• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Carrizo APU Massive Leak – 40W TDP, Excavator Cores, 512 SP and Assorted Benchmarks

Davedree said:
I am most certainly not ignoring the architectural improvments considering Carrizo using rescaling and high dense library= means they have more transistors in the same footprint.
More transistors (3.1 billion vs 2.4billion means you get more done but in Carrizo this is for a 15w-35w tdp envelope.


"the bandwidth limitation will be exactly the same as kaveri"
Lol yeah and the end of the day Kaveri and Carrizo are both limited by shared ddr3 bandwidth, Something both you and Boom ignored in the Kaveri thread and something I had to repeat, I even predicted the igp performance of the kaveri igp and was proven right. Just like in this thread there was a link to x2 the performance based on sisoftware benches, which you said would translate into performance, which it hasn't

edit and it all is based on acceptable 1080p performance.
Click to expand...

you still haven't given me any reference for your rhetoric, your just repeating it.

Where are the iGPU clock rates of Carrizo?
where are the Carrizo vs kaveri memory performance figures?
where is this wide selection of Carrizo gaming performance?
 
humbug said:
you still haven't given me any reference for your rhetoric, your just repeating it.

Where are the iGPU clock rates of Carrizo?
where are the Carrizo vs kaveri memory performance figures?
where is this wide selection of Carrizo gaming performance?
Click to expand...

It's hard when I don't own the hardware Humbug but from the slides it's clear Carrizo is mobile only. 28shp isn't very good for clockspeeds, but it has the advantage of density compared to intel 22nm and tsmc 28nm.
HDL is used in Carrizo, it's putting more metal layers/transistors 3.1 billion (i think it's 8-9 layers) inc the southbridge into in a die area of 244mm2. Kaveri is 2.4 billion at 245mm2. The difference is these transistors (mostly for the gpu) are denser and tightly packed and so I predict the power consumption and thermal limits will constrain the top end headroom of overclocking. It doesn't mean it won't scale but it will require 65w-95w and then are the performance gains really worth it, this is my point (even if you increase the tdp to 95w for the desktop is it enough gpu performance to provide a substantial gain over Kaveri at 1080p or even intel pro 6200 when the weak point is shared ddr3? (We don't know but we will try and find out).

The excavator cores are 23% smaller than steamroller cores (are these refined for efficiency and less for performance?) . Also Carrizo features an advanced AVFS which is continually adjusting for the tdp via clocks speeds and voltages so this would need to be looked at for overclocking.

L1 cache is doubled to 64k per module, but l2 cache is reduced to 1mb per module. But the traces are smaller so logic and latency overall should be improved.

The same can be thrown back to you, Why do you think Carrizo can scale in clockspeed and provide a substantial performance increase which will be able to provide the 1080p experience from an igp that we have been waiting many years for? Why are you so against the fact the Carrizo is purely for mobile and that i'm out of my depth, when neither of us can prove it at the moment.

Let's look at it logically, why didn't Steamroller make it onto am3+ 32nm or am3+ 28nm ?

Ok why didn't excavator make it onto fm2+ 28nm. Why have we got Godaveri which is a p-state refined kaveri.
( I'm fairly confident why)

Kaveri bios and p-states in early models were affected badly by tdp limits.
For a 7850k Max clocks on all cpu cores on desktop Kaveri is 3ghz when the 720 mhz gpu is fully utilized.
There is a way around it but reviews and benchmarks will normally show throttled results. For example 7850k usually quoted as 2100 when it's potential is around 3000+

Where are the iGPU clock rates of Carrizo?
carrizo clock speeds = 720mhz to 800mhz

where are the Carrizo vs kaveri memory performance figures?
give it time.

where is this wide selection of Carrizo gaming performance?
give it time.

Whilst the thermal limits was my own theory, this link provides insight into carrizo and states the clock speeds, but I can't confirm if they are true.

http://arstechnica.co.uk/informatio...sixth-time-lucky-amd-details-the-carrizo-apu/

edit i edited the shared ddr3 bit
 
Last edited:
NBC tested the A10-8700P:

http://www.notebookcheck.net/HP-Pavilion-17-g054ng-Notebook-Review.147794.0.html

Compared to Kaveri it looks a decent improvement,even if the laptop is somewhat bugged and only runs the RAM at 1333MHZ.

It seems AMD is using a configurable TDP of between 15W to 35W with Carrizo and many OEMs are choosing the former.

Plus some benchmarks from The Stilt:

http://www.overclock.net/t/1560230/jagatreview-hands-on-amd-fx-8800p-carrizo/360#post_24279106
http://www.overclock.net/t/1560230/jagatreview-hands-on-amd-fx-8800p-carrizo/380#post_24282962

Some notes on the power management:

http://www.overclock.net/t/1560230/jagatreview-hands-on-amd-fx-8800p-carrizo/410#post_24287228
 
Last edited:
CAT-THE-FIFTH said:
Its a shame TBH - it really does look like Carrizo is a big upgrade over Kaveri.

Edit!!

Having said that I did a Google search and there are a surprising number of A10 8700P laptops now available.
Click to expand...

I'm not aware of a single one that's not gimped though. Think they are actually slower than some of the CPUs they replaced in asthmatic mode.
 
One of these or the R series apu's I reckon combined with DDR4 would give AMD something that could compete with Intel's NUC pc's.
 
CAT-THE-FIFTH said:
Carrizo based A12-8800B tested:

http://www.notebookcheck.com/Test-HP-EliteBook-745-G3-Notebook.157955.0.html

In CB the A12-8800B is around 20% slower than a Skylake Core i3 6100U but the difference is much lower for X264. Its also around 30% faster for IGP performance too and around the same as a Kepler based GT920M discrete card which has 384 shaders.
Click to expand...

Get me one in either an 11.6" or 12-13.6" ultrabook and I'll buy one.
Till then, it's still a product AMD don't give me an option to buy :p
 
jigger said:
At 95 watts it could have 16/32 no problem. I want one like that.
Click to expand...

hmmm.... dunno, Intel's 22nm 8 Core 16 Threader is 140 Watts.

I know these are 14nm but i still think 16 cores 32 Threads at a reasonable clock rate would be 160 Watts. those 125 watt FX-8### get hot overclocked.
 
Last edited:
My 4.5Ghz Q6600 was getting close to 270 IIRC. An overclcoked X99 chip is probably not far off that.

A 200 watt Zen chip would be nothing. If it's fast enough give me 300 watts. Every watt used is a watt faster!
 
Last edited:
I want to know what speeds of memory it will support. I'm guessing it wont support Quad-channel which is a shame (emphasize guess, mind). Also, how many PCI-E 3.0 lanes can I get out of it. This is key for me.
 
You must log in or register to reply here.
Back
Top Bottom