AMD - Piledriver chitchat

[HalloweenJack]

I saw this:

http://forums.anandtech.com/showpost...8&postcount=83

IPC of Kabini seems to be quite decent.

and typical reaction from the intel support forum , oops I mean anandtechs impartial user base....

[CAT-THE-FIFTH]

It seems AMD has made less of a loss this quarter and has increased margins too:

http://www.marketwire.com/press-rele...md-1780582.htm

Revenue fell by 6% over the previous quarter.

Funnily,Intel revenue fell a bit more than AMD:

http://www.techpowerup.com/182863/Intel-Reports-First-Quarter-Revenue-of-$12.6-Billion.html

Revenue fell by 6.7% over the previous quarter.

The PC market fell by 14% this quarter.

Edit!!

A better overview of the Jaguar results I mentioned in post 1280:

http://www.planet3dnow.de/cgi-bin/ne...?id=1366210905

[watercooled]

Wow, that's beating Core2 in some areas, and in the ballpark of even more modern 'big' architectures!

I wonder how well it clocks down? If Atom can reach mobile SoC ballpark at 32nm, this could make things interesting in the mobile space.

[kalniel]

and typical reaction from the intel support forum , oops I mean anandtechs impartial user base....

Not all of us over there are like that, thank you... (or do you lump me in as well because I happen to have gone Intel this round? - you would have as well if you had to chose between nehalem and phenom and you had need of high IO bandwidth )

[scaryjim]

Edit!!

A better overview of the Jaguar results I mentioned in post 1280:

http://www.planet3dnow.de/cgi-bin/ne...?id=1366210905

I'm concerned by the Xeon E3 1220L Whetstone numbers in those charts. It's basically the same chip as an i3 2100T, but should have an advantage in single-threaded tests due to turbo boost, and be roughly level in multi-threaded tests. In Dhrystone that holds out, but in whetstone the i3 monsters it. In the multithreaded tests Kabini is about half the performance of a 2.5GHz i3, which would make sense given its positioning in the thin/light laptop market against 1.3GHz i3 derivatives - if all this translates to real-world performance it could genuinely give ultrabooks a run for their money....

There's a lot of assumptions in there, but it's looking like a good chip: IPC is similar to Core 2, so you're getting a 2GHz Core 2 Quad plus discrete-class Radeon graphics (based on GCN, iirc) in an ultrabook thermal envelope. Finally, something to get excited about in the mobile world

[CAT-THE-FIFTH]

Not all of us over there are like that, thank you... (or do you lump me in as well because I happen to have gone Intel this round? - you would have as well if you had to chose between nehalem and phenom and you had need of high IO bandwidth )

I am over there as well,and I consider it to be a poorly moderated tech forum ATM. The whole approach to the "AMD sucks thread" and the response threads to it was bad enough.

There are some good members over there and some good threads,but you only have to look at some members and how they are allowed to troll away.

Its basically Intel will destroy all competition and everyone else,ie,AMD,ARM,Qualcomm,Samsung,etc is doomed. It really is pathetic at times,and verges on the same level as AMDZone(for AMD) IMHO.

[watercooled]

I know we've been here time and time again, but I fail to understand why anyone with a functioning brain cell would wish for a monopoly in any area, let alone one run by Intel...

[scaryjim]

I know we've been here time and time again, but I fail to understand why anyone with a functioning brain cell would wish for a monopoly in any area, let alone one run by Intel...

Basic misunderstanding of how the business world works? Presumably they think that once all their competitors fail, Intel will be overwhelmed by their world-dominating position and become a philanthropist-led organisation that will sell all its subsequent processors at cost because its work will have been done...?

Just reviewed those benchmark results and realised that the single thread Whetstone score for Kabini is < 20% higher than for Ontario, despite it being clocked twice as fast (the same goes for multithreaded if you factor in having twice as many cores). That's a bit worrying for Kabini flops (unless it's just a case of Ontario punching way over its weight)...

In fact, the results are a bit all over the place. The Core 2 6600 - which I now realise must be an E6600 based on the difference between single and multithreaded results - is more than twice as fast as a T7300 at the single core benches - despite both being Core 2. In fact, the only differences are clock speed (2.4GHz v 2.0GHz) and FSB (1066MHz v 800MHz). That shouldn't account for 100% difference in benchmark scores...

[DanceswithUnix]

That looks to be a QA rig, so all they will care about is if the performance on a platform degrades. Performance of one platform relative to another won't matter, so no attempt would be made to keep thinfs the same.

For starters, one of the few things visible there is a Linux kernel version, and they are wildly different between platforms. If the kernels are that different, chances are the platforms are using wildly different versions of gcc as well.

Edit to add: As for Intel, being a monopoly is what got IBM where it is today

[watercooled]

Yeah, looking again, they do look a bit all over the place. There seem to be a fair few anomalous results and a lack of sanity-checking.

Edit: QA rig?

Edit2: Ah quality assurance? Yeah I suppose it could make sense with software changes.

[CAT-THE-FIFTH]

Look at the Linux version used,ie,3.8 for the Jaguar based setup.

Compare it to the core i7 950 based setup used which runs the same version.

The leaked Jaguar scores in CB indicated IPC with all cores enabled was very similar to a SB Core i3.

[scaryjim]

Look at the Linux version used,ie,3.8 for the Jaguar based setup.

Compare it to the core i7 950 based setup used which runs the same version.

The leaked Jaguar scores in CB indicated IPC with all cores enabled was very similar to a SB Core i3.

It's not just about linux version used those, it's about compiler optimisations and all sorts of other variables. Those scores simply aren't reliable - after all, the i3 2100T and E3 1220L use the same linux version and the Xeon has half the whetstone performance of the i3. Given they're hewn from identical silicon, that just isn't likely. They give a hint to the overall picture, but they can't be taken as genuine performance comparisons. I could come up with half a dozen scenarios to explain some of the results, but I'd just be guessing.

What it does say, is that Kabini is likely to be particularly good with well-threaded software. 4 genuine cores stack up pretty well, and the single thread performance doesn't look dire. Depending on price point, kabini could end up in a lot of mainstream laptops. If it did, that'd be great news for AMD...

[CAT-THE-FIFTH]

This is an interesting read:

http://www.eurogamer.net/articles/di...c-for-next-gen

[CAT-THE-FIFTH]

Kabini is shipping:

http://www.xbitlabs.com/news/cpu/dis...r_Revenue.html

HD8570 is tested:

http://www.eteknix.com/amd-hd-8570-o...ance-revealed/

[Noxvayl]

I would like to see devices in stores. It seems there is a lag with AMD products that doesn't happen for Intel.

The other thing that annoys me is the AiO computers not having good AMD options, the AMD APUs are much better choices in those machines.

[watercooled]

I wonder when we'll see some proper reviews?