AMD CPU speculation... and expert conjecture

[JAYDEEJOHN]

Understand, Intel knows good enough.
They also believe good enough in a GFX sense as well.
Just let them be a player mentality, even with sacrifice.

Also, DT isnt in the same position of expediency it once was, as solutions and markets change.
Under current economies, smaller cheaper good enough solutions are sought now.
Newer nodes and their costs per mm silicon also has a play in this, even if the economy rebounds.
Intels solution with GT3e will be good enough, not so sure about the others, and yes, AMD with GCN will see huge improvements and will eliminate the entry level bracket, where only the Intel solutions with a entry discrete will remain, to some degree.

I believe AMD is working hard on their cache solutions now, as well as BW and front end.
We may not see all this in one solution, but within 2-3 for sure

 

[esrever]

Edit.
http://www.tomshardware.com/news/Richland-APU-iGPU-Clock-Speeds-Specs,20748.html

looks like this will have the GCN nice at least that's good news i was just angry that they would use the same design as trinity with a clock bump when Intel is releasing a nice tock this time.
"GPU on the Richland APUs will be part of the HD8000 Series based on the GCN architecture"

Now with this phenom yes its time to do something new. Was a module design smart no to many latency problems with sharing cache, don't even understand why Amd did this thought they had smart engineers unless the marketing team was behind it somehow or dumb management who don't understand anything.

I think toms and the source they took that from is wrong. In AMD's slides it says radeon cores 2.0 which are the same as trinity.

 

[sarinaide]

GT3 will be 7660G level performance in high end ultrabooks for longer battery life.

Possibly or thereabouts, considering AMD's mobility iGPU's are still very low level that is possible.

 

[Blandge]

I think toms and the source they took that from is wrong. In AMD's slides it says radeon cores 2.0 which are the same as trinity.

I agree, and here's why:

Take this for what you will, but I've not seen anything directly from AMD that indicates Richland had adopted GCN.

Edit, Source: http://wccftech.com/amd-launching-28nm-kaveri-apu-steamroller-cores-2013/ (Same source that Tom's used to make their erroneous assumption about GCN.)

 

[de5_Roy]

wcdefghijklmtech is funny.
so far, it's far more likely for the low power apus like kabini and temash to get gcn-based igpu. incidentally, tsmc who makes the gcn gpus will also make the lp apus. would be better if the gcn cores are from the newer sea island gpus.
gcn being in richland would be downgrade imo (28nm to 32nm).
edit: i don't like amd dancing around the apu specs like this. would it hurt so much if they confirmed igpu type in ces? it's not like intel's gonna show them up(they can't). they've started being dodgy about kaveri and steamroller as well.

 

[JAYDEEJOHN]

Release your PR, eliminate a few SKUs, new CEO, restructuring etc.
I can see a period of this type of thing happening, for now.

 

[cgner]

The stocks have taken a hit (like any bubbles needing to burst) but they're also selling more product quarter over quarter. They still sold 1/2 a billion iOS devices last year. Staggering!

I didnt say they sell fewer devices, just fewer than they had hoped to sell. A bit of a difference there.

 

[noob2222]

Take this for what you will, but I've not seen anything directly from AMD that indicates Richland had adopted GCN.

AMD hasn't said that its VLIW either, only thing they have actually said is 20-40% improvement.

As far as you claiming wcctparrot as the source, all they do is parrot other sites. The speculation that richland was going to be gcn started as early as july of 2012. Since then every other website has been contradicting each other as to what it actually is.

http://news.softpedia.com/news/Richland-is-AMD-s-Next-Gen-APU-and-Fits-in-FM2-Socket-279049.shtml

WCCT as a source, hardly. They aren't even a source of their own, but they are useful to find parroted information.

 

[Blandge]

AMD hasn't said that its VLIW either, only thing they have actually said is 20-40% improvement.

As far as you claiming wcctparrot as the source, all they do is parrot other sites. The speculation that richland was going to be gcn started as early as july of 2012. Since then every other website has been contradicting each other as to what it actually is.

http://news.softpedia.com/news/Richland-is-AMD-s-Next-Gen-APU-and-Fits-in-FM2-Socket-279049.shtml

WCCT as a source, hardly. They aren't even a source of their own, but they are useful to find parroted information.

Well I do not dispute anything you've said, but given that AMD has announced that Kabini, Tamesh, and Kaveri will all be using GCN, and has remained silent about Richland, I think it can be safely assumed that Richland will not have adopted GCN.

 

[esrever]

I can see AMD just upping the mobile parts of richland to 45w and call it a performance increase at the very top. Richland is generally going to be trinity 2.0 with some marketing and software put in.

 

[sarinaide]

Honestly I speak and get free press from my AMD correspondent but at no time has AMD ever said what Richland's iGPU core is based upon. Last year's fud said it was going to have a reworked core, updated 8000D series part based on existing GCN archictecture, around that time Steamroller Kaviri was said to feature the updated GCN 2.0 parts. Now it leaks stories of a revised VLIW part. While I will not get into performance disputes it makes far more sense for AMD to use a 28nm GCN core cutting power drastically and allowing more aggressive stream processor counts.

If they get 20% on revised VLIW then it will be alone a feat of excellence if they get 20-40% then Steamroller Kaviri will probably be the most anticipated part in the pipeline as it may very well be mainstream lowend gaming level by then.

 

[Cazalan]

It's clear in the roadmap slides that AMD released at CES 2013 in January.

The 28nm chips due later in the year get the GCN features. The 32nm chips are stuck with what they got. Minor process tweaks and frequency bump. Both Trinity and Richland are marked as "2nd Generation A-Series APU"

 

[jdwii]

It's clear in the roadmap slides that AMD released at CES 2013 in January.

The 28nm chips due later in the year get the GCN features. The 32nm chips are stuck with what they got. Minor process tweaks and frequency bump. Both Trinity and Richland are marked as "2nd Generation A-Series APU"

So it seems it is the same thing as trinity, then were is Amd getting their 20-40% increase.

When well they ever learn

CPU clock speed increased 7.8%
CPU Turbo increased by 4.7%

GPU Clock speed increased 5.5%
And a 14.3% increase in memory bandwidth

What is the laptops increase in memory bandwidth and clock speed?

At least Intel's haswells latest benchmarks make things look a little better.

 

[JAYDEEJOHN]

Dont just look at it from a enthusiasts POV, look at it from AMDs POV.
It will please the OEMs

 

[jdwii]

Dont just look at it from a enthusiasts POV, look at it from AMDs POV.
It will please the OEMs

Why would it?

 

[esrever]

Why would it?

bigger numbers than last year's models means easier to market.

 

[jdwii]

bigger numbers than last year's models means easier to market.

Sounds about right.

 

[JAYDEEJOHN]

TDP to perf

 

[sarinaide]

Seeing the results so far on Haswell prior to testing the ES, looks like GT3 will match around a A8 3850/3870k level, problem is GT3 is limited to the high end parts which cost 2.2 and 3.5x more. Will still see a deficit of around 25-35% to Trinity and probably around 40-45% to Richland.

I am actually more interested in the frame latencies, synthetic wise Intels iGPU's are made out to be better than they are, its the stutter that really is the issue, where AMD is smooth HD is a lag fest.

 

[de5_Roy]

(rumored) new xbox ps consoles have amd all over them
http://www.hardcoreware.net/microsoft-durango-vs-sony-orbis-rumored-specs-comparison/

 

[JAYDEEJOHN]

HW on old consoles cant compare to this rumored set.
Ive stated earlier, itll be somewheres close to 8x perf overall, which includes everything, BW, gou, cpu.
Thats alot of room for the devs to do the the pooch and still make it look good.

 

[Tlight131]

I wish if they just dump AM3+ and Modules all together and go for the ultimate gamer setup which can be tied up with an AMD GPU. That would make it AMAZING just like abbadon_34 said. That would kill Intel and nVidia unless these two overpriced companies get together and form an ultimate gaming setup for 1000$ CPU and GPU only lol.

A10 GPUs are impressive on a cost basis.

 

[wh3resmycar]

APUs on a gaming desktop makes no sense. i mean you'll build around a setup focused primarily on the IGP. i have celeronSB right now and i'm planning whether on getting a i3-3220 or move to an APU setup. initial cost is good with the APU but longevity is questionable. if in the in the long run i'd get a discreet GPU why would i bother with an AMD setup?

hell, i can put a used 6750 in this celeronSB and never look back.

 

[de5_Roy]

apus are marketed to the entry level segment and people who replace their machine when they upgrade.
their longevity is as good as pentiums and celerons + entry level discreet gpu.

 

[esrever]

APUs on a gaming desktop makes no sense. i mean you'll build around a setup focused primarily on the IGP. i have celeronSB right now and i'm planning whether on getting a i3-3220 or move to an APU setup. initial cost is good with the APU but longevity is questionable. if in the in the long run i'd get a discreet GPU why would i bother with an AMD setup?

hell, i can put a used 6750 in this celeronSB and never look back.

why do people buy a cpu and upgrade it later?