The AMD Execution Thread [2007 - 2017]

[function]

They're probably selling better than Vishera.

If you're building a cheap desktop machine without a discreet GPU or one of those HTPC things there's really nothing better than Trinity. Richland should be even better.

Shame it doesn't seem to be widely available in laptops though.

 

[Alexko]

they got that from PD will very little in the way of uarch changes, SR has a large amount of uarch changes, if they only get 15% consider it a dissapointment.

Agreed. But 15~20% over Richland might be OK.

Does it ? How are the trinity apu's selling ?

Better than Vishera.

 

[jimbo75]

Trinity shipments doubled last quarter, there may actually be a shortage developing due to demand and not enough supply because of AMD cancelling wafers at GF.

 

[function]

Desktop Richland:

http://techreport.com/news/24277/leaked-richland-specs-reveal-higher-clock-speeds

Another decent bump while still on the same old 32 nm process. That A10 6700 at 65W is really nice, if I was building a HTPC or a mini desktop for someone nothing else would even get a look in.

It makes you realise just how premature Bulldozer was. And Piledriver isn't even using that resonant clock mesh tech - assuming that Richland is it doesn't appear to hurt clocks any.

 

[jimbo75]

Of most interest is the graphics numbering for the lowest SKU A6-6300, 8370D.

That's interesting because what we believe to be the highest Kabini SKU, the A6-5200 (I'm going to assume this will be renamed as the A6-6200), has graphics numbered as 8400 - http://www.amd.com/us/press-releases/Pages/amd_unveils_new_apus.aspx

Let's assume the 6300 has the same 128 SP's as the 5300, and if that leak is correct they are at 760 MHz. This should give us a reasonable indication of Kabini's stats, however the single channel memory troubles me a bit as I'd have thought that would hurt performance a lot.

Best guess says 25W, 4 cpu cores at unknown speeds and 128 SP's at 800 MHz though. AMD would not give it the 8400 name unless it was faster in graphics than the Richland 6300.

For those of you who aren't really keeping up with these bottom end APU's, I reckon the performance should be fairly close to the A6-3500 http://www.guru3d.com/articles_pages/amd_a6_3500_apu_review,10.html

It looks like a huge perf/watt increase anyway, as these A6's are rated 65W (from what I've seen of them, 50W appears to be a bit closer to the average). Still, looking at a doubling of Perf/Watt in all likelyhood.

The A6-6400K will probably be about the same (in gaming) as the Llano 3870K btw.

 

[no-X]

Desktop Richland:

http://techreport.com/news/24277/leaked-richland-specs-reveal-higher-clock-speeds

Another decent bump while still on the same old 32 nm process. That A10 6700 at 65W is really nice, if I was building a HTPC or a mini desktop for someone nothing else would even get a look in.

Yes, CPU performance almost identical to A10-5800K (Trinity), GPU performance higher and TDP lowered by 35 watts. Really nice.

 

[Alexko]

Desktop Richland:

http://techreport.com/news/24277/leaked-richland-specs-reveal-higher-clock-speeds

Another decent bump while still on the same old 32 nm process. That A10 6700 at 65W is really nice, if I was building a HTPC or a mini desktop for someone nothing else would even get a look in.

It makes you realise just how premature Bulldozer was. And Piledriver isn't even using that resonant clock mesh tech - assuming that Richland is it doesn't appear to hurt clocks any.

Trinity is using it, so Richland should too.

 

[jimbo75]

Woops I forgot Kabini is GCN, so clocks might be closer to 650-700 MHz I guess.

 

[UniversalTruth]

According to the agency Fitch, AMD is nearing bankruptcy

 

[rpg.314]

I am not surprised. The long cycles of product development in ICs don't help either.

It's only a matter of time now, it seems.

 

[Kaotik]

Intel will probably rescue them one way or another in case this would happen, since they would have trade comissions around the world chopping them to pieces before they could say "monopoly" if AMD went down

 

[AlexV]

since they would have trade comissions around the world chopping them to pieces before they could say "monopoly" if AMD went down

Yes but no. Whilst this is a commonly held view, from the large family of "you might've beaten me up quite badly, but man, did the sole of your boot break on my nose!", it has less and less weight with each passing day. AMD might well get a white knight (they definitely need it), however Intel is increasingly unlikely to actively engage in keeping them alive, as their role of anti-anti-trust buffer is quite well covered.

 

[silent_guy]

Intel will probably rescue them one way or another in case this would happen, since they would have trade comissions around the world chopping them to pieces before they could say "monopoly" if AMD went down

It's not against the law to operate a monopoly... as long as you don't abuse it. Intel just would have to be even more careful to avoid doing this.

It's not their fault that AMD can't design competitive products (or earlier, fabs.)

 

[AlphaWolf]

It's not against the law to operate a monopoly... as long as you don't abuse it. Intel just would have to be even more careful to avoid doing this.

It would take about 10seconds for someone to claim they were abusing it. Hell they've already been abusing it for years.

 

[Kaotik]

however Intel is increasingly unlikely to actively engage in keeping them alive, as their role of anti-anti-trust buffer is quite well covered.

Are they? They've had to pay AMD already due abusing their position, so they do have history of abusing even their current situation proven in court.

 

[digitalwanderer]

How could there be anything wrong with AMD, they got a new gaming bundle out!

(WARNING: The link is a blatant attempt to pimp a story at EB! )

 

[UniversalTruth]

A lot of things are wrong with AMD. Especially this one:

As you know, in the last quarter AMD had to pay Globalfoundries penalty in the amount of $ 320 million for a reduction in the volume of purchases of wafers.

I don't understand how they did dare to do it. I mean they could have launched a huge advertisement campaign with this money, or manufacture the chips and sell them to gain some market share.
Simply ridiculous.

 

[AlphaWolf]

The $320 million would be a significant reduction in the actual cost to them. So rather than buying product they probably couldn't sell they agreed to pay a penalty.

 

[AlNom]

How could there be anything wrong with AMD, they got a new gaming bundle out!

"Silhouette-enhancing tessellation"

 

[UniversalTruth]

So rather than buying product they probably couldn't sell they agreed to pay a penalty.

My point being is that this is impossible. It is all relative and there is always price at which everything will sell. That's why I suggested an investment type of strategy to replace this ridiculous penalty...

Instead they probably have lower actual cost but higher indirect losses too.