.

 

[DrMrLordX]

14.04.3 lts.

Okay . . . unless you've manually updated the kernel, you aren't going to get amdgpu with that. You need at least 15.04. I would recommend 15.10, especially if all you're doing is testing (read: you really don't need the long term support to run some benches).

But we expect the corporate executives whom Anandtech was working with to save the company. It's these same executives that are responsible for not making sure AMD was put in the best light for this article.

Personally, I have no confidence in anyone but the engineers working there, and secondarily the support the engineering teams are getting from execs like Su who should at least nominally understand what it is to be an engineer. I have no idea who was responsible for providing hardware to AT for the review. Whoever it is deserves a tongue-lashing, at the very least.

Nope, this time I'll go with the more likely hypothesis: AMD marketing is professionally retarded.

I hate to say it, but yeah, it's looking that way. There just don't seem to be any people in there who really understand or care about little marketing minutiae that can make a big difference down the road. The whole way that Carrizo has been handled by OEMs and reviewers reveals much about deficiencies within the skill sets of those remaining at AMD.

 

[Phynaz]

As you mentioned before, they still missed the 2.4X power efficiency goal. Could that have been their focus - to get closer to that figure than showing high performance?
How much power consumption increase can be expected from adding a second DIMM? Is it in the 1 or 2W range incl. the additional SOC NB/I/O power?

There is also this unlikely hypothesis: AMD marketing purposely let it look like its product got unfairly reviewed on a well known enthusiast site to activate evangelists among the enthusiasts spreading the word, that the products are better than shown. This get's the "word of mouth" tag, with all its implications.

Your first point is a good observation, and makes sense.

The conspiracy theory is too far out there for me

 

[Abwx]

How can you get closer to your power efficiency goal when using 5-10% less energy drops performance by 30% or more?

It is the other way around, when starting from energy the ratio must be root squared, so 5-10% less energy (0.95-0.9) yield 2.5-5% less performance...

 

[Abwx]

15/25W/DDR-1600 = 20m 2.95s
35/42W/DDR-2133 = 15m 23.23s

Default settings.

OpenCL could not be tested since there is no display driver which could be installed. The drivers from AMD are blocked from installing since they are "bad of quality"

About 30% difference.

They're not comparable since different Blender versions produce different render times and the older file was used in that list. However, I did run the newer file (BMW27.blend) on a Core i5 2400S with a time of 9 minutes 14.69 seconds. This gives us:

Core i5 2400S:   35945 samples/s/core
15/25W/DDR-1600: [B]34475 samples/s/module[/B]
35/42W/DDR-2133: [B]34030 samples/s/module[/B]

Looks like Excavator IPC still isn't great. I'll run this on my FX8350 and put results up later.

About 1% difference.

 

[AtenRa]

As you mentioned before, they still missed the 2.4X power efficiency goal. Could that have been their focus - to get closer to that figure than showing high performance?
How much power consumption increase can be expected from adding a second DIMM? Is it in the 1 or 2W range incl. the additional SOC NB/I/O power?

There is also this unlikely hypothesis: AMD marketing purposely let it look like its product got unfairly reviewed on a well known enthusiast site to activate evangelists among the enthusiasts spreading the word, that the products are better than shown. This get's the "word of mouth" tag, with all its implications.

Conspiracy theory number two,

AMD gave those craptops to be reviewed on purpose, knowing how awful they will performance, in order to make BristolRidge look even better.

If latest info is correct and A12-9830 is a 35W TDP 3.0GHz base 3.7GHz turbo with 2400MHz dual channel ram, it will look like its Conroe II vs those Carrizo craptops in the AT review.

 

[mrmt]

As you mentioned before, they still missed the 2.4X power efficiency goal. Could that have been their focus - to get closer to that figure than showing high performance?
How much power consumption increase can be expected from adding a second DIMM? Is it in the 1 or 2W range incl. the additional SOC NB/I/O power?

There is also this unlikely hypothesis: AMD marketing purposely let it look like its product got unfairly reviewed on a well known enthusiast site to activate evangelists among the enthusiasts spreading the word, that the products are better than shown. This get's the "word of mouth" tag, with all its implications.

This is not a Carrizo review at least strictu sensu, as Anandtech wasn't reviewing all the theoretical capabilities of Carrizo chips, but what actual OEMs decided to deliver in terms of actual producs, regardless of the theoretical limits of the platform.

 

[USER8000]

Conspiracy theory number two,

AMD gave those craptops to be reviewed on purpose, knowing how awful they will performance, in order to make BristolRidge look even better.

If latest info is correct and A12-9830 is a 35W TDP 3.0GHz base 3.7GHz turbo with 2400MHz dual channel ram, it will look like its Conroe II vs those Carrizo craptops in the AT review.

I hope for their sake that is true,but at 15W TDP only 1866MHZ memory is supported:

http://cdn.wccftech.com/wp-content/uploads/2016/02/AMD-Bristol-Ridge-FP4-Family.png

But even then sites like notebookcheck have already tested Carrizo at a 15W TDP:

http://www.notebookcheck.net/HP-Pavilion-15-ab052ng-Notebook-Review.152289.0.html

If you read the review,AMD actually gave them an extra SODIMM to test the laptop in dual channel. That is from last year.

So the Y700 and some of those HPs tested seem to be not representative of the actual production laptops out there.

Hence,sites like them will compare Carrizo in dual channel to Bristol Ridge in dual channel.

Plus,if OEMs can't sell Carrizo laptops,especially when they were among the first AMD laptops to ship with higher resolution and/or IPS screens,it's not going to give much confidence in OEMs to invest in the more expensive Bristol Ridge design.

 

[Beer4Me]

AMD is quickly beginning to look like ViA.

 

[el etro]

Mea culpa. Now I see, what you meant. Yeah, it looks like the reference design was better then. Or this Kaveri system wasn't the worst one in efficency.

For me its not really misleading, the slide just don't disclose the things as detailed as it should be for more people understand.

The "typical-use" energy efficiency probably mean the use in low, with bursts workloads such as web-browsing, office, and multitasking. A load that is low at a average, with occasional very high spikes(maybe the best-case AVFS work). Like most PC users use is.
I think it markets the benefits of the power saving techniques they implemented in Carrizo.

 

[cbn]

OK let me say here that all this crap is AMD FAULT and im dead serious this time.

DONT let the OEM destroy you, but that is exactly what the OEMs did to Carrizo. They completely destroyed a very nice product with awfully bad designed Laptops.

Im asking for a $400-500 35W TDP dual channel 2133MHz iGPU only 1600x900 14" Carrizo laptop that could make AMD and its product rise above the competition. But nooooo, OEMs know best. They create those abominations that nobody here and i mean nobody here even a die hard AMD fan boy would ever want to buy.

If AMD doesnt change its priorities and understand that if the public doesnt want to buy your products, OEMs would only make you look worse if you dont put guidelines.

Maybe AMD can assist the process of creating 35W Bristol Ridge laptops by having Stoney Ridge at 35W as well?

Then with enough Bristol Ridge 35W processors available the OEM can set-up the motherboard for dual channel.

A dual channel motherboard can be used by single channel processor right? (I would assume the cost is a bit more, but with enough 35W Bristol Ridge available it should be worth it).

 

[mrmt]

For me its not really misleading, the slide just don't disclose the things as detailed as it should be for more people understand.

The "typical-use" energy efficiency probably mean the use in low, with bursts workloads such as web-browsing, office, and multitasking. A load that is low at a average, with occasional very high spikes(maybe the best-case AVFS work). Like most PC users use is.
I think it markets the benefits of the power saving techniques they implemented in Carrizo.

This is exactly what would make the slide misleading.

 

[Dresdenboy]

For me its not really misleading, the slide just don't disclose the things as detailed as it should be for more people understand.

The "typical-use" energy efficiency probably mean the use in low, with bursts workloads such as web-browsing, office, and multitasking. A load that is low at a average, with occasional very high spikes(maybe the best-case AVFS work). Like most PC users use is.
I think it markets the benefits of the power saving techniques they implemented in Carrizo.

I already looked into this. See

That depends on the equations used (not our simple comparison of max power at a task). According to the footnotes, AMD used the ETEC equation 1 given on page 10 here:
http://www.energystar.gov/ia/partners/product_specs/program_reqs/Computers_Program_Requirements.pdf

Then they divide some an SpecIntRate, Passmark and PCMark score by that number. Passmark is heavily influenced by IDIV latency, PCMark uses GPGPU, SpecIntRate probably runs well out of the L2 cache. But for their own tests, they usually used 2 DIMMs.

 

[Shivansps]

Amd should have added a boot block to carrizo, so it can only boot up in dc.

 

[MiddleOfTheRoad]

AMD is quickly beginning to look like ViA.

Actually, I'm looking forward to getting my hands on an Athlon X4 845 -- an Excavator chip not totally neutered by cheapskate OEM's. The single channel memory is just stupid... Especially for a chip whose main strength is integrated graphics (which needs as much bandwidth as possible).

 

[DrMrLordX]

Amd should have added a boot block to carrizo, so it can only boot up in dc.

You're at least the third person to come out and say this, and of course, you're correct. That would have solved a lot of problems.

But I've already gone on too much about what they should have done with Carrizo (and/or what they should do with mobile Bristol Ridge). Either they wise up or they don't . . . everything they need to know has already been said.

 

[DeathReborn]

You're at least the third person to come out and say this, and of course, you're correct. That would have solved a lot of problems.

But I've already gone on too much about what they should have done with Carrizo (and/or what they should do with mobile Bristol Ridge). Either they wise up or they don't . . . everything they need to know has already been said.

It has but quite frankly any x86 CPU with a IGP should as a minimum come with DC memory or SC and HBM (or EDRAM).

 

[dark zero]

It has but quite frankly any x86 CPU with a IGP should as a minimum come with DC memory or SC and HBM (or EDRAM).

Agreed. Even Intel should release all their Graphics with at least 32 Mb of EdRAM

 

[jhu]

Core i5 2400S:   35945 samples/s/GHz/core
15/25W/DDR-1600: 34475 samples/s/GHz/module
35/42W/DDR-2133: 34030 samples/s/GHz/module

Just ran this (2.76b) on Core i5-6200U (9m 15s) and FX8350 (5m 40s):

Core i5-6200U:   69189 samples/s/GHz/core
FX8350:    37187 samples/s/GHz/module

I don't think the sustained clockspeeds on the Carrizo machines are correct. Excavator should have higher IPC than Piledriver.

 

[Dufus]

Question, why does this CPU-z ver 1.75 pic show a Lenovo Y700-15ACZ with dual channel RAM if it's not supported on that model?

http://forum.notebookreview.com/threads/ideapad-y700-15acz-review.783192/page-5#post-10196477

 

[Shivansps]

Its likely there are 2 models of Y700... Still all Dell Carrizo models are ST only :S

 

[Abwx]

Question, why does this CPU-z ver 1.75 pic show a Lenovo Y700-15ACZ with dual channel RAM if it's not supported on that model?

http://forum.notebookreview.com/threads/ideapad-y700-15acz-review.783192/page-5#post-10196477

The AMD based Y700 has dual channel but is delivered with a single RAM stick with a slot being unpopulated, on the other hand the Intel based Yoga 700 is genuinely single channel with a single slot :

http://www.expertreviews.co.uk/lapt...1in-review-different-yoga-same-flaws/page/0/1


Should had the reviewer at AT done some basic googling that he would had noticed that it s the Intel version that is single channel crippled, instead his article is used as a mean to spread fud on AMD using incorrect and even inverted realities:

http://fudzilla.com/news/processors/39903-oems-appear-to-kill-amd-s-carrizo

 

[jhu]

The AMD based Y700 has dual channel but is delivered with a single RAM stick with a slot being unpopulated, on the other hand the Intel based Yoga 700 is genuinely single channel with a single slot :

http://www.expertreviews.co.uk/lapt...1in-review-different-yoga-same-flaws/page/0/1


Should had the reviewer at AT done some basic googling that he would had noticed that it s the Intel version that is single channel crippled, instead his article is used as a mean to spread fud on AMD using incorrect and even inverted realities:

http://fudzilla.com/news/processors/39903-oems-appear-to-kill-amd-s-carrizo

If you had read the original arricle:

We had four Carrizo devices on hand to test for a week, along with a single Kaveri system. These devices were sourced by AMD

So AMD provides the laptops for review. Then from the fudzilla article:

Of course AMD fanboys will be claiming that there is a conspiracy by the OEMs to back Intel, but that is rather daft. It is simply that the OEMs are hoping to trouser the extra cash by designing a cheap and cheerful system and sell it against a more expensive Intel model and hope that users will not notice.

The ones who are being mean to AMD are the OEMs, not the article author. Whatever AMD was trying to do with a unified platform for Carrizo/Carrizo-L, it has backfired.

 

[USER8000]

If you had read the original arricle:


So AMD provides the laptops for review. Then from the fudzilla article:


The ones who are being mean to AMD are the OEMs, not the article author. Whatever AMD was trying to do with a unified platform for Carrizo/Carrizo-L, it has backfired.

Its more like AMD marketing has backfired:

http://www.notebookcheck.net/HP-Pavilion-15-ab052ng-Notebook-Review.152289.0.html

They even supplied a SODIMM so that Notebookcheck could test a 15W TDP Carrizo system in dual channel.

Yet,now all the systems they had on hand have issues running dual channel memory since most appear to be pre-production laptops and AMD is not aware it seems the production ones are fine!!

 

[Abwx]

If you had read the original arricle:


So AMD provides the laptops for review. Then from the fudzilla article:


The ones who are being mean to AMD are the OEMs, not the article author. Whatever AMD was trying to do with a unified platform for Carrizo/Carrizo-L, it has backfired.

Can you provide a relevant leak that say that it has backfired, all i read are speculations but not a single commercial model is given as exemple...

As posted ad nauseam can you point me a laptop in the list below that would be single channel..?.


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

Here in Europe that s about the only ones available, the 459€ HP is the one reviewed recently at NBC and has dual channel....