AMD Piledriver rumours ... and expert conjecture

Page 61 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.
We have had several requests for a sticky on AMD's yet to be released Piledriver architecture ... so here it is.

I want to make a few things clear though.

Post a question relevant to the topic, or information about the topic, or it will be deleted.

Post any negative personal comments about another user ... and they will be deleted.

Post flame baiting comments about the blue, red and green team and they will be deleted.

Enjoy ...
 
In most games the gpu is what pulls it's weight

Not as much recently; In the recent THG CPU review, only the DX11 titles were GPU limited using a 7970. While at the top of the high end, this still shows the CPU has a MAJOR part to play, and can not be discounted. The A4 got crushed [the A8 held its own with the pack though].

http://www.tomshardware.com/reviews/gaming-fx-pentium-apu-benchmark,3120-10.html

I'll say it one final time: The decision to not port DX10 to XP basically ensured DX9 remained the dominant graphical API, because developers will not abandon 50% of their potential market. All DX11 has done so far is replace DX10.
 
It does surprisingly well provided you are not trying to play at 1080p and were not cpu limited. Besides the gpu the other main benifit of mobile Llano for gaming to go is the unlocked cpu multiplier. Unlike the desktop version of Llano the multiplier is unlocked and allows for every easy overclocking. Some have gotten a little over 3ghz stable. In most games the gpu is what pulls it's weight. Around 2.4ghz mobile Llano makes up for it's weakness compared to Intel very well.

OCing laptops is not exactly what my needs are but playing on a medium setting is not bad
Intel missed the boat in mobile I hope Trinity is a good uplift enough that I can get a ulv 17-35 watt version and still equal lano
 
will Intel graphics ever be as good, I don't think so unless they have some kind of breakthrough or 'acquisition'.
but the CPU side of things mobile included IB will be a gap widener..
but where AMD is going with APU graphics is far ahead.

yes Intel can catch up by buying a third of nvidia
 
OCing laptops is not exactly what my needs are but playing on a medium setting is not bad
Intel missed the boat in mobile I hope Trinity is a good uplift enough that I can get a ulv 17-35 watt version and still equal lano

That was just one aspect of use when it comes to AMD laptops. Many when not gaming or doing any tasks that needed the performance often underclocked when on battery to run cooler while lasting longer. When I was in class with mine I often underclocked all the way down to 533mhz. Can go further but didn't and dropping the volts below .8v made it run very cool. One can undervolt the e-350 ect to gain up to and over an hour more on battery.
 
Overpriced is what the market will not bear ,should that be the case sales should suffer. Everyone has a limit as to what they are prepared to pay .Since there are still lots of 6870-6850 in the pipeline amd can afford to keep prices higher for now.
Prices are not written in stone
As far as PD and Trinity are concerned prices will be commensurate with improvements over BD 25% improved should equal 25% price hike. Again if the prices are too high sales will suffer and then they will go on sale.
Like 1000$ intel parts they will languish on the shelf .
Trinity will be over priced as lano parts will still be available .
Nvidia is in trouble I'd be surprised if they can come back with low power improved parts this year




Intel can get away with overcharging Amd can't!
 
In most games the gpu is what pulls it's weight

Not as much recently; In the recent THG CPU review, only the DX11 titles were GPU limited using a 7970. While at the top of the high end, this still shows the CPU has a MAJOR part to play, and can not be discounted. The A4 got crushed [the A8 held its own with the pack though].

http://www.tomshardware.com/reviews/gaming-fx-pentium-apu-benchmark,3120-10.html

I'll say it one final time: The decision to not port DX10 to XP basically ensured DX9 remained the dominant graphical API, because developers will not abandon 50% of their potential market. All DX11 has done so far is replace DX10.

That's not quite right. We're talking mobile gaming, meaning Sabine platform not Llano although most lump them into the same category.

Those bench's were done at 1920x1080 with 4xAF and 8xAF with a high end discrete GPU. That completely defeats the purpose of an APU to begin with. Half their die space is devoted to a GPU that your just disabling. Redo those bench's with the built in GPU vs the HD3K on the SB's and watch the APU's smoke them. Which is the entire point of the APU to begin with. You won't be doing high end gaming with a APU, it just won't work. It's good for mobile gaming and low power budget PC's. That was a good article for desktop decisions as it show's that a decent discrete GPU greatly diminishes the APU's worth.

Bench's done with just the 6620G (build into the 3530MX).

http://forum.notebookreview.com/gam...g-benchmarked-various-ram-configurations.html

These were done to show how memory speed effects the GPU's performance and what to expect out of it.

As I have the HP DV6q model with a 3530mx w/o the 6750 addon, I can confirm that it games well at 1366x768 with low~medium graphics settings.

This is not for hardcore gamers / enthusiasts who want 1920x1080x120hz with 4~8x AA and 8~16x AF at Ultra Hard Core graphics settings. Its for people who need a cheap portable computer that can do all the standard office stuff, watch HD BR movies, and light gaming. For about $220 more you can get the 1920x1080 screen and a 6750m (low to mid level discrete GPU) added to that notebook. It's been suggested you get them together because the 6620G won't get acceptable frame-rates at 1920 but the 6750 + 6620G together can. Also DDR3-1600 4GBx2 kits are approx $50 USD so there is room for some performance benefit (OEM's like to stick DDR3-1333 on this platform).

It sounds counter-intuitive but you want to overclock + under-volt the APU. Not only is the multiplier unlocked but all the p-states are also unlocked. This lets you set P0/B0 at 2.6Ghz and under-volt P1~P4. When your gaming or in demanding applications it will clock to 2.6Ghz but when your just doing office stuff or movies it'll clock to one of the lower under-volted power states to extend battery life. Having done this for a few months now, I can say I'm quite happy with my purchasing decision. My GF enjoys this laptop to no end.

So for his situation, a significant other that needs a new mobile PC that has long battery life and can become a functional low end gaming platform, anything with a Sabine APU makes sense. For desktops, if you have to use a discrete GPU then go SB, if you don't want / need a discrete GPU, or the physical space doesn't allow for one, then go with an APU.
 
I think it will be a combination of ipc and clock
As Paladin said a rearrange of the front end of BD and a good lift in speed which makes sense
they claim a 35 watt will be equal to lano 65 watt which is a big lift what ever that means
When AMD said it would be better than the expected 20 50 they previously claimed they had to have found a fix for BD
and are implementing it in PD. which also could explain the date being pushed back

AMD also said BD would be the most efficient single and multithreaded CPU ever. They kinda missed the mark there, don't ya think?

AMD's upcoming Pitcairn..."

http://www.fudzilla.com/home/item/25998-amds-upcoming-pitcairn-gpu-detailed

Wont be much faster than the 6850/6870. Only GPu that looks worth while is the HD7900 series and those cost way too much as it is.

you do realize that 3Dmark physics is very multithreaded right? That means with LLano vs PD, your testing CMT, that is not a per-core test, AMD already stated CMT = 80% of a dual core, and in most tests, thats pretty dead on. Take away the 13% disadvantage and that gives PD a ~7-14% ipc over llano on a single core test.

Looks like pd should be an improvement over stars in single threaded tests. If you really thought PD would improve over stars in multithreaded testing, then CMT would either have to be 100% scaling, or ipc 20% higher than stars.

So you are blaming me for the benchmarks that show equal clock now? I know all about the "80%" of a core. Never said it wasn't. But AMD will market it as a 4 core A CPU and it will replace Llano as such.

You compare top end Llano to top end Trinity. You can't make excuses. If AMD is claiming 20% better then its either 20% better or not, no fudging the numbers due to them not being 100% full fledged cores. Thats just trying to make it look better than it really is.

Try giving that excuse to BD which has 8 "Cores" yet still fails to beat, consitently, a 2600K with SMT, known to at best give a 20% bump, in multithreading.

The clock speed doesn't surprise me too much. It's basically an FX-4120 with a 100mhz bump.
Minor cache improvements + clock speed should get the 20% over llano without much issue.

While not totally new there is mention of a Bulldozer B3 stepping as early as October '11. That would be the 6th incarnation of Bulldozer.

I think Trinity will do well this year but Haswell will be a problem for them in 2013. 22nm tri-gate is a game changer.

Before Haswell. Ivy bridge is 22nm 3D Tri Gates. Haswell is a new arch on 22nm 3D Tri Gates. IB will be the first problem and there is a post in the IB thread of IBs GPU being 67% faster than SB, which would go in line with Intels 50% claim. It also shows IB getting 30FPS+ on games like Dirt 3 and FC2 which is pretty nice for a Intel IGP.
 
check in 2006 Intel was floundering bad stock price marked down badly AMD was selling 100 to every one Intel part
about the exact reverse of today.
times change in the business world fortunes can change quickly especially if your competitor is paying your customers
to not buy your product as Intel was caught doing at least they paid billions for their crime.


http://www.pcpro.co.uk/features/372859/amd-what-went-wrong

AMD in their peak (Q2 of 2006) was still vastly outsold by Intel. Digging up an old report from July 2006 about Q2 2006 sales figures:

- Server market share: AMD 33%, Intel 66%.
- Desktop market share: AMD 22%, Intel 73%. Notably, Intel had 81% of the desktop revenue share, AMD had 18%.
- Mobile market share: AMD 13.3%, Intel 86.7%.

Intel was also worth $110.52B in Q2 of 2006, while AMD was worth $15.99B at that time.

So while you may have been selling A64s and X2s 100 to 1 versus Pentium 4s and Pentium Ds in your custom build store, the market as a whole was still buying tons and tons of Pentium Ms, Yonahs, and P4s/PDs. The P-M and Yonah were actually good chips and it was reasonable that they sold well. Dell and others were stuffing P4s and Pentium Ds into machines, for lots of reasons ranging from "nobody got fired for buying Intel" PHBs placing orders for 10k lots of OptiPlexes, the general public who just remembered the "Intel Inside" chimes and TV ads and buying OEM machines, and Intel's monopoly abuse tactics that bought them >$1B in fines and multiple antitrust suits from various governments.
 
AMD in their peak (Q2 of 2006) was still vastly outsold by Intel. Digging up an old report from July 2006 about Q2 2006 sales figures:

- Server market share: AMD 33%, Intel 66%.
- Desktop market share: AMD 22%, Intel 73%. Notably, Intel had 81% of the desktop revenue share, AMD had 18%.
- Mobile market share: AMD 13.3%, Intel 86.7%.

Intel was also worth $110.52B in Q2 of 2006, while AMD was worth $15.99B at that time.

So while you may have been selling A64s and X2s 100 to 1 versus Pentium 4s and Pentium Ds in your custom build store, the market as a whole was still buying tons and tons of Pentium Ms, Yonahs, and P4s/PDs. The P-M and Yonah were actually good chips and it was reasonable that they sold well. Dell and others were stuffing P4s and Pentium Ds into machines, for lots of reasons ranging from "nobody got fired for buying Intel" PHBs placing orders for 10k lots of OptiPlexes, the general public who just remembered the "Intel Inside" chimes and TV ads and buying OEM machines, and Intel's monopoly abuse tactics that bought them >$1B in fines and multiple antitrust suits from various governments.

Imagine and they laid off 10% of their work force while AMD had the fastest parts and could barely keep up with orders
goes to show you being big is a determent if your parts can't keep up.
 
So you are blaming me for the benchmarks that show equal clock now? I know all about the "80%" of a core. Never said it wasn't. But AMD will market it as a 4 core A CPU and it will replace Llano as such.

You compare top end Llano to top end Trinity. You can't make excuses. If AMD is claiming 20% better then its either 20% better or not, no fudging the numbers due to them not being 100% full fledged cores. Thats just trying to make it look better than it really is.
I assume your referring to the xbit labs article, http://www.xbitlabs.com/news/cpu/display/20111026223104_AMD_Expects_Trinity_t😵ffer_20_30_Performance_Increase.html


The 20% improvement represents AMD's projections "using digital media workload" and actual performance advantage over currently available Fusion A-series "Llano" vary depending on the applications and usage models

Maybe intead of getting bent out of shape because someone made an attemp to analyze what a cpu physics score means with trinity, stick with the facts instead of twisting them to mean what wasn't said. Then again, maybe you just read the title of the book, I don't know. Trinity isn't 20% overall, and its not the end of the world.
 
Those bench's were done at 1920x1080 with 4xAF and 8xAF with a high end discrete GPU. That completely defeats the purpose of an APU to begin with. Half their die space is devoted to a GPU that your just disabling. Redo those bench's with the built in GPU vs the HD3K on the SB's and watch the APU's smoke them. Which is the entire point of the APU to begin with. You won't be doing high end gaming with a APU, it just won't work. It's good for mobile gaming and low power budget PC's. That was a good article for desktop decisions as it show's that a decent discrete GPU greatly diminishes the APU's worth.

You are aware the entire point of the article was to bench CPU's? All that was shown was that llano was hopelessly CPU bottlenecked compared even to $80 processors in most every major title, even when paired with the best GPU on the market.

My point being, llano will never be suitable for gaming with any significant graphical settings, simply because the CPU side is a cripple. Now theres nothing wrong for having a low power IGP for laptops/netbooks that is actually halfway decent, but as I've said before, I expect tablets/smartphones to make those two form factors obsolete within a decade. I simply don't see APU's as a major factor going forward.
 
Those bench's were done at 1920x1080 with 4xAF and 8xAF with a high end discrete GPU. That completely defeats the purpose of an APU to begin with. Half their die space is devoted to a GPU that your just disabling. Redo those bench's with the built in GPU vs the HD3K on the SB's and watch the APU's smoke them. Which is the entire point of the APU to begin with. You won't be doing high end gaming with a APU, it just won't work. It's good for mobile gaming and low power budget PC's. That was a good article for desktop decisions as it show's that a decent discrete GPU greatly diminishes the APU's worth.

You are aware the entire point of the article was to bench CPU's? All that was shown was that llano was hopelessly CPU bottlenecked compared even to $80 processors in most every major title, even when paired with the best GPU on the market.

My point being, llano will never be suitable for gaming with any significant graphical settings, simply because the CPU side is a cripple. Now theres nothing wrong for having a low power IGP for laptops/netbooks that is actually halfway decent, but as I've said before, I expect tablets/smartphones to make those two form factors obsolete within a decade. I simply don't see APU's as a major factor going forward.
Last I checked AMD was pushing to put apu's in tablets+smartphones 😛.

I am just about to give up on people these days. Screw it. I'm caps locking this.

APU'S ARE MADE WITH THE IDEA OF HAVING GOOD CPU+GOOD GPU ON ONE CHIP, WITH NO DISCRETE GPU TO HELP. THEY ARE NOT MADE WITH THAT IDEA IN MIND.

Llano's cpu does not bottleneck its gpu, so there is no issue with the chip. Apu's are amazing, admit it and move on.
 
@gamerk316: imo, this is exactly what 'they' refuse to understand. that article showed how a sub-$200 desktop cpu would perform for gaming. that's it. there was no mobile issue, no igpu issue. llano igpu's superiority was already demonstrated, but it's ability to drive a gaming gfx card was not demonstrated.
and who's to say what cpu should be tested or not? that's like ordering someone to abide by a restricted pc configuration (apple, anyone?).
when llano came out in june~ last year, i really, really wanted to build a llano pc. my plan was to use the igpu for a few months then add a radeon hd 6850 or gtx 560ti right around now and turn it into a cheap gaming pc. since the a8 3850s could be overclocked to 3+ ghz, i figured it's be great for a cheap (compared to other 'cheap' quadcore cpu available at that time - 2.8 ghz core i5 2300), single gfx card gaming pc. plus, i'd have a cheap 32 nm quad core cpu for multitasking and stuff. i knew that the apus were entry level, but they should at least be able to drive a mid range/upper mid range gfx card like cheap, entry level pentiums or slightly costlier core i3 cpus. i am a bit relieved that i didn't go that way. i still want a capable cpu + an igpu like llano's. llano set the bar for igpu, for desktop and mobile.
 
It could be the end for our old mate Triny in this thread.

He is predicting 25% overall for CPU and 60% for GPU. 😱



At least you understand that instead of being smart and underestimating performance improvements Like a smart company should Amd Always overestimates their performance expectations, Like 50% more through-output when in reality its less then 20%. Amd really needs to run 50+ benchmarks of programs that people actually use and then find the price/performance ratio or does that make to much since!
 
This might explain some of the power savings in how they were able to get the higher clocks with Piledriver (FX-4120+GPU).

AMD packs Cyclos clock tech into Piledriver.

"To back its claims of power savings up, Cyclos has some real-world figures. Based on Piledriver-based x86 processing cores running at 4GHz and above, the resonant clock mesh technology drops clock distribution power by up to 24 per cent at peak and between five and 10 per cent on average in the company's testing. Clock-skew, a serious issue in high-speed processors, is claimed to be unaffected by the drop in power draw."


http://www.bit-tech.net/news/hardware/2012/02/21/amd-packs-cyclos-piledriver/1

 
It could be the end for our old mate Triny in this thread.

He is predicting 25% overall for CPU and 60% for GPU. 😱


25-30% is my prediction but I certainly won't lose any sleep if AMD failed and went bankrupt
though I like the idea of HSA it could very well be a paper tiger
and I stated 25-30 based on a10 to a8 using +ipc and + clock
no where did I say it would be +25-30% on a same freq basis
a10 4.2 to a8 lano 2.9
in the end trinity will be 25-30 + on cpu 60% igp minimum could be 80% product vs product
I also think price for trinity will be 25-30% more
you are too afraid to make a prediction
 
Status
Not open for further replies.