AMD's Carrizo APU With Excavator Cores Significantly Improves Efficiency

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.

marcelospera

Honorable
Sep 24, 2013
11
0
10,510
Yes! I really respect AMD for all work and efforts they do. I want to see AMD strong again and with great products. I believe Carrizo is one of them. Bring it to the desktop world!
 

irish_adam

Distinguished
Mar 30, 2010
236
64
18,760


You have to remember this is not a massive overhaul of the architecture so 5% increase in IPC isnt that bad. They are stuck with modifying their old cores until Zen comes out next year, there is nothing that can do about that. You have to remember that AMD is a tiny company compared to Intel, the fact that they even come close performance compared to Intel (at the lower end) is still a massive achievement for them.

Also K12 is the ARM core, Zen is their next x86-64 core. They'll survive till then but they wont survive if it doesnt contend with Intel at least in the mid range. With the shift to 14nm (ish) process which by the time zen comes out (mid-2016) will be quite mature so they should be able to bring the power consumption down and they'll have worked hard on the IPC so we should see a big bump in that too. Rumour has it that the first desktop Zen APU will use about 95W for the 8 core version which sounds quite promising as it will have quite a large GPU on it that will draw a lot of that power. It will be interesting if and when they release a non APU version of zen as if its powerful enough to take on an i5 with the graphics on the die they might not bother, or at least to start with anyway just like you cant get an Intel CPU without a GPU unless you cough up stupid money.
 

airborn824

Honorable
Mar 3, 2013
226
0
10,690
Impressed. Very good direction AMD, keep it up. Jim Keller can already be seen in this old tech. Imagine what his influence will do with all the 2016 tech.. DROOOOL!! In Keller we trust.
 

MyDocuments

Reputable
Jun 21, 2014
39
0
4,530
Will the AMD APUs ever be competitive on performance with Intel without the L3 cache... is it really too hard to add the L3 cache into the HSA system... will the extra silicon and transistors for a L3 cache blow the power efficiency budget... in which markets does AMD now hope to challenge Intel (if ever); Mobile, Desktop, Server... too many questions and so little time, c'mon AMD you can do it!
AMD may now be the only answer to getting the Intel x86 chip prices lower (unless the ex-Cyrix/VIA licensee can pull something magical out of the hat, unlikely I guess).
 

xenol

Distinguished
Jun 18, 2008
216
0
18,680
I want to see this in action against the Atom Z3000 series. Might actually be worthwhile to play something on a tablet!
 

InvalidError

Titan
Moderator

There isn't much AMD can do about that since all the fabs-for-hire are just barely starting to produce sub-20nm chips now. Just be glad for AMD's sake that Intel's Broadwell ended up delayed by a whole year due to "complications."
 

Reepca

Honorable
Dec 5, 2012
156
0
10,680
No impressed, 5% IPC? give me a break! This is just more AMD marketing hype and not much more. They already moved the south bridge onto the chip for the small cores last year (which they are now calling carrizo-L or some such nonsense) so doing this for the big core while may save a few bucks for the manufacturers is nothing to crow about.

I don't see this as being competitive unless AMD really drops the price so that their top end chip undercuts the Intel i3. Hopefully AMD can stay in business long enough to get their new K12 chips out there on 14nm process or you can stick a fork in AMD as they well be toast at that time. May be toast anyway looking at their level of debt, any pop up interest rates will kill them.

5% increased IPC isn't that bad. Especially when you consider it's going to get a performance increase beyond that anyway.

So you have C-IPC = 1.05K-IPC. However, note "40% less power consumption". But if we turn that lower power consumption into higher clocks, so that we have the same power consumption but higher frequencies, we get more overall performance. The exact amount is something I'm not quite willing to compute just now, since I'm unfamiliar with what voltage is necessary for what frequency, how power scales with voltage, etc - all I know is that it isn't linear, so it would be a bit tricky to find out. The point remains, though, that we should be getting, for the same power consumption, not only increased IPC but also increased clocks. The net improvement is the product of these two (%ipc * %clocks = %net).

Also: "a more GPU-oriented configuration, which is what Carrizo has".
 
28 nm? we're in 2015!

AMD Kaveri APU transistor density at GF 28nm SHP is higher than Intel Haswell (4C/GT2) 22nm transistor density. As the article notes, a 23% reduction in die-size on the same 28nm-node by AMD for the Excavator "High Density (Thin) Libraries" is likely competitive with Intel transistor density at 14nm.

 


Lol, the names that intel and AMD come up with (Devils canyon, Godavari) are so goofy sometimes.

I am not too sure about AMD but Intel typically names their CPU code names after locations. Devils Canyon is located in California. Nehalem is a city in Oregon (while is where that uArch was deigned) and Haswell is a city in Colorado.
 
28 nm? we're in 2015!
-------------------------------
Remember, the Samsung new 3d NAND is a step backwards from the 15-20nm parts to a full 40nm process. GPU's are on 20-30nm still, and nVidia's Maxwell chips are amazingly efficient. Going smaller isn't always required and sometimes there's a lot more to a chips efficiency than just sheer process feature size.

Boy, I never thought I'd say 'size doesn't matter' Lol.

Increasingly though, these chips have to first and foremost be -mobile friendly- as that is the trend with laptops/convertibles/tablets. In that, they might just succeed admirably as their GPU's still have Intels IGP as their official whipping boy. So - they have a decided advantage as the screen size tends to increase in resolution. Let's hope they can recapture both some mind-share in the consumers and some market-share along with it in sales.

A strong AMD and Intel benefits -us- the consumer.
 

andre_888

Distinguished
Mar 23, 2010
33
0
18,540
The names that Intel and AMD choose for their products is quite smart : it makes consumers curious about geography, the people who inhabit them and how the world is changing.

Names like Pentium, Athlon etc. are nothing more than attempt to market a technology sounding item.
 

hyper_cool

Distinguished
Dec 31, 2007
4
0
18,510
It looks very promising and probably a good CPU. But sad part is it will never make it to nice Laptop builds from OEM companies. It will make it in to crappy laptop models, of course courtesy by Intel. Intel will not stay put. The intel ceo is probably making few phone calls already asking OEM not to use AMD again on high models except on crappy ones. Well, AMD you still have my business. I'm happy with my laptop running A10 Dual graphics too :)
 

hyper_cool

Distinguished
Dec 31, 2007
4
0
18,510
It looks very promising and probably a good CPU. But sad part is it will never make it to nice Laptop builds from OEM companies. It will make it in to crappy laptop models, of course courtesy by Intel. Intel will not stay put. The intel ceo is probably making few phone calls already asking OEM not to use AMD again on high models except on crappy ones. Well, AMD you still have my business. I'm happy with my laptop running A10 Dual graphics too :)
 

bit_user

Titan
Ambassador
At the recent AMD AMA (Ask Me "Anything") about FreeSync, someone asked how they picked their CPU code names. The AMD rep said the engineers (?) nominate and vote on it.

But they should hold a public contest. And perhaps not just for internal code names, but perhaps the official branding. They might get some good names out of it, plus some free publicity. Self-important marketing types probably hate this idea.
 

Nakul Kale

Reputable
Feb 25, 2015
1
0
4,510


Lol, the names that intel and AMD come up with (Devils canyon, Godavari) are so goofy sometimes.


Lol, the names that intel and AMD come up with (Devils canyon, Godavari) are so goofy sometimes.


Lol, the names that intel and AMD come up with (Devils canyon, Godavari) are so goofy sometimes.


Lol, the names that intel and AMD come up with (Devils canyon, Godavari) are so goofy sometimes.


Lol, the names that intel and AMD come up with (Devils canyon, Godavari) are so goofy sometimes.


Lol, the names that intel and AMD come up with (Devils canyon, Godavari) are so goofy sometimes.

Actually these names are real names referring to mountain place somewhere in Us, and kaveri godavari are some of the river names
 


Yes, thats it. Intel and AMD are secretly trying to spur interest in geography by naming their chips after places.
 


Thats more true than we might care to admit. AMD despite having issues on the desktop, has been VERY competitive on the laptop parts. But that success has not been reflected in market share.
 

ykki

Honorable


I wonder why this happens. Maybe bad marketing from AMD side or intel has monopolized the market to a saturation point.

 

childofthekorn

Honorable
Jan 31, 2013
359
0
10,780


FX-8800p
 
Dang it :pt1cable: I had some 'TL;DR' info for you guys, but lost it when I re-booted. Basically:

Carrizo = Excavator Cores + GCN '1.2' CUs
Carrizo-L = Puma+ Cores + GCN '1.2' + integrated FCH

GCN 1.2 (not an AMD term) is the latest Radeon core iteration (as in 'Tonga' -- the R9 285)

Puma+ is the an enhanced 'Cat Core' - Bobcat--> Jaguar--> Puma

From the article:
...HSA support has already been completely implemented ...

I do not believe this is a correct statement. The Kaveri APU contained certain HSA features, The Carrizo APU will be the first fully-compliant HSA 1.0 processor.

 
Status
Not open for further replies.