AMD Carrizo Pre-release thread

[ShintaiDK]

Not sure why you're describing this as an AMD-only-issue. Intel also has CPUs with various selectable TDP levels for the same chip; cTDP, SDP and whatever.

Read again instead of skipping the parts to twist the context on purpose.

Its not an AMD issue. But its an issue when people try to compare apples and oranges.

 

[ShintaiDK]

We won't compare chips, we'll compare products. Just like we do in Intel's case.

Exactly:
http://forums.anandtech.com/showpost.php?p=37583898&postcount=2029

 

[Dresdenboy]

The problem is people try to compare 35W chips, with or without cTDP down with 15W chips, with or without cTDP down.

One can just claim the 15W 5200U Broadwell-U chips run at 7.5W.

But as seen with a few posters, nice way to use 35W numbers against 15W or sub 15W numbers to distort reality.

I see following way to go from here:
* cite sources
* mention given TDP configuration
* if no TDP config is given, mention that too

 

[maarten12100]

We won't compare chips, we'll compare products. Just like we do in Intel's case.

Agreed now if OEMs would make decent products with Carrizo. HP has shown that even if a chip is crap you can make a good product with it. (See the Kaveri based Probook)

This time the chip has the biggest leap ever from an energy efficiency perspective allegedly which is putting it ahead of most of Intel's lineup allegedly and no OEM bats an eye. I wonder :whiste:

Exactly:
http://forums.anandtech.com/showpost.php?p=37583898&postcount=2029

The conclusion is beyond stupid in that particular post comparing to a product gulping down 25-30W when running those benches then claiming the 15W chip performs far less. I will take back these words once we have some notebook in hand and they also show 25-30W consumption but for now I'm going to declare that statement right there invalid until we have actual load number to go with some Carrizo notebooks.

 

[ShintaiDK]

This time the chip has the biggest leap ever from an energy efficiency perspective allegedly which is putting it ahead of most of Intel's lineup allegedly and no OEM bats an eye. I wonder :whiste:

Maybe because its not as good as you think.

 

[Abwx]

Exactly:
http://forums.anandtech.com/showpost.php?p=37583898&postcount=2029

http://www.notebookcheck.net/Intel-Core-i5-5200U-Notebook-Processor.127831.0.html

So people can check the corresponding power numbers...

Even the lowest scoring at 2.23 has a chip that get over 15W, the rest i wont even talk about it other than to point that the chips are in the 25-28W range.

 

[dark zero]

http://www.notebookcheck.net/Intel-Core-i5-5200U-Notebook-Processor.127831.0.html

So people can check the corresponding power numbers...

Even the lowest scoring at 2.23 has a chip that get over 15W, the rest i wont even talk about it other than to point that the chips are in the 25-28W range.

That's why I avoid the U tier... is not always saying the truth.... on the other hand.. WHY AMD LEFT THE SOCKETED CHIPS?
Sorry if that was so much, but that is retarded... socketed chips are better than soldered ones.

 

[mrmt]

This time the chip has the biggest leap ever from an energy efficiency perspective allegedly which is putting it ahead of most of Intel's lineup allegedly and no OEM bats an eye. I wonder :whiste:

This. The answer you are looking for is in your text already: You are assuming that AMD claims are true. OEMs shunning the product are a strong, almost conclusive indication that the clams are not.

 

[slaven]

http://users.atw.hu/instlatx64/AuthenticAMD0660F01_K15_Carrizo_CPUID.txt - QuadCore AMD FX-8800P, 2500 MHz (25 x 100) (Carrizo) 660F01 CPUID dump


MSR C0010064: 8000-0183-0000-4812 [34.00x] [1.3250 V] [13.10 A] [PState Pb0]
MSR C0010065: 8000-0174-0000-5C10 [32.00x] [1.2625 V] [11.60 A] [PState Pb1]
MSR C0010066: 8000-015D-0000-7C0D [29.00x] [1.1625 V] [ 9.30 A] [PState Pb2]
MSR C0010067: 8000-0144-0000-A409 [25.00x] [1.0375 V] [ 6.80 A] [PState Pb3]
MSR C0010068: 8000-013A-0000-B407 [23.00x] [0.9875 V] [ 5.80 A] [PState Pb4]
MSR C0010069: 8000-0133-0000-C005 [21.00x] [0.9500 V] [ 5.10 A] [PState P0]
MSR C001006A: 8000-0128-0000-D854 [18.00x] [0.8750 V] [ 4.00 A] [PState P1]
MSR C001006B: 8000-011D-0040-F04C [14.00x] [0.8000 V] [ 2.90 A] [PState P2]

* * *

At 3400 MHz CPU takes 1.3250 V * 13.10 A = 17.4 Watt. So, if its true, all other Watts are for Video, NB, SB and other uncore.

 

[maarten12100]

Maybe because its not as good as you think.

Those were all statements made by AMD or AMD employees so it doesn't have to do what I think. If they lied they will be sue-ed for that reason it must perform until proven it doesn't.

I don't think a laptop set to 15 or 20W and gimped with a single stick of ram for a dual channel 128bit memory controller and comparing it to a laptop that consumes 25-30W under load is viable. Even more than that we don't even have CPU load power number on those laptops and when they come NBC will only give numbers where all is loaded as in the secondary gpu is too and the igp is too giving no indication of the power consumption under those CPU benches.

That is why I'm hoping for a laptop with just the FX-8800p and no garbage 384/512 SP tonga based dgpu. I'm willing to accept a good product and there is no point of letting go of this mindset until it is proven to me that Carrizo isn't a good product.

Again bad notebooks don't tell me anything and too little information is just that.

This. The answer you are looking for is in your text already: You are assuming that AMD claims are true. OEMs shunning the product are a strong, almost conclusive indication that the clams are not.

Of course I do and will continue to until proven false. I'm cautioned as always though.

And OEMs shunning to me is more of an indication of how stupid/bribed OEMs are. They make plastic chassis-es for 500+ notebooks while the production cost of a aluminium chassis are 20 euros. Actually 5 if you do powder pressing and only mill the screwholes. Additionally they use HDDs even in high end offerings. Their TIM applying skills need work their cooler design and fan choice is poor. Their choice of ram is poor. Their choice of screen is beyond poor. They use weak no response keyboard for the sake of shaving of a single millimeter. And their trackpads are almost all equally bad. They use small batteries and act like batteries cost a lot of money.

Apple is crushing because they made a quality product and priced it accordingly due to their branding power. If Asus would make top to bottom products of great quality they could at least get 20% of the sales Apple has.

Also from all companies the lack of pursuing new and exciting technology baffles me. As a university student I have free access to Sciencedirect along with other databases and all those great ideas only get them patents but never actually products. We are talking things that have been established many years ago.

/rant

http://users.atw.hu/instlatx64/AuthenticAMD0660F01_K15_Carrizo_CPUID.txt - QuadCore AMD FX-8800P, 2500 MHz (25 x 100) (Carrizo) 660F01 CPUID dump

At 3400 MHz CPU takes 1.3250 V * 13.10 A = 17.4 Watt. So, if its true, all other Watts are for Video, NB, SB and other uncore.

For one core, right?
So that bring the total power on a cpu ST bench to 20-25W right?
(due to screen, uncore and the negligible gpu load.

 

[slaven]

maarten12100, don't know. You can compare the results to other (Kaveri for example) on that site and find out.

 

[ShintaiDK]

Those were all statements made by AMD or AMD employees so it doesn't have to do what I think. If they lied they will be sue-ed for that reason it must perform until proven it doesn't.

Do you need a long list with claims they was never sued for, despite being blatantly wrong? Its pretty much been the company mantra the last 10 years.

Nobody gonna sue them for something like that.

 

[Abwx]

http://users.atw.hu/instlatx64/AuthenticAMD0660F01_K15_Carrizo_CPUID.txt - QuadCore AMD FX-8800P, 2500 MHz (25 x 100) (Carrizo) 660F01 CPUID dump


MSR C0010064: 8000-0183-0000-4812 [34.00x] [1.3250 V] [13.10 A] [PState Pb0]
MSR C0010065: 8000-0174-0000-5C10 [32.00x] [1.2625 V] [11.60 A] [PState Pb1]
MSR C0010066: 8000-015D-0000-7C0D [29.00x] [1.1625 V] [ 9.30 A] [PState Pb2]
MSR C0010067: 8000-0144-0000-A409 [25.00x] [1.0375 V] [ 6.80 A] [PState Pb3]
MSR C0010068: 8000-013A-0000-B407 [23.00x] [0.9875 V] [ 5.80 A] [PState Pb4]
MSR C0010069: 8000-0133-0000-C005 [21.00x] [0.9500 V] [ 5.10 A] [PState P0]
MSR C001006A: 8000-0128-0000-D854 [18.00x] [0.8750 V] [ 4.00 A] [PState P1]
MSR C001006B: 8000-011D-0040-F04C [14.00x] [0.8000 V] [ 2.90 A] [PState P2]

* * *

At 3400 MHz CPU takes 1.3250 V * 13.10 A = 17.4 Watt. So, if its true, all other Watts are for Video, NB, SB and other uncore.

Power above is in ST instance, at 15W max frequency is 3.3GHz, 3.4 require 17W so TDP must be extended to the next cTDP state to get the max ST frequency..

While we re at it Notebookcheck numbers are 100% made up apparently, what they did is to apply the %ages published by AMD for Carrizo to a Kaveri equiped IdeaPad Z50-75 numbers that they extracted from nowhere, and wich was averaged with the single actual review they made of this SKU :

http://www.notebookcheck.net/AMD-Kaveri-FX-7500-Notebook-Processor.117332.0.html

They didnt even use this truncated average, they had to pick the imaginary review numbers, of course...

Imaginary because a IdeaPad Z50-75 doesnt perform like this, even with a lower grade Kaveri than the FX7500 :

http://www.notebookcheck.net/Lenovo-IdeaPad-Z50-75-Notebook-Review-Update.127031.0.html


........

 

[Abwx]

And OEMs shunning to me is more of an indication of how stupid/bribed OEMs are. They make plastic chassis-es for 500+ notebooks while the production cost of a aluminium chassis are 20 euros. Actually 5 if you do powder pressing and only mill the screwholes.

AMD are currently winning back some marketshare thanks to relatively low cost and good quality of HP offerings among others, rest is propaganda from usual suspects that never produce any number that would back their blanks statements.

And It s no surprise since they know too well that all they are posting is fud, just browse this thread if you like reading through marketing motivated urban legends and myths, notice that the concept of number is totaly absent form thoses purely textual posts...

 

[maarten12100]

maarten12100, don't know. You can compare the results to other (Kaveri for example) on that site and find out.

Compared to 7850K results seems like those were indeed ST numbers.
For the closest clock listed which is 3,5GHz I get 21,25 W
[35.00x] [1.2500 V] [17.00 A] [PState P1]

That makes an improvement of about 20% in terms of power consumption at a given clock.
Of course even at ST scenarios Kaveri actually does more like 50W in power consumption + 20W platform power at those clocks. See A8-7600 for reference.
Obviously not all the gains are found at this level but it seems like it matches that curve at such a high clock.

To confirm this thesis about it matching the curve we can compare at different clocks so let's take the lowest clock. At which the curve favours excavator.

Kaveri 1700MHz
[17.00x] [0.9000 V] [ 5.70 A] [PState P4]

Carrizo 1800MHz
[18.00x] [0.8750 V] [ 4.00 A] [PState P1]

We get 3,5W and 5,1W respectively the difference is now ~46,6% in favour of Carrizo at core level. This matches the curve found in the slide.

Where the steamroller line begins matching it up with the excavator line we get these 2 ratios which are 0,65 / 0,46 = ~1,41 in other words according to the slide it should be 41% ahead of steamroller in terms of clock at a low power level. This matches the data we found.

However I'm going to place a critical side note here:
The slide says Core Pair and I calculated using the results of a single core. I therefore expect the actual power consumption for a pair of cores to match the 41% found rather than the 46,6% I found by the CPUID dump.

Do you need a long list with claims they was never sued for, despite being blatantly wrong? Its pretty much been the company mantra the last 10 years.

Nobody gonna sue them for something like that.

They were sue-ed for Llano while that was a very good product. If investors smell money they will bite.

You want such a list for Intel. Which do you think is longer? That doesn't make it right though. Bulldozer comes to mind as a lie by their PR teams. The engineers were screaming it failed there was too little time but marketing gave the go ahead an marketed as a good product.

 

[frozentundra123456]

And the spin begins. In any case, just the skeptic in me I guess, but I take any manufacturer's claim with a grain if salt until I see the confirmatory results in real world shipping products. So according to this thread, all the claims of Intel chips using more power than they say must be false as well, because, well, everything manufacturers say is true, otherwise they would be sued, right? And all the great claims for Skylake must be true too, because manufacturer's never lie, or cherry pick the absolute most favorable data, because they would get sued, right?

Honestly, this new tack that one must believe every marketing claim of AMD (or anyone else) because of the possibility they would be sued is one of the most absurd rationalizations yet to come to these forums. If this were true, AMD, Intel, and pretty much every other manufacturer would have time for nothing but defending themselves from lawsuits.

 

[shady28]

Are there any reviews / benchmarks on these chips? Every site I just went to has nothing but regurgitation of AMDs PR slides.

 

[ShintaiDK]

They were sue-ed for Llano while that was a very good product. If investors smell money they will bite.

Llano was due to an inventory writedown, while telling investors that it sold like hotcakes.

Nothing to do with any performance claims etc.

 

[mikk]

Mikk, you got them p.... down!
See from the first google hit:

http://www.anandtech.com/show/4764/ivy-bridge-configurable-tdp-detailed

Did you read properly what I said? I have to repeat me. This is much more restricted and limited on Intel and your links just proves it. Thanks for the proof.

If they lied they will be sue-ed for that reason it must perform until proven it doesn't.

There was a lawsuit after the Bulldozer IPC claims?

Honestly, this new tack that one must believe every marketing claim of AMD (or anyone else) because of the possibility they would be sued is one of the most absurd rationalizations yet to come to these forums. If this were true, AMD, Intel, and pretty much every other manufacturer would have time for nothing but defending themselves from lawsuits.

Yes this is really absurd. It's a typically AMD argument.

 

[Abwx]

Of course even at ST scenarios Kaveri actually does more like 50W in power consumption

An early 7850K consume 67W under Prime 95, 54W on regular MT softs, that is at 3.7GHz, on ST at 4GHZ it consume 28W.

All numbers at stock voltages, 1.352V for the 3.7GHz base frequency, and wich can be checked at Hardware.fr review of the 7850K.

 

[Arachnotronic]

This back and forth on Carrizo v.s. Intel is getting pretty dull.

Perhaps a change of discussion topic is in order?

Who here is planning on buying a Carrizo-based laptop? What models of the ones that we know about are you currently considering, if your answer to the last question is "yes"?

 

[Abwx]

This is much more restricted and limited on Intel and your links just proves it. Thanks for the proof.

I read 47W TDP on the left of the screen shot and i see 100W+ for the SoC on the right..

People want to discuss technical matters about Carrizo and are not interested in your constant whining, post in Intel related threads since it s all about promoting Intel...

Now can you stop polluting this thread with your myths and other Intel propaganda..?..

 

[maarten12100]

There was a lawsuit after the Bulldozer IPC claims?

Obviously there should've been. Pretty sure they sacked their marketing though.

 

[Abwx]

Of course even at ST scenarios Kaveri actually does more like 50W in power consumption

An early 7850K consume 67W under Prime 95, 54W on regular MT softs, that is at 3.7GHz, on ST at 4GHZ it consume 28W.

All numbers at stock voltages, 1.352V for the 3.7GHz base frequency, and wich can be checked at Hardware.fr review of the 7850K.

Those numbers are somewhat the worst possible since GF process power/frequency start exiting from a square law as early as 3GHz..

That s why Haswell s better efficiency at DT level over Kaveri slowly shrink as power is reduced, to the point that at 15W its perf/Watt advantage is barely 15%, and is due to lower leakage that allow to better feed the cores..

This back and forth on Carrizo v.s. Intel is getting pretty dull.

Perhaps a change of discussion topic is in order?

Who here is planning on buying a Carrizo-based laptop? What models of the ones that we know about are you currently considering, if your answer to the last question is "yes"?

Fact is that only 15"6/17" are available at this point because it s the main market, as people are buying them as main PCs concurrently with AIOs wich is why HP also used it for this format.

I guess that most people here seems more interested by smaller formats, including me of course, but i wouldnt be surprised that HP will offer a laptop based A8-8600P with no dGPU at all, typicaly in the format below :

www.notebookcheck.net/HP-Pavilion-13-a093na-x360-Convertible-Review-Update.130928.0.html

Set apart for the screen that would be a very adequate offering price/perfs wise and an opportunity for firms like Acer that did lost ground in some markets.

At some point there are people interested by a given set up, and whoever who provide it will make $$, HP seems to have enforced this logic, contrary to Asus for instance that will not release Carrizo laptops in the US according to the last news, i guess that they risk the same beating there as what Acer endured last quarter in the EMEA...

 

[Dresdenboy]

Did you read properly what I said? I have to repeat me. This is much more restricted and limited on Intel and your links just proves it. Thanks for the proof.

Ok, you didn't get the fun.

This was just the first hit and about IB, way too old to support your point. Please show us, why Intel went backwards after this development.