AMD what will happen in 2015?

[cbn]

The a8 provides 90% performance of a10 in games for a fraction of price.

AMD could probably get that level of performance (A8-7600) from a 35 watt Mobile Kaveri (with configurable TDPup for very small form factor desktop.)

There is no strong reason to have Kaveri on the desktop Socket FM2+. It would be better off as a mobile SKU where its integration would help package it more compactly in a laptop.

Furthermore, in a low power laptop scenario the lack of bandwith (OEM DDR3 1600) doesn't hurt its gaming performance since the GPU is now downclocked compared to the speed it was on the desktop.

 

[mrmt]

so intel make good cpus and good enough gpus but amd makes "AMD sells poor CPU with a good but bandwidth choked iGPU." that is beyond skewed and is pure hypocrisy.

So in your POV Intel and AMD are doing essentially the same thing, developing the same basic strategy? Could you please develop more?

 

[AtenRa]

The vast majority of the time the use case is gaming though.

GPGPU is just too niche to spend all that die area for desktop iGPU at this time.

The context of the conversation was GPGPU no matter if it is a niche or not. I never talked about gaming in this instance so try to stay on topic.

A10-7850K is cheaper, faster and consumes less energy than having a traditional CPU + dGPU for GPGPU loads.

 

[AtenRa]

So in your POV Intel and AMD are doing essentially the same thing, developing the same basic strategy?

Broadwell dual core that was just released, has 50% or more of its die dedicated to the iGPU.

 

[cbn]

The context of the conversation was GPGPU no matter if it is a niche or not. I never talked about gaming in this instance so try to stay on topic.

A10-7850K is cheaper, faster and consumes less energy than having a traditional CPU + dGPU for GPGPU loads.

Okay, you've made your point. (And I am sorry for making the mistake of using gaming for a GPGPU comparison).

But do you think AMD should continue with the development of large iGPU on desktop, or are there better ways for the company to use their resources?

 

[cbn]

A10-7850K has 512 Steam Processors, R7 250 has 384.

So for GPGPU the A10-7850K is cheaper with lower power consumption AND Faster than Athlon + R7 250.

Actually the A10-7850K 512 stream processors are clocked lower at 720 MHz, and the R7 250 384 stream processors are clocked at up to 1050 Mhz (typically 1000 MHz).

http://www.tomshardware.com/reviews/a10-7850k-a8-7600-kaveri,3725.html

http://www.amd.com/en-us/products/graphics/desktop/r7#

So if going by stream processors x clockspeed, the R7 250 is actually slightly faster.

 

[Maxima1]

But do you think AMD should continue with the development of large iGPU on desktop, or are there better ways for the company to use their resources?

Intel is also increasing iGPU. HBM is around the corner. Discrete will be dead at some point.

 

[cbn]

Intel is also increasing iGPU.

Yes, and I think AMD should not do that for desktop.

Instead, I wish they would use the die size savings for more CPU. This coupled to some form of modest chipset integrated graphics.

Then I think they will be in a more efficient position to compete against Intel on desktop.

Laptop is a different issue, I see the iGPU as more of an asset in that scenario.

 

[mrmt]

Actually the A10-7850K 512 stream processors are clocked lower at 720 MHz, and the R7 250 384 stream processors are clocked at up to 1050 Mhz (typically 1000 MHz).

I think that iGPU vs discrete on GPGPU wasn't even a debate in the first place. We are not seeing clusters of iGPU chips on HPC machines, but we do see clusters using dGPUs. For all intents and purposes, dGPU trumps iGPU in GPGPU any time of the day.

 

[cbn]

For all intents and purposes, dGPU trumps iGPU in GPGPU any time of the day.

Yes, I believe that is true because the low end R7 250 dGPU that is being used a comparison to A10-7850K has a very poor performance to price ratio.

A larger discrete card would give much better performance for very little additional money.

 

[BSim500]

Prices from Amazon

A10-7850K = $153.00

Athlon 860K = $90.00

R7 250 = from $87.00

total = From $177.00

A10-7850K has 512 Steam Processors, R7 250 has 384.
So for GPGPU the A10-7850K is cheaper with lower power consumption AND Faster than Athlon + R7 250.

That has to be the trashiest and most desperate comparison of all time. You "forgot" to mention that for $17 less also on Amazon, you could get an R7 260 for $70 (after $20 rebate) with 768 stream processors (literally double the performance for $17 less), and save a further $10 with an X4 760K. Drop down to a G3258 for $70, and the $107 could buy a 260X / GTX 750 with 2.0-2.5x the performance and actual gaming playability at 1080p at something more than 25-30fps in most games (not to mention +70% higher GPGPU scores for 16% more money). Anyone pushing the "white elephant" R7 250 vs APU's in serious perf/$ budget build comparisons really is cherry picking to the point of total absurdity as it by far has the worst perf/$ ratio out of all of AMD's dGPU's and is about as "obscure niche" as you can get, even by "let's make APU's look relatively better by picking the absolute worst over-priced budget dGPU in existence" fanboy standards...

AMD needs to more than double the GFX performance of Carrizo's 7850K replacement just to meet the current gaming baseline of an XBox One and/or remain relevant to 2013/2014 budget dGPU's that are actually selling the most in popular-budget-build-lists over the past two years (7770 / 7790 / 260 / 260X / 750 / 750Ti range of cards). "Low end GFX performance" is a moving target. In 2 days time it'll be 2015, and your build will presumably have to last the builder for the next 2-3 years (through 2017-2018 because if you're that hard up, you won't be upgrading every year) and do so with 2011-2012 era equivalent budget dGPU horsepower that's actually far worse value for money for $150 vs a $70 R7 260 + $70-80 CPU.

 

[cbn]

every choice comes at the opportunity cost of some other choice. Spending generations getting APUs to where they provide tangible benefits to a tangible portion of the market is R&D budget, software budget, development time and actual silicon chip transistor budget that could have been spent on other goals (like entering the Phi/CUDA space more aggressively, or lower power CPUs, or more aggressive ARM adoption, or any other number of things).

Yes, that is how I feel about the situation.

Maybe there is some benefit on desktop to a very large iGPU, but the cost seems so ultra high relative to the very small gain that is being achieved.

 

[cbn]

Regarding a plan to remove Kaveri from FM2+ here is what I came up:

It's called S.T.O.P. F.M. ( System to openly promote FX motherboards)

high end FM2+ (A10) gets replaced with the lower end of AM3+ processors using integrated graphics chipsets.

Mid range FM2+ (ie, A8-7600) gets replaced by 35 watt Mobile Kaveri (maybe in some cases cTDP up could be applied to increase clockspeed of the mobile processor for certain niche desktop uses). In other cases, it might be AMD uses a lower TDP AM3+ processor at the 65 watt power level.

Low end FM2 gets replaced by higher TDP variants of Jaguar/Puma APU (2.4 Ghz quad core should work nicely).

Then FM2/FM2+ no longer needs to exist as the void can easily be filled in by other chips.

And Kaveri lives on as mobile processor only.

 

[redzo]

AMD needs to get back and compete with intel in desktop high end market wars at all costs. Otherwise they're screwed.
The magic of APU power marketing will not save them. Intel has the ability to stick cpu + gpu more economically onto the same die and it will screw them back.
HSA is irrelevant because mass adoption takes too long. It's the same phenom native quad core thingy crap all over again: nobody gives a crap if it has no real benefits.
They need to get their shit together and come up with a competing x86 core architecture fast.

And my god!!! They fail so hard even when they succeed. Their current offering is so competitive but they still don't gain anything off of that: amd mobile designs are so freaking hard to find! It's the famous and powerful athlon64 era all over again: athlon64 wipes the floor with intel P4, but magically everybody buys intel because AMD is too small and they cannot afford to pay money to sell more like intel did back then with their failed P4 core.

 

[Arachnotronic]

They need to get their shit together and come up with a competing x86 core architecture fast.

Easier said than done. If it were so easy, AMD would have caught up by now. It's been, what, 8 years since Core 2? Intel's R&D budget has only moved up and its process lead only widened, while AMD is a much smaller, much more resource constrained company.

I disagree that AMD needs to compete with Intel in the high end desktop market. This market is lost; AMD's best chance is to move outside of these markets, a strategy that AMD's management has signaled that it is doing. Whether it succeeds or not, only time will tell.

p.s. when Athlon 64 wiped the floor with P4, AMD was selling pretty much everything it could make. Opteron was also a hit because of the superior perf/watt and multi-socket scalability relative to the clunky solutions Intel had with the P4/Core 2 Xeons.

 

[ShintaiDK]

AMD needs to get back and compete with intel in desktop high end market wars at all costs. Otherwise they're screwed.

You better donate them some 10-20B$ and give them some 5-6 years and it _may_ happen with a good bit of luck included.

AMD needs to focus on the things they can earn money on, not what they would just lose money in an even faster rate at.

The only thing they currently can earn money on is semi custom.

 

[Phynaz]

We are talking about GPGPU specifically and you are bringing gaming in to the discussion ??

No, YOU are. We are talking about something different.

 

[el etro]

AMD needs urgently to deliver Carrizo, at least in CES 2015. Carrizo results can made AMD stocks going up, depending on the results. Carrizo needs too to be cheaper than Kaveri was, because Kaveri was a bunch overpriced at the launch.

 

[Arachnotronic]

Intel's IGP already has superior media, graphics will be next.

Intel has definitely talked up media capabilities at various technical events, and for most usages this is more important. I would note that Intel's Gen. 8 media is behind what many of the top mobile SoC vendors are currently offering (Qualcomm, MediaTek, Rockchip). Intel is doing better here than AMD (another reason why Temash/Mullins didn't do well commercially), but it still has work to do.

 

[redzo]

Easier said than done. If it were so easy, AMD would have caught up by now. It's been, what, 8 years since Core 2? Intel's R&D budget has only moved up and its process lead only widened, while AMD is a much smaller, much more resource constrained company.

I disagree that AMD needs to compete with Intel in the high end desktop market. This market is lost; AMD's best chance is to move outside of these markets, a strategy that AMD's management has signaled that it is doing. Whether it succeeds or not, only time will tell.

p.s. when Athlon 64 wiped the floor with P4, AMD was selling pretty much everything it could make. Opteron was also a hit because of the superior perf/watt and multi-socket scalability relative to the clunky solutions Intel had with the P4/Core 2 Xeons.

You better donate them some 10-20B$ and give them some 5-6 years and it _may_ happen with a good bit of luck included.

AMD needs to focus on the things they can earn money on, not what they would just lose money in an even faster rate at.

The only thing they currently can earn money on is semi custom.

Under no circumstances I condemn business decisions that keep one company floating on the short run.

High end x86 is where profitability is minimal or none, but it's the place where the backbone for a complete capable product gets born. I am more on the pessimistic side anyways. History proved that they cannot face the colossus even when they win it all. They need that win it all again otherwise there is no coming back. They will go under and under. Marketing people have nothing to sell. They need something big with pure pedigree. Only a pure x86 score can offer them that.

And something else about my thoughts. I strongly believe in the point of "singularity" when x86 will take over mobile arm. I think it is only a matter of time.

 

[cbn]

AMD needs to get back and compete with intel in desktop high end market wars at all costs.

I don't think that will happen any time soon.

But by making processors like Kaveri mobile only (not FM desktop), I am thinking they would be able to increase volume on the existing AM3+ processors and therefore lower prices.

An FX-6300 in six months from now at $69.99 would be better competition for the G3258 or whatever processor Intel eventually has as its replacement (eg, Maybe a fast stock clocked Skylake Pentium dual core, etc)

 

[ShintaiDK]

Under no circumstances I condemn business decisions that keep one company floating on the short run.

High end x86 is where profitability is minimal or none, but it's the place where the backbone for a complete capable product gets born. I am more on the pessimistic side anyways. History proved that they cannot face the colossus even when they win it all. They need that win it all again otherwise there is no coming back. They will go under and under. Marketing people have nothing to sell. They need something big with pure pedigree. Only a pure x86 score can offer them that.

And something else about my thoughts. I strongly believe in the point of "singularity" when x86 will take over mobile arm. I think it is only a matter of time.

Thats the sad part really. AMD being on the most likely candidate for the winning side (x86) bets the entire house on a 28nm ARM chip to emerge sometime in 2016. Zen may not even be anything at all anymore.

When you look on R&D budgets you notice why AMD doesnt stand a chance.

R&D budget per quarter 2007 vs 2014 (rough numbers).
AMD ~500M$ vs ~280M$
nVidia ~160M$ vs ~340M$
Qualcomm ~500M$ vs ~1300M$
Intel ~1400M$ vs 2800M$

AMDs R&D budget should have been around 1000M$ per quarter to just be with the crowd.

One of those 4 is a misfit. And the reason why they cant compete. No hopes or dreams gonna change that. In R&D you may not get what you pay for, but on the other hand you never get what you didnt pay for.

 

[cytg111]

Intel's IGP already has superior media, graphics will be next.

- And they already have superior throttling too, so yea graphics will problary be next... Oh snap.

 

[redzo]

An FX-6300 in six months from now at $69.99 would be better competition for the G3258 or whatever processor Intel eventually has as its replacement (eg, Maybe a fast stock clocked Skylake Pentium dual core, etc)

Pointless, absolutely pointless. AMD FX would be so irrelevant in terms of efficiency and performance in average joe's mind... That's exactly my point. How many people think that much when they make a purchase: only a few. Not mentioning business sales. Business sales are zero. Hell, it is getting hard even for myself to keep up with this cherry pickings.

 

[escrow4]

- And they already have superior throttling too, so yea graphics will problary be next... Oh snap.

Intel's iGPU may not be all that flash but a basic locked i5 or i7 with that iGPU up against an APU - who wouldn't take the extra CPU grunt every time? APUs are a tiny niche. Plus Intel constantly improves that iGPU year on year, and its not a 5% CPU IPC bump. Come Skylake low end 720p gameplay with some settings turned up shouldn't be a problem.