AMD Carrizo Pre-release thread

[frozentundra123456]

Again you just make up nonsense to support your flawed claims.

Even if one accepts the data of some forum poster for a product that you cant even buy yet, according to TH, Cinebench 11.5 scores are as follows:

i5 5200U = 2.85, i7 5600U = 3.13.


Even the i3 scores 2.3, so I dont know where the poster is getting the 20% difference, even comparing to the lowest end i3.

 

[Sweepr]

~2.75 points @ 15W? My Ivy Bridge ULV chip posted this score back in 2012. Not even close to Broadwell-U, a Core i7 5500U which is not the fastest bin scores 3.25 and will trash Carrizo in any ST task. To make things worse it will compete with Skylake-U, not Broadwell-U.

 

[Abwx]

i5 5200U = 2.85, i7 5600U = 3.13.

You could as well write that it s 50, where are the power graphs and measurements...??.

Nowhere to be find since those numbers are achieved at way more TDP than 15W, no wonder that Intel, contrary to AMD didnt publish the slightest tests or power curves, heck we had to read AMD s financial paper to read that :

Testing by AMD Performance Labs using an AMD FX-8800P with AMD Radeon™ R7 graphics, 2x4 DDR3-2133, 256 GB SSD, Windows 8.1 64bit, driver 15.10 scored 2753 in 3DMark® 11 performance. Core™ i7 5500U with HD 5500 graphics, DDR3-1600, 256 SSD, Windows 8.1 64bit, driver 4156 scored 1350 in 3DMark 11 Performance.

Half the score in 3DMark...

If the process was that efficient it would extend to the GPU effiency and we wouldnt see such a large gap..

So we have "ultra efficient" transistors lose all their qualities once they are used for a GPU....

 

[Abwx]

Not to mention:

To mention that i used a much more favourable case than what IDC posted since he include second order effects that i downplayed, as i m talking of relatively low frequencies where they have low influence...

 

[Abwx]

~2.75 points @ 15W? My Ivy Bridge ULV chip posted this score back in 2012. Not even close to Broadwell-U, a Core i7 5500U which is not the fastest bin scores 3.25 and will trash Carrizo in any ST task. To make things worse it will compete with Skylake-U, not Broadwell-U.

Score is a little above 2.4 at 15W, 2.75 is for extended TDP a la Intel.

65W for a DT BDW at 3.3GHz and you think that half this chip would be at 15W in the vicinity of 3GHz..??..

More likely 32-35W given that only part of the CPU is halved, but whatever keep on spreading such claims that are auto contradicted...

Now looking at real numbers there s a 11.6 item that has a BDW limited at 8W, yet it start Cinebench at up to 59W, Notebookcheck retrived the pic but i saved it before they did so...

And i m still waiting to see Intel s own numbers, so far the only one is a 1.6x better parf/Watt that has obviously no substance as aknowledged by Intel complete silence on the subject.

 

[Sweepr]

Half the score in 3DMark...

If the process was that efficient it would extend to the GPU effiency and we wouldnt see such a large gap..

So we have "ultra efficient" transistors lose all their qualities once they are used for a GPU....

Except that score comes from the fastest Carrizo bin at 35W TDP. Are you being dishonest now too? First the 20% better than Broadwell-U @ CB11.5 joke which is not even true for MT scores (let alone ST) and now this.

Another handpicked NotebookCheck extrapolation using the lowest-bin Broadwell chips? Sorry but this is different from some of the forums where you post where people worship AMD without questioning anything.

 

[ShintaiDK]

Its not the first time he mixes 35W and 15W on purpose.

 

[Sweepr]

Its not the first time he mixes 35W and 15W on purpose.

Expect a lenghy and creative explanation on why AMD's 35W are lower than Intel's 15W. Grabs some popcorn.

 

[Abwx]

Except that score comes from the fastest Carrizo bin at 35W TDP.
Are you being dishonest now too?

I didnt imply that the 3DMark score was at 15W but that the score is twice Intel s in AMD comparison, and i dont think that they did run the i7 at 15W but in similar conditions.

Also i pointed that AMD advantage is at low power, at 35W i pointed that Intel should have a slight lead CPU wise.

First the 20% better than Broadwell-U @ CB11.5 joke which is not even true for MT scores (let alone ST) and now this.

Another handpicked NotebookCheck extrapolation using the lowest-bin Broadwell chips? Sorry but this is different from some of the forums where you post where people worship AMD without questioning anything.

Contrary to you i check power usage before posting a score, that s what you call cherry picking, isnt it, to not cherry pick is to take the scores at face value once it s intel, nevermind that the laptops consume 3x the alleged TDP of the SKU...

So far for AMD we have very accurate numbers thanks to someone who test this plateform, on the other hand with Intel there s zero informations and absolutely no communication from this brand, if BDW was that efficient no doubt that they would had released a comparison showing that it s better than Carrizo.

 

[SAAA]

1.6x the frequency mandate 2.56x more power, so if it s genuinely 15W at 2GHz it will be 38W at 3.2GHz, there s absolutely no way that it could be otherwise unless our universe physical laws are completely changed.
...

I'm pretty sure a Broadwell dual core ULV at turbo speed is consuming less than 38W:

Here a -quad- core with a turbo of 3.7GHz is consuming around 67W more from idle running AVX load, what force on earth could make a slower dual core use much more than 15W at cinebench?
Using the same square law I get at most a 25W delta for 3.2GHz (but with this AVX load) while at more reasonable base clock of 2.6GHz it's ~16W...
I don't know how much the idle values are but sure as heck not more than the load delta.

 

[mrmt]

Expect a lenghy and creative explanation on why AMD's 35W are lower than Intel's 15W. Grabs some popcorn.

It doesn't matter that OEMs are shunning Carrizo more than any other AMD processor in their recent history to the ADF see how bad this product is.

 

[Abwx]

I'm pretty sure a Broadwell dual core ULV at turbo speed is consuming less than 38W:

What i said is that from 2 to 3.2 comsumption rise by 156%, if te starting point is 15W as in the exemple given by ShintaiDK the finish will be at 38W.



Here a -quad- core with a turbo of 3.7GHz is consuming around 67W more from idle running AVX load, what force on earth could make a slower dual core use much more than 15W at cinebench?
Using the same square law I get at most a 25W delta for 3.2GHz (but with this AVX load) while at more reasonable base clock of 2.6GHz it's ~16W...
I don't know how much the idle values are but sure as heck not more than the load delta.

AVX loading will often throttle Intel CPUs, hence the relatively low power that can be measured.

As for power comsumption and scores here two laptops using the same CPU with scores of 2.23 and 2.85, this imply 64% difference in TDP, unfortunately Dell seems to not allow power measurements of their samples but the difference amplitude is telling, undoubtly thoses chips are often working well over their rated TDPs.

http://www.notebookcheck.com/Test-Toshiba-Satellite-Z30-B-100-Notebook.144873.0.html


http://www.notebookcheck.com/Test-Dell-XPS-13-2015-FHD-Notebook.139927.0.html

Also the discretanpcy between CB 11.5 and R15 scores make me think that there s a driver benchmark detection for the latter test..

It doesn't matter that OEMs are shunning Carrizo more than any other AMD processor in their recent history to the ADF see how bad this product is.

Blank statement without any value, not even an entertaining one...

Either provide arguments that support yours saying or else stop derailing AMD related threads, but you know that you are spreading fud since you pretend that you listened to Lisa Su conference, one more time she said that there s currently 35 designs...

Now can we have a tech discussion without those constant derailing..?.

 

[dark zero]

Despite looks decent Carrizo ends DOA. Intel trash tier, I mean, U Tier (I am calling that tier trash due their poor performance) is superior to Carrizo... When they were on Kaveri they had a chance, but now they don't have any chance anymore.

 

[jime1]

is the carrizo out in the market yet ?

 

[dark zero]

is the carrizo out in the market yet ?

Nope and Hardly we might see Carrizo since it's so bad.. Maybe Carrizo-L might enter on the market.

 

[Abwx]

Despite looks decent Carrizo ends DOA. Intel trash tier, I mean, U Tier (I am calling that tier trash due their poor performance) is superior to Carrizo... When they were on Kaveri they had a chance, but now they don't have any chance anymore.

At 28W, wich is well above Intel U series theorical TDPs, it s still competitive with Intel 28W CPUs, GPU wise it s a no contest even with Iris Pro as opponent...

Kaveri is well below Carrizo effiency wise at those TDPs, so why would this be a more compelling solution than Carrizo..??.

is the carrizo out in the market yet ?

Once W10 is released.

Nope and Hardly we might see Carrizo since it's so bad.. Maybe Carrizo-L might enter on the market.

Back your statement with numbers or facts, otherwise it s just fud, i mean all you are posting are blank statements, as if the purpose was to get those unsanstancied, and erroneous, claims to be posted and that s all...

 

[maarten12100]

People are turning this into a Anti AMD thread. They must be bleeding blue.

Figure 1. Actual measurements of CB 11.5 on a FX-8800P developers platform at set package power levels

i3 5010u with 920M consuming 31.3/34.7W
http://www.notebookcheck.net/Asus-F555LJ-GeForce-920M-Notebook-Review.144232.0.html

i5 5200u no dgpu consuming 28.7/31.5W
http://www.notebookcheck.net/Acer-Aspire-V3-371-Notebook-Review.135831.0.html

i5 5200u with 840M consuming 36/46.7W
http://www.notebookcheck.net/Acer-Aspire-E5-571G-Notebook-Review-Update.136078.0.html

i5 5250u no dgpu consuming 37.6/43.4
http://www.notebookcheck.net/Apple-MacBook-Air-11-Early-2015-Notebook-Review.141993.0.html

The list goes on and on those will all have equal scores at much higher power except for the 840M based one since that will beat it at higher power. At the same power I doubt they'll stand a chance. But again there are non available and we need some reviewers to confirm hypothesis.

Will things other than the Carrizo package consume the additional 20W so the power consumption of the entire notebook still matches Intel notebooks. Stay tuned to find out more about how this almost certainly won't be the case.

 

[mikk]

I doubt you can compare Notebookcheck numbers from the whole notebook with real power SoC tests from others.

You have to wait for Notebookcheck numbers from Carrizo. And even then such power adapter measurements might be inaccurate. Furthermore you need Retail notebooks from OEMs because power consumption can differ from Notebook to Notebook.

GPU wise it s a no contest even with Iris Pro as opponent...

You ask for facts but don't have much either. All you have is 3dmark which doesn't use much CPU. What matters are games and Carrizo on low TDP might suffer from CPU+GPU use. Not to mention that we don't have Skylake tests to compare.

 

[maarten12100]

I doubt you can compare Notebookcheck numbers from the whole notebook with real power SoC tests from others.

You have to wait for Notebookcheck numbers from Carrizo. And even then such power adapter measurements might be inaccurate. Furthermore you need Retail notebooks from OEMs because power consumption can differ from Notebook to Notebook.




You ask for facts but don't have much either. All you have is 3dmark which doesn't use much CPU. What matters are games and Carrizo on low TDP might suffer from CPU+GPU use. Not to mention that we don't have Skylake tests to compare.

Very much true a direct comparison isn't valid since the rest of the notebook is not accounted for. Still I doubt it will consume that much power. I'm thinking 20-25W for the entire system under full load for the FX-8800p. Whether parts of the chip are throttled is the real question. I have no doubt Intel stays bellow their 15W TDP target under load if just one of the 2 components is fully loaded. AKA the CPU can be loaded or the GPU can be loaded when there is a mixed load.

I was under the impression that NBC uses a software based method to determine power consumption.

 

[Abwx]

I doubt you can compare Notebookcheck numbers from the whole notebook with real power SoC tests from others.

You have to wait for Notebookcheck numbers from Carrizo. And even then such power adapter measurements might be inaccurate. Furthermore you need Retail notebooks from OEMs because power consumption can differ from Notebook to Notebook.

We have the numbers from SA Thevenin , that s much more accurate than what is provided by Notebookcheck.

You ask for facts but don't have much either. All you have is 3dmark which doesn't use much CPU. What matters are games and Carrizo on low TDP might suffer from CPU+GPU use. Not to mention that we don't have Skylake tests to compare.

Iris Pro is 10% faster on 3DMark but at a 47W TDP, CPU doesnt matter much since it doesnt need to work at full frequency to load the GPU, for the record it s 3GHz on a Kaveri 7850K.

At low TDP Carrizo will do better as Intel s GPU has quite poor perf/Watt and they dont have better CPU perf/Watt.


Besides, in power limited environment Carrizo CPU will have more throughput than a Core M with the same power dedicated to the CPU part, you seems to forget this detail by assuming that Intel has better per/watt for the CPU, wich is no more the case...

 

[monstercameron]

I'm not quite sure if glofo 28nm is comparable to 14nm finfets. Clock for clock Intel still wins cpu efficiency for sure.

 

[Abwx]

I was under the impression that NBC uses a software based method to determine power consumption.

They use a multimeter, a Voltcraft VC 940.

Sometimes they also use tools like Intel power gadget but they dont use those numbers for their power comsumption measurements.

 

[Abwx]

I'm not quite sure if glofo 28nm is comparable to 14nm finfets. Clock for clock Intel still wins cpu efficiency for sure.

What matters is the throughput at a given power, same frequency means nothing given the large difference in uarch, and in this respect Carrizo has about 20% better perf/Watt at 15W, Core M doesnt even manage to be on par with a Beema in this metric.

 

[monstercameron]

What matters is the throughput at a given power, same frequency means nothing given the large difference in uarch, and in this respect Carrizo has about 20% better perf/Watt at 15W, Core M doesnt even manage to be on par with a Beema in this metric.

I'm not how accurate this is because even atom is more performant at lower tpds, albeit not clock for clock. Maybe you would have an argument for the gpu alone but the CPUs, Intel is more efficient.

 

[Abwx]

I'm not how accurate this is because even atom is more performant at lower tpds, albeit not clock for clock. Maybe you would have an argument for the gpu alone but the CPUs, Intel is more efficient.

Indeed Baytrail is still more efficient than Core M but only up to 12-13W..

As for Intel efficency and at 15W Carrizo get 20% higher score in CB 11.5, so the number speak by itself, Intel CPU is less efficient at this TDP and up to 28W at wich point efficencies are comparable.

Now if you can provide numbers or estimations that say otherwise i would reconsider my opinion but so far i wasnt contradicted anywhere, my estimation of Intel s 14nm perf/Watt improvement is still unchallenged and the only one that tried to do so and who is knowledgeable (and often works for Intel according to his sayings) has long abandonned this intenable position after he scrutinized the numbers more closely.