AMD Trinity Reviews

[Over_Lord]

Gonna be 15th may in a few hours. When does the NDA lift?

Eagerly waiting for the reviews. Do post here, I will.


http://www.anandtech.com/show/5831/amd-trinity-review-a10-4600m-a-new-hope

http://hexus.net/tech/reviews/cpu/39333-amd-a10-4600m-trinity-apu/

http://www.expertreviews.co.uk/laptops/1292020/amd-trinity-review-hands-on-with-a-prototype-laptop

http://www.pcper.com/reviews/Mobile/AMD-A10-4600M-Trinity-Mobile-Review-Trying-Cut-Ivy

http://vr-zone.com/articles/amd-trinity-a10-4600m-apu-review-jumping-the-shark-/15830.html

http://hothardware.com/News/AMD-Trinity-A104600M-Processor-Review/

ROUNDUP:

http://www.overclock.net/t/1257540/various-amd-trinity-apu-review-roundup

 

[W1zzard]

you realize that desktop products will be announced at a later time?

 

[Mathragh]

For some reason I'm quite exited about this launch.
Really looking forward to see that tech at work, even if it only includes laptop chips for the time being.

 

[btarunr]

Some nice 14-inch, 15.6-inch and 12-inchers will be launched.

Desktop Trinity APUs arrive later.

 

[tacosRcool]

I want them to be put in AMD's equivalent to an ultrabook!

 

[Over_Lord]

http://notebook.pconline.com.cn/testing/contrast/1205/2781947_all.html#content_page_1

Here you go, use TRANSLATE

http://www.overclock.net/t/1257289/pc-online-china-amd-trinity-apu-review

 

[Over_Lord]

http://www.anandtech.com/show/5831/amd-trinity-review-a10-4600m-a-new-hope

http://hexus.net/tech/reviews/cpu/39333-amd-a10-4600m-trinity-apu/

http://www.expertreviews.co.uk/laptops/1292020/amd-trinity-review-hands-on-with-a-prototype-laptop

http://www.pcper.com/reviews/Mobile/AMD-A10-4600M-Trinity-Mobile-Review-Trying-Cut-Ivy

http://vr-zone.com/articles/amd-trinity-a10-4600m-apu-review-jumping-the-shark-/15830.html

http://hothardware.com/News/AMD-Trinity-A104600M-Processor-Review/

ROUNDUP:

http://www.overclock.net/t/1257540/various-amd-trinity-apu-review-roundup

 

[Steevo]

Yet another turd, being polished by the worst PR team ever.


I'm glad I got my 1100T before they killed these great chips. Stock voltage to do 3.7Ghz/4.1Turbo


The retards at AMD need to get with the program, right now my next system will be Intel/Nvidia as I am tired of their half assed bullshit attempts at a "bulldozer" that is more like a Yugo with a snow blade.

 

[Mathragh]

I'm glad I got my 1100T before they killed these great chips. Stock voltage to do 3.7Ghz/4.1Turbo

Not trying to talk things right here, but do you realize that the fact that this chip(piledriver, and bulldozer for that matter) performs the way it does is possibly in no small way as consequence of the state of the 32nm process compared to the 45nm process?

If you compare things like voltage and clockspeed for the athlons produced at 45nm, and Llano(Athlons) produced at 32nm, you'll see that 32nm actually didnt really improve anything when it comes to high frequency and voltage at higher frequencies.(Which turns out especially bad when you just went all in with a arch aimed at higher frequencies)

Llano's(athlons) at 32nm dont even seem to be able to break the 4GHz barrier(just look up different 'K' Llano model reviews), while you indeed correctly state that at 45nm they could easily break it, even with stock voltage.

One can only wonder how things would've looked if the 32nm did scale better compared to 45nm (like it seems to be at Intel, where 32nm chips did seem to clock higher and use lower voltages compared to 45nm chips).

Just a quick comparison taking the facts quite liberally. The highest clocking athlon/phenom at 45nm with a 100W TDP clocks at around 3,4GHz, while the highest clocking Llano with the GPU part disabled and a 100W TDP clocks at 3GHz.
this would mean that actually, when using the same kind of logic and using the same TDP, the 32nm process gets you about 13% less Hertz.

Taking this with pounds of salt, one could extrapolate that in order to fairly compare piledriver(or bulldozer for that matter) with K10(phenom), you would need to increase the clocks of the former by 13% for things to look *fair*. This would mean that the FX-8150 all of a sudden would turbo at 4200MHz*113%= 4,76GHz instead of 4,2GHz.


This doesnt in any way improve things for products out now, and doesnt make AMD any more attractive, but it just got me thinking. It could atleast explain why AMD still went with the bulldozer arch instead of sticking with K10 and just adding functionality to that(if that were possible atleast, I'm not an engineer).

Feel free to critisize upon/make fun of/ give feedback on whatever stated above.

 

[Absolution]

Why the heck are discrete graphics solutions in laptops being compared with the inbuilt graphics core? Just to keep intel on top of the chart? -.-

Or why didnt AMD send out a dual graphics laptop for reviews...

Edit : http://cdn.overclock.net/5/55/5523dae4_trinity_a10.jpeg

 

[Dent1]

Yet another turd, being polished by the worst PR team ever.

Yet another Intel fanboy, being polished up as unbiased.

I'm glad I got my 1100T before they killed these great chips. Stock voltage to do 3.7Ghz/4.1Turbo


The retards at AMD need to get with the program, right now my next system will be Intel/Nvidia as I am tired of their half assed bullshit attempts at a "bulldozer" that is more like a Yugo with a snow blade.

Why are you comparing an enthusiast 1100T with an Trinity which is an low end APU? If you want enthusiast performance that is what Piledriver is for.

 

[Vulpesveritas]

Not trying to talk things right here, but do you realize that the fact that this chip(piledriver, and bulldozer for that matter) performs the way it does is possibly in no small way as consequence of the state of the 32nm process compared to the 45nm process?

If you compare things like voltage and clockspeed for the athlons produced at 45nm, and Llano(Athlons) produced at 32nm, you'll see that 32nm actually didnt really improve anything when it comes to high frequency and voltage at higher frequencies.(Which turns out especially bad when you just went all in with a arch aimed at higher frequencies)

Llano's(athlons) at 32nm dont even seem to be able to break the 4GHz barrier(just look up different 'K' Llano model reviews), while you indeed correctly state that at 45nm they could easily break it, even with stock voltage.

One can only wonder how things would've looked if the 32nm did scale better compared to 45nm (like it seems to be at Intel, where 32nm chips did seem to clock higher and use lower voltages compared to 45nm chips).

Just a quick comparison taking the facts quite liberally. The highest clocking athlon/phenom at 45nm with a 100W TDP clocks at around 3,4GHz, while the highest clocking Llano with the GPU part disabled and a 100W TDP clocks at 3GHz.
this would mean that actually, when using the same kind of logic and using the same TDP, the 32nm process gets you about 13% less Hertz.

Taking this with pounds of salt, one could extrapolate that in order to fairly compare piledriver(or bulldozer for that matter) with K10(phenom), you would need to increase the clocks of the former by 13% for things to look *fair*. This would mean that the FX-8150 all of a sudden would turbo at 4200MHz*113%= 4,76GHz instead of 4,2GHz.


This doesnt in any way improve things for products out now, and doesnt make AMD any more attractive, but it just got me thinking. It could atleast explain why AMD still went with the bulldozer arch instead of sticking with K10 and just adding functionality to that(if that were possible atleast, I'm not an engineer).

Feel free to critisize upon/make fun of/ give feedback on whatever stated above.

Problem: 100w llano part has the equivalent of a radeon hd 5570 without the RAM under that 100w TDP envelope.

 

[D007]

Yet another Intel fanboy, being polished up as unbiased.




Why are you comparing an enthusiast 1100T with an Trinity which is an low end APU? If you want enthusiast performance that is what Piledriver is for.

Lol you did check his specs right? You read he has an AMD/ATI build? Just sayin...

 

[Dent1]

Lol you did check his specs right? You read he has an AMD/ATI build? Just sayin...

Which makes Steevo's post even worse. He claims to support AMD, but bashes AMD in an unconstructive way with inaccurate facts.

 

[Zen_]

It will be interesting to see what the price advantage is for Trinity over Intel products in retail ultrabooks. They are still too expensive for many students, or an impulsive toy buy over the usual lot of $400-500 laptops / tabs. At the target $600 that changes. I don't think most people really care at all what the hardware is as long as it's fast enough and the battery life isn't miserable.

 

[Mathragh]

Problem: 100w llano part has the equivalent of a radeon hd 5570 without the RAM under that 100w TDP envelope.

Please try to accurately read what I wrote. I specifically stated it was a Llano part with the GPU part disabled.

 

[Vulpesveritas]

Please try to accurately read what I wrote. I specifically stated it was a Llano part with the GPU part disabled.

Well, that's funny, there are no 3ghz llano parts with unlocked multipliers that have their GPUs disabled. Plus, if the GPU cores aren't lasered off, they'll still draw some power, as will the pins.

 

[Mathragh]

Well, that's funny, there are no 3ghz llano parts with unlocked multipliers that have their GPUs disabled. Plus, if the GPU cores aren't lasered off, they'll still draw some power, as will the pins.

Unlocked multies dont matter at all for chip quality, furthermore, as far as I understand, fusing a part off does mean it is physically separated from the rest of the chip, meaning no current can/will run trough it.

You can also look at it another way, if you dont buy my story: Why do AMD 32nm chips need so much voltage compared to Intel 32nm chips?

 

[Vulpesveritas]

Unlocked multies dont matter at all for chip quality, furthermore, as far as I understand, fusing a part off does mean it is physically separated from the rest of the chip, meaning no current can/will run trough it.

You can also look at it another way, if you dont buy my story: Why do AMD 32nm chips need so much voltage compared to Intel 32nm chips?

Architecture differences. Plus I am still quite sure even with the cores with disabled graphics, I'm not quite sure given everything that the FM1 athalons have the GPU fused off. If they're only software-disabled, then that could be the issue with OCing right there. Plus, all the good llano yields went to mobile. 1.5/2.3ghz A8's which OC up to 3.2ghz (not turbo) under 50w. Given all the issues llano had with yeilds, we really didn't get to see all that STARS might have offered us...

 

[Mathragh]

Architecture differences. Plus I am still quite sure even with the cores with disabled graphics, I'm not quite sure given everything that the FM1 athalons have the GPU fused off. If they're only software-disabled, then that could be the issue with OCing right there. Plus, all the good llano yields went to mobile. 1.5/2.3ghz A8's which OC up to 3.2ghz (not turbo) under 50w. Given all the issues llano had with yeilds, we really didn't get to see all that STARS might have offered us...

I might have been a bit overzealous when i started comparing intel 32nm with globalfoundries 32nm.

However, my point still stands, with Llano unable to reach 4GHz+, even with adequate cooling.

 

[jrs3000]

According to anandtech review they say Trinity is behind 20-25% so how far was bulldozer behind?

 

[Vulpesveritas]

I might have been a bit overzealous when i started comparing intel 32nm with globalfoundries 32nm.

However, my point still stands, with Llano unable to reach 4GHz+, even with adequate cooling.

The better batches could, if they had the power suplly and cooling for it, given the 3.2Ghz on 50w, stuck in a laptop, and having integrated graphics. No reason it couldn't go higher if not for the heat.
Seeing as short term 4.6ghz+ oc runs have been done on the APUs with liquid cooling and all. Main problem being that desktop users get the bad / lower end chips. All the binned chips are in mobile. And even there you run into the bit where the CPU isn't even able to utilize all the voltage being poured in, given that many of the pins have to go through the GPU anyhow to deliver voltage to the core I think... so the GPU will heat up anyhow. I think.

Also @jrs Anandtech's review has some off things. At 3.2ghz turbo trinity vs 2.4ghz llano single thread, Trinity is 37% faster with 32% faster clocks, so IPC is faster clock for clock, but floating point performance is lower when all cores are being used per core due to the said shared resources. Integer performance (what most CPU calculations are) is faster though. Bulldozer is on average 10-20% slower than Llano/Phenom II clock-for-clock/core-for-core.

 

[Dent1]

According to anandtech review they say Trinity is behind 20-25% so how far was bulldozer behind?

Very good question.

I'm only guessing here. But I'd think Bulldozer would be about the same gap behind, about 20-25% behind. These Trinity Mobile APUs are cut down and gimped for low power consumption and battery life in laptops where as the Bulldozer was full blown enthusiast CPU.

The real question willl the desktop Trinity APUs close the gap. Or better yet, will the Piledriver further close the gap.

 

[Grant]

Desktop AMDs certainly did close the gap. Check out these awesome motherboards with AMD Trinity FM2. They rock!

Best AMD Trinity CPU FM2 Motherboards 2012