AMD Carrizo Pre-release thread

[coercitiv]

Celeron G1820 with ECS B85 motherboard is less than $4 more than Athlon 5350 with ASRock AM1 motherboard.

It is also much faster (and can be upgraded to use an i7). This, in contrast, to the AM1 where the Athlon 5350 is actually the top SKU.

So Ivy Bridge Celeron is a low cost product, but Kabini Athlon is not. Is that correct?

 

[cbn]

So Ivy Bridge Celeron is a low cost product, but Kabini Athlon is not. Is that correct?

Its a Haswell Celeron.

And the processors are in the same price ballpark.

The point is that Intel offers much better performance per dollar....and benefits from vastly better upgrade potential.

 

[coercitiv]

The point is that Intel offers much better performance per dollar....and benefits from vastly better upgrade potential.

We are not talking about who has the better performance per dollar, we are discussing what constitutes a "low cost platform", since you clearly disagree AM1 fits that description.

 

[cbn]

We are not talking about who has the better performance per dollar, we are discussing what constitutes a "low cost platform", since you clearly disagree AM1 fits that description.

Low cost for accomplishing what computing purpose?

I would say that AM1 is high cost for accomplishing what it does.

Having AM1 bundled to Windows with bing OS would lower this cost.

 

[ShintaiDK]

Low cost isnt low cost if people dump it at the landfill/recycling yard as quickly as they bought it.

 

[gorion]

I think, with Braswell's lower CPU base clocks, at least for Desktop usage, AMD should go for the "kill" and release Beema AM1. Maybe even increase the frequency at the top end (if they can), at the expense of TDP (instead of 15W, would still have 25W TDP).

http://www.cpubenchmark.net/cpu.php?cpu=AMD+A8-6410+APU

I believe an AM1 version of an A8-6410 would be almost competitive, performance-wise, with a Q6600. At 15W. How is that not a winner?

A6-6410 is already competitive quite better than J1900 at 15w:
http://www.cpubenchmark.net/compare.php?cmp[]=2290&cmp[]=2131
Braswell is worse CPU wise. I wonder if sales in this segment are so low that for AMD it's not worth the hassle.

 

[dark zero]

Actually the A8 6410 is toe to toe against the Core i3 4300U

 

[Shivansps]

A4-4000 has a higher clocked GPU than Kabini. Increasing GPU clocks is ok as long as it brings some gains. Adding a secondary memory channel is clearly out of the question for the purpose of this product. Also, I'd rather see a proper test of memory scaling on Kabini, than comparing memory scaling on systems with different CPUs.


Less TDP is also correct.

See chart above. Also... chill?! It's just a small core / low cost platform.

Not sure WTH you are talking about, mobile Kabini has been mostly 15W parts, also if desktop Beema is 15W with active fan is just pointless.
Most of the deal breaking features of AM1 are socket limited, the SC mem, the 4x pci-e and the 2 sata max...

Also lets check a common A4-4000 result with SC memory.
http://www.3dmark.com/is/2276489

How a common result with dual channel memory
http://www.3dmark.com/is/1848385

Now the 5350
http://www.3dmark.com/is/2657525

SO... i dont see much a diference in those extra 120mhz that the A4-4000 has...

Nobody cares about better ram speed, even a 1333 in dual channel is way better than anything you can place in single channel.

 

[AtenRa]

Also lets check a common A4-4000 result with SC memory.
http://www.3dmark.com/is/2276489

How a common result with dual channel memory
http://www.3dmark.com/is/1848385

Now the 5350
http://www.3dmark.com/is/2657525

SO... i dont see much a diference in those extra 120mhz that the A4-4000 has...

Nobody cares about better ram speed, even a 1333 in dual channel is way better than anything you can place in single channel.

A4-4000 iGPU is VLIW4 when Kabini iGPU is GCN, not to mention one is 65W TDP and the other only 25W TDP.

 

[Shivansps]

A4-4000 iGPU is VLIW4 when Kabini iGPU is GCN, not to mention one is 65W TDP and the other only 25W TDP.

and thats relevant because? nobody cares, we are talking about better igp clocks will not make it better, everyone can see that on the A4-4000.

 

[AtenRa]

and thats relevant because? nobody cares, we are talking about better igp clocks will not make it better, everyone can see that on the A4-4000.

There are games that will gain from higher clocks with the Kabini, that is why comparing it to a different architecture like the A4-4000 is meaningless.

 

[coercitiv]

Not sure WTH you are talking about, mobile Kabini has been mostly 15W parts

A6-5200, the 2Ghz mobile Kabini was 25W TDP. Since it's the one with the highest clocks for Kabini, the same parameter you were asking to be increased earlier, I would argue it's important in this conversation.

also if desktop Beema is 15W with active fan is just pointless.

I can understand why Bay-Trail D had a clear advantage with power usage under load and more flexibility due to passive cooling, but to call a desktop CPU pointless because it uses active cooling is far fetched.

 

[monstercameron]

OK can we get this thread back in order, make another thread about everything else please.

 

[richaron]

OK can we get this thread back in order, make another thread about everything else please.

Agreed. I'm still interested in learning about available products and seeing some benchies. Pls try not to feed the regular trolls.

Anyone got new info?

Low cost isnt low cost if people dump it at the landfill/recycling yard as quickly as they bought it.

...Good point.

 

[coercitiv]

OK can we get this thread back in order, make another thread about everything else please.

Well, I created a separate thread for the iGPU scaling debate.

Apologies for the rest of the OT.

 

[Abwx]

Cinebench 11.5 scores...

https://semiaccurate.com/forums/showpost.php?p=240581&postcount=911

Roughly 20% better than Broadwell at 15W....

 

[ShintaiDK]

Cinebench 11.5 scores...

https://semiaccurate.com/forums/showpost.php?p=240581&postcount=911

Roughly 20% better than Broadwell at 15W....

No its not, and you know it. And thats the A12-8800B.

 

[Abwx]

No its not, and you know it. And thats the A12-8800B.

That s a business dedicated FX8800P actualy, besides could you tell us why it wouldnt be better than said BDW..?..

You know that BDW score is barely 2 at 15W..??..

 

[ShintaiDK]

That s a business dedicated FX8800P actualy, besides could you tell us why it wouldnt be better than said BDW..?..

You know that BDW score is barely 2 at 15W..??..

Are you trying to compare the lowest bin to the top bin?

 

[Abwx]

Are you trying to compare the lowest bin to the top bin?

No, the comparison is Carrizo at 15W with BDW at 15W, as power is increased Carrizo advantage shrink and at 28-30W the gap is completely closed with Intel taking the advantage as power is even more increased.


There s no top bin actualy, that s just power managements that allow a "top binned" CPU to work at higher TDP for a short time, but if a hard limitation at 15W is implemented there will be no difference between CPUs of a same line, if it wasnt for the GPUs there would be no point in buying the most expensive parts...

 

[ShintaiDK]

Why do you avoid the SKU part? A 2Ghz 5005 is a 15W part. But so is the 5650 at 3.2Ghz.

 

[Abwx]

Why do you avoid the SKU part? A 2Ghz 5005 is a 15W part. But so is the 5650 at 3.2Ghz.

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.

It s not by chance that there is only 3 Carrizo SKUs, if you want more perfs just increase its TDP, exactly what Intel is doing but through a segmented offering allowed by much bigger volumes...

 

[ShintaiDK]

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

 

[Abwx]

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

You call nonsense all things that you dont understand...??.

That s not an academic forum but if you want a prove that CPUs TDP scale, at best, as a square of frequency you have all the theorical explanations here :

https://en.wikipedia.org/wiki/MOSFET

And the relevant formulae, excluding some second order influence, is this one :

To up frequency by 1.6 require increasing the current (Id) by the same ratio.

To get this the supply voltage (VDS) must be increased by sqrt(1.6).

The rest you can deduct it by yourself but if ever you have difficulties here what follow for our exemple:


Frequency delta x voltage delta = 1.6 x sqrt(1.6)^2 = 1.6 x 1.6 = 2.56....

Of course this can be easily deducted when comparing the voltages of a same CPU at different frequencies, two CPUZ pic would be enough actualy, you want me to provide some...??.

 

[ShintaiDK]

You are trying to extrapolate a BWs usage from the lowest bin to the top bin. Using only numbers you got from Carrizo and its node.

Not to mention: