The Kaveri Pre-Launch Thread (A10-7800 and A10-6800k @3,5 Ghz)

[Sweepr]

$173, thats ~22% more expensive than the fastest Richland and definitely on par with entry-level quad-core Haswells (3.0-3.2GHz).

 

[krumme]

"CrossFire XDMA" block in there. Looks like we might get Dual Graphics that actually works for once!

Yes. First time ever. Lol. In a notebook add about 512 extra shaders and mantle and you got a dirt cheap bf4 machine playing fine.

It all depends on price then.

 

[kapulek]

$173, thats ~22% more expensive than the fastest Richland and definitely on par with entry-level quad-core Haswells (3.0-3.2GHz).

$173 with BF4.

 

[SPBHM]

Yes. First time ever. Lol. In a notebook add about 512 extra shaders and mantle and you got a dirt cheap bf4 machine playing fine.

It all depends on price then.

price and availability, it was the main problem with llano, trinity and richland laptops I think

if mantle delivers enough gain it could make BF4 viable without discrete graphics for a less expensive laptop.

$173 is way to much even with a free game (probably easy to get for something lower than $60 right now)

 

[inf64]

Well the game costs ~55 bucks or so. Subtract that from total price and you get a nice deal for APU+Bf4.

 

[CHADBOGA]

Well the game costs ~55 bucks or so. Subtract that from total price and you get a nice deal for APU+Bf4.

Providing you want to play that game and then only play it at middling levels of frame rates.

That's probably not the biggest of niches.

 

[mikk]

Besides, Broadwell is to be mobile-only, and will have no effect on the desktop sector where Kaveri will also, along with mobile, be competing it.

Broadwell isn't mobile only. Xeon E3 and Broadwell-K.

 

[Fjodor2001]

I'm interested in what low power Notebook Kaveri APU models they'll have. Has any such info been leaked / presented yet? I think I've only seen info on the desktop models .

Also, is it known whether the 28 nm GloFo process technology used for Kaveri is good for low power chips?

 

[inf64]

Or you can sell the key and practically pay 55 bucks less for 7850?

 

[Homeles]

That is one weird looking die... CPU cores swimming in a pool of GPU. The GPU looks like a massive cache... terrible image, really. I really wish they'd strip these things down to the silicon when taking publishing shots of them. Can't wait until better die shots start floating around.

Also, from what I can see, the Steamroller cores in Kaveri aren't the ones from that leaked die shot from the summer. As was concluded back then, that mystery die image is definitely Excavator.

 

[monstercameron]

http://imgur.com/a/Yzo8L
in case you missed it.

 

[Homeles]

Kaveri definitely looks like a great product, however I'm not convinced that their position in 2014 vs. Intel will be better than it was in 2013. Time will tell, of course.

Really, both companies are going to have some fine products on their hands this year. Progress is good... and AMD actually looks like they might be making some for once. I'll bet AMD is eagerly awaiting to use a process that isn't ion-doped garbage. The Bulldozer family should look very nice once it's ported over to 20nm.

 

[Asterox]

http://imgur.com/a/Yzo8L
in case you missed it.

How much i see no comment on this, but there should certainly be some good comments "because there is question of tiny 600mhz higher defoult CPU frequency+Higher IPC under 45W TDP".

- Richland APU A10-6700T, 2.5ghz/3.5ghz Turbo Core/384 VLIW4 Radeon Cores, 45W TDP

http://www.cpu-world.com/CPUs/Bulldozer/AMD-A10-Series%20A10-6700T.html


- Kaveri APU A8-7600, 3.1ghz/3.3ghz Turbo Core/384 GCN Radeon Cores, 45W TDP

 

[Centauri]

I find it interesting that the 7850K has the same GPU clock as the 7600.

I hope it isn't an indication that there isn't much headroom.

 

[Homeles]

Sorry to rain on your parade, but it's impossible to make a meaningful evaluation of a processor based on clock speed alone. This is particularly true with mobile products, where thermal and power constraints can result in ICs running under their max default clocks.

I'm actually highly suspicious of those speeds. 28nm planar is notably worse than 32nm PDSOI.

 

[Blitzvogel]

Kaveri looks exciting, but reviews will tell the real story. The BF4 bundle was a really good idea on whoever's part. Should help to cement Mantle as a truly viable addition to the PC gaming world.

 

[Abwx]

Sorry to rain on your parade, but it's impossible to make a meaningful evaluation of a processor based on clock speed alone. This is particularly true with mobile products, where thermal and power constraints can result in ICs running under their max default clocks.

I'm actually highly suspicious of those speeds. 28nm planar is notably worse than 32nm PDSOI.

Slides say SHP , same naming scheme as the previous process
wich was SOI based , density seems higher than Intel s 22nm....

 

[Homeles]

AMD explicitly said they would not go with SOI on 28nm. And it's great and all that they're denser than Intel, because they get slaughtered on absolutely every other metric.

 

[Abwx]

AMD explicitly said they would not go with SOI on 28nm. And it's great and all that they're denser than Intel, because they get slaughtered on absolutely every other metric.

Particularly on the GPU side of things.

 

[Centauri]

Sorry to rain on your parade

Might wanna learn the meaning of that saying before you try using it again.

 

[Spawne32]

Remember guys, they are touting this APU around BF4, which is a very CPU dependent game. I've heard nothing but horror storys about AMD vs Intel in BF4 as far as how it performs with the lacking single threaded performance on the FX series. So for them to use BF4 as the basis of the sale rather then some other game that would obviously perform better with dual graphics APU + dedicated GPU, shows that they are pretty confident about the improvements made. If it performs just about as well as i3-i5 haswell, and is priced lower, then it makes it a formidable desktop processor for the average user.

 

[Homeles]

Might wanna learn the meaning of that saying before you try using it again.

You first. Also, my comment wasn't directed at you.

Particularly on the GPU side of things.

Whoops, looks like you got lost. I was very clearly talking about transistor performance.

Remember guys, they are touting this APU around BF4, which is a very CPU dependent game. I've heard nothing but horror storys about AMD vs Intel in BF4 as far as how it performs with the lacking single threaded performance on the FX series. So for them to use BF4 as the basis of the sale rather then some other game that would obviously perform better with dual graphics APU + dedicated GPU, shows that they are pretty confident about the improvements made. If it performs just about as well as i3-i5 haswell, and is priced lower, then it makes it a formidable desktop processor for the average user.

I see it more as them touting BF4 because of several reasons. One, it's the AAA game that "everybody wants," and they've got a strong partnership with DICE/EA. BF4 is the Mantle poster child.

 

[Xpage]

I really wish they had FD SOI on 28nm, i think the savings in power would really help these parts and would have allowed kabini to maybe exist on this 28nm with better performance than TMSC.

Any news on FD SOI on the 20nm process? I would hope they use it to try to compete vs 14nm Fin-Fets

 

[NostaSeronx]

Are you guys sure that 28SHP is not 28FD? http://i.imgur.com/u6fbE3d.png

It could be Bulk for the GPU, PHYs, etc and FD-SOI for the CPUs, SRAM, etc. Explaining the whole it is perfect for the APU.

SHP is SOI (custom(A custom SOI implementation)) moniker.

LPH is built for high-end SoCs but it is from Samsung.
HPP is built for best perf/watt but it

28 HPP should be ~1.8 Billion
28 LPH should be ~1.9 Billion
28 FDSOI should be ~2.2 Billion
^-- Transistors in 245 mm², VLIW4 is less dense than GCN so the number should be off a bit.

28 HPP (0.5 Node from 32nm) -> 28 LPH (0.55 Node from 32nm) -> 28 FD (0.5 Node above Bulk 20nm)

 

[DaZeeMan]

I really wish they had FD SOI on 28nm, i think the savings in power would really help these parts and would have allowed kabini to maybe exist on this 28nm with better performance than TMSC.

Any news on FD SOI on the 20nm process? I would hope they use it to try to compete vs 14nm Fin-Fets

According to this article, AMD will be taping out 20nm and 14nm FinFETs in the first two quarters of 2014, with Global Foundries taking their 14nm XM finFET process into mass production sometime in late 2014.

http://www.xbitlabs.com/news/other/...nm_FinFET_Chips_Within_Next_Two_Quarters.html

The article notes that we probably won't see any products from this until late 2014/more likely 2015.

Still about two steps behind Intel, but it's progress at least.