So is the whole APU strategy of AMD a bust?

Discussion of all forms of processors, from AMD to Intel to VIA.

Moderators: Flying Fox, morphine

So is the whole APU strategy of AMD a bust?

Postposted on Sun Feb 23, 2014 3:11 pm

I was very excited years ago when AMD, after the buyout of ATI, announced their Fusion strategy. I thought it was a perfect idea of synergism between CPU and GPU firms. But after many years and the adoption of the term APU, it still seems kinda like a bust. Isn't the A10-7850 the latest and greatest APU from AMD but can't this still be beaten with a lower end CPU and lower end discrete card?

Am I being too pessimistic? Could the next generation of the APU be the one that finally takes off? Maybe the next process shrink from TSMC?
I dont think, therefore I am not.
WhatMeWorry
Gerbil
 
Posts: 81
Joined: Mon Apr 13, 2009 6:01 pm
Location: Illinois

Re: So is the whole APU strategy of AMD a bust?

Postposted on Sun Feb 23, 2014 3:33 pm

WhatMeWorry wrote:Am I being too pessimistic?


Yes
X2 3800+ @ 2.3, 2GB Mushkin pc3200, Lanparty Ultra-D, X850XT @ 560/585
jss21382
Gerbil Jedi
 
Posts: 1647
Joined: Wed Apr 30, 2003 8:20 am
Location: Grand Rapids, MI

Re: So is the whole APU strategy of AMD a bust?

Postposted on Sun Feb 23, 2014 3:50 pm

they are fairly poor atm but i still have hopes they might actually make something interesting
they can do it - look at the ps4 and xbone APU's, if something similar to those were released as PC APU's then they would be fairly useful
4 steamroller cores
around 800-1000 GCN cores
4GB onboard ddr5 256bit bus (i would like 8gb but maybe too expensive for budget cpu?)

then i would actually consider getting one for my htpc

im probably looking at this too simple but why cant they pretty much just get a 7850 and use its memory controller and put 4 steamroller cores on it and call it a day? get the motherboard manufacturers to solder 4GB of ddr5 onto it and done - perfect htpc

i do know its not that easy but up until now they have been designing the APU's around the cpu and adding GPU cores to it, why cant they do the opposite and design them around a mid range GPU like a 7850/7870 and add cpu cores to it using its memory controller instead of a cpu memory controller?

i would happily pay $500+ for that setup (motherboard/cpu/memory)
f0d
Gerbil
 
Posts: 70
Joined: Sun Jan 23, 2011 3:07 pm

Re: So is the whole APU strategy of AMD a bust?

Postposted on Sun Feb 23, 2014 3:56 pm

WhatMeWorry wrote:So is the whole APU strategy of AMD a bust?


Yep.

Its SSE5 all over again. No matter how good the strategy is, Intel can dictate what will be used because Intel sets the tone. Creating instructions for AMD's market, which is >20% of all cpu's is not very cost effective. Particularly as many of those machines are family machines outside of the workspace, as AMD has secured themselves the budget market. It isn't to say that it isn't a good theory and plan, because it is. It's software-dependent, and it simply isn't being coded for. If it continues to be moved through OpenCL, it will get better, but the biggest problem is that it is only going to make a dramatic difference in very specific areas. As a general processor, AMD is not keeping up with Intel. If AMD had another leap forward in technology, as they did with the Athlon64, we could see a change. But AMD is just not where it needs to be in efficiency or in raw power.
Sony a7
Sony Zeiss 55/1.8 SSM, 24-70/4 SSM
Minolta 100/2, 100-300 APO
TheEmrys
Minister of Gerbil Affairs
Silver subscriber
 
 
Posts: 2184
Joined: Wed May 29, 2002 8:22 pm
Location: Northern Colorado

Re: So is the whole APU strategy of AMD a bust?

Postposted on Sun Feb 23, 2014 4:03 pm

Even when it was announced, AMD took care to point out that it doesn't see a true unification of the GPU+CPU until sometime between 2016-2020. I haven't seen recent slides so I can only give the general numbers.

The potential of the APU is when the duplicate hardware is removed, and the strengths of both CPUs and GPUs are combined into a single mutlipurpose chip that doesn't require special drivers or OS support to properly utilize. Speaking in broad terms, the CPU can leave the FP units out of its design and rely on the wide FP capability of the GPU instead, and vice-versa the GPU can let the CPU handle integer work and anything else that it isn't suited for. Kaveri still is comprised of separate CPU+GPU cores that can only just now actually share memory/cache data, but each half of the APU is still a separate CPU/GPU as far as OS and drivers are concerned. There's still a ways to go before both sides of the APU can be truly integrated and that's where the true potential of an APU still lies. That was my understanding of the initial APU presentation, and probably the most optimistic one as well. :P
Kougar
Gerbil XP
 
Posts: 360
Joined: Tue Dec 02, 2008 2:12 am
Location: Texas

Re: So is the whole APU strategy of AMD a bust?

Postposted on Sun Feb 23, 2014 4:12 pm

Yeah, I'm beginning to think it's not such a good idea. You don't have to combine the CPU and GPU on one piece of silicon to realize the benefits of HSA or GPGPU computing. And it's not like the cost savings are significant either. Cooling also gets compliated as the cooler needs to dissipate heat from both chips and both the CPU and GPU are being held back because of cooling and TDP limits, among other things. Imagine combining an FX-8350 and an HD7970 on one chip: It will not only be prohibitively expensive to build, expensive to buy, but will will also be tough to cool. Speaking of being expensive to build, I'd say it could actually be more expensive than building a separate FX-8350 and an HD7970 because you can at least mix and match a good FX with a good 7970 the way Intel can mix two good dies back in the Core 2 days to build a C2Quad, whereas if Intel built them as one whole quad core die, a defect on one 'half' of the die will render the entire thing useless if Intel really wanted to sell the entire chip as a quad. What's so bad about putting them on separate chips anyway? As I've said, you can still do HSA and GPGPU or Mantle or whatever without insisting on putting both CPU and GPU on one piece of silicon. Besides, the manufacturing process can compromise or favor one part over the other as we've seen with Kaveri where the 28SHP process limits the CPU cores in favor of making the GPUs denser.

So yes, I think the whole idea of forcing CPU and GPU onto One Piece of silicon isn't really a very good idea. The only thing I can see these APUs being good for is HTPCs where the space savings could be useful and performance isn't on top of the list.
The three pillars of my digital life: AMD FX-8350, Google Nexus 7 (Qualcomm Snapdragon S4 Pro), Intel Core i5-2450M
ronch
Gerbil Elite
 
Posts: 663
Joined: Mon Apr 06, 2009 7:55 am
Location: C:\Program Files\

Re: So is the whole APU strategy of AMD a bust?

Postposted on Sun Feb 23, 2014 4:28 pm

I built a system for my parents using an A4-5300, for which I paid CAD 52. That's all a normal, casual user needs.

Try paying less with a discrete CPU and GPU.
Yan
Gerbil
 
Posts: 34
Joined: Fri Dec 21, 2012 9:37 pm
Location: Ottawa

Re: So is the whole APU strategy of AMD a bust?

Postposted on Sun Feb 23, 2014 4:40 pm

I think these chips are a bust when running on a 128 bit DDR3 bus.
As it stands, all those GPU cores/shaders/whatevertheheckthey'recalled are utterly strangled for bandwidth. With the current FM2x chips, I think AMD would have been better off reclaiming some of the silicon currently devoted to GPU resources and plonking in another CPU module, maybe two. That would have maintained at least rough x86 parity with Thuban/83xx chips, while including enough GPU power to be useful for OpenCL uses and light gaming. Add a cheap discrete GPU to upgrade. As delivered, I really don't understand how any 'gamer' worth his/her salt is supposed to be tempted by Kaveri.

If the next socket doesn't doesn't arrive soon and include either a stacked-onboard-DDR5 die or 2+gb of DDR5 memory (no 64-bit buses allowed either!) nailed to the Mobo, AMD will become utterly irrelevant.
Last edited by Geonerd on Sun Feb 23, 2014 9:10 pm, edited 1 time in total.
Geonerd
Gerbil
 
Posts: 67
Joined: Mon Dec 19, 2011 2:29 pm
Location: Sunny Aridzona

Re: So is the whole APU strategy of AMD a bust?

Postposted on Sun Feb 23, 2014 6:48 pm

I think many people commenting here are looking at the term "APU" too optimistically. IMO, "APU's" were implemented to save platform costs by bringing the integrated graphics on-die (see SoC) first and foremost. As a side effect: Until that point, if you were running off on-board graphics, it was an anemic chip, oftentimes with a tiny little passive heatsink or none at all. Integrating the graphics onto the CPU gives it a bigger (usually active) heatsink (more thermal headroom = more performance) and puts it closer to the CPU.

Also, Intel has had "APU's" since...Clarkdale in 2010? AMD just coined the term "APU" for their Llano chips (2011) with on-die graphics. Since graphics was brought on-die, there have obviously been evolutionary steps that have improved IGP performance (process node shrinks, architecture changes, etc). AMD saw that they couldn't beat Intel at CPU performance, so they dedicated more effort/die space to the GPU. Intel has (roughly) kept pace with AMDs IGP improvements while focusing on power efficiency.

That evolution has led us to where we are today as IGPs can actually begin to reasonably replace dGPU's for low-level 3D gaming needs (not to mention you no longer need a dGPU for Office-use computers, HTPCs, and the majority of consumer content). However, physically, a dGPU will be able to perform better than an IGP simply because of thermals. dGPUs have their own separate cooling solution as well as a boat-load of dedicated power input. The modularity of dGPUs is obviously a bonus as well. Although IGPs will continue to improve, it's not economically worthwhile to give every PC user the equivalent of a 7970 (for example) in IGP. It costs money to build, and people won't see the need to pay that much money for, say, Office-type PCs that will never use that much power. The level of IGP performance will always increase, but the top end of the graphics market will be occupied by dGPUs for the foreseeable future.
i5-3570K, ASRock Z77 Pro4-m, Asus GTX660 TOP, 120 GB Vertex 3 Max IOPS, 2 TB Samsung EcoGreen F4, 8GB G-Skill @1.25V, Silverstone PS07B
DPete27
Gerbil Jedi
Silver subscriber
 
 
Posts: 1679
Joined: Wed Jan 26, 2011 12:50 pm
Location: Madison, Wisconsin

Re: So is the whole APU strategy of AMD a bust?

Postposted on Sun Feb 23, 2014 8:30 pm

Geonerd wrote:I think these chips are a bust when running on a 128 bit DDR3 bus.
As it stands, all those GPU cores/shaders/whatevertheheckthey'recalled are utterly strangled for bandwidth. With the current FM2x chips, I think AMD would have been better off reclaiming some of the silicon currently devoted to GPU resources and plonking in another CPU module, maybe two. That would have maintained at least rough x86 parity with Thuban/83xx chips, while including enough GPU power to be useful for OpenCL uses and light gaming. Add a cheap discrete GPU to upgrade. As delivered, I really don't understand how any 'gamer' worth his/her salt is supposed to be tempted by Kaveri.

If they next socket doesn't doesn't arrive soon and include either a stacked-on-chip-DDR5 die or 2+gb of DDR5 memory (no 64-bit buses allowed either!) nailed to the Mobo, AMD will become utterly irrelevant.

This. Regardless of how powerful they make the on-die GPU, the system memory bus doesn't have enough bandwidth to allow for any reasonable expectation of gaming performance. If you do the math, it just isn't there; even with top-end DDR4 in a couple years. Compare the bandwidth of system RAM with even a mid-range discreet card, and the difference is staggering.

In order for AMD to exceed the performance of low-end discreet GPUs, they would need to either include some fast on-die RAM (ala the Xbox One), or embed GPU RAM onto motherboards. The first option would increase the cost of CPU production far too much, and the second option would significantly increase the cost of motherboards. I don't think either is realistic.

If AMD engineers are capable of doing basic math, you know they were aware of this from the beginning. That said, I doubt they were ever aiming very high on the performance scale.
i5 2500k - P67 - GTX660 - 840 Pro 256GB - Xonar Essence STX - Senn HD595's
The Egg
Gerbil Elite
 
Posts: 509
Joined: Sun Apr 06, 2008 4:46 pm

Re: So is the whole APU strategy of AMD a bust?

Postposted on Sun Feb 23, 2014 10:16 pm

APUs aren't trying to beat discrete GPUs, they're trying to beat integrated graphics. It's certainly true that right now APUs don't bring enough to the table to justify the tradeoffs, but that will start to change as the industry starts to move toward small form factors and mainstream HTPC.
NovusBogus
Gerbil Elite
 
Posts: 520
Joined: Sun Jan 06, 2013 12:37 am

Re: So is the whole APU strategy of AMD a bust?

Postposted on Sun Feb 23, 2014 10:58 pm

It's no bust. they would have been dead already if it weren't for APU's. and then again, could purchasing ATI, diverting R&D money to the APU/Fusion Program, be the real reason for them falling behind in such a way in traditional CPU architecture? who knows..
..
TaBoVilla
Gerbil Team Leader
Silver subscriber
 
 
Posts: 251
Joined: Sun Mar 27, 2005 2:37 am
Location: Madrid, Spain

Re: So is the whole APU strategy of AMD a bust?

Postposted on Sun Feb 23, 2014 11:15 pm

In term of non gaming application using the GPU, total bust.

In term of providing a low cost PC solution (desktop & laptop) its the only way.

And for the APU saved AMD... it might have, but it destroyed ATI value in the process.

ATI without AMD would have been a bigger force then nvidia. ATI at the time of the acquisition was moving full force in mobile SoC and GPGPU.
One initiative was completely killed by AMD giving Qualcom Adreno (rAdeon) all of the ATI IP, R&D team, and clients (motorola, sony, nokia,.. )
the other squandered opening the door for nvidia to catch up and take control of the market with cuda.

And result : AMD+ATI is a combination that billions in debt... ATI on its own would have billion in cash. AMD is an anchor with its x86 legacy IP.
sschaem
Gerbil Team Leader
 
Posts: 247
Joined: Tue Oct 02, 2007 11:05 am

Re: So is the whole APU strategy of AMD a bust?

Postposted on Mon Feb 24, 2014 12:07 am

TheEmrys wrote:Its SSE5 all over again.

Actually it is more like 3DNow! (which actually got implemented). SSE5 never even saw the light of day.

Aside from the issue of paying too much for ATI, I find it hard to fault AMD for deciding to pursue this strategy though. The back-to-back successes of Athlon and Athlon 64 proved that they could play Intel's game, and even innovate within the x86 ecosystem, pushing it in new directions. It was logical for them to try to continue the string of successes by trying something new.

Unfortunately, they woke the sleeping giant. Intel's edge in manufacturing, and the fact that they started firing on all cylinders again just when AMD was starting to falter, has ultimately relegated AMD to much the same role they had back in the K6-x days -- a second-tier vendor of parts for budget systems. At least they've got the console design wins now. And the Opterons (not sure how much of the server market they have these days, I guess it is down a fair bit).

I think AMD may have also been blindsided a bit by the massive shift to mobile, and the decline of traditional desktops/laptops. It'll be interesting to see how their ARM gamble pays off. I don't expect them to get huge penetration in the mobile device market, but with their expertise on the server side there's a chance they can push ARM into the datacenter in a big way. And all those mobile devices need datacenters to provide the services they rely on.
(this space intentionally left blank)
just brew it!
Administrator
Gold subscriber
 
 
Posts: 37843
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: So is the whole APU strategy of AMD a bust?

Postposted on Mon Feb 24, 2014 9:12 am

just brew it! wrote:I think AMD may have also been blindsided a bit by the massive shift to mobile

How one is blindsided by a massive cargo ship is beyond me :)
There is a fixed amount of intelligence on the planet, and the population keeps growing :(
morphine
Grand Admiral Gerbil
Silver subscriber
 
 
Posts: 10023
Joined: Fri Dec 27, 2002 8:51 pm
Location: Portugal (that's next to Spain)

Re: So is the whole APU strategy of AMD a bust?

Postposted on Mon Feb 24, 2014 9:59 am

I think there are two different things being talked about, and perhaps the confusion was mine. I had thought that AMD's APU strategy was to accelerate many things that are labor intensive for cpu's through an on-die APU, which also happened to be a GPU that handled the graphics duties.

By this, I would say that AMD's strategy for getting better-than-on-board gpu's has been a push, as the on-board gpu's were getting better.

As far as the Accelerating part of the APU, it just has not materialized.
Sony a7
Sony Zeiss 55/1.8 SSM, 24-70/4 SSM
Minolta 100/2, 100-300 APO
TheEmrys
Minister of Gerbil Affairs
Silver subscriber
 
 
Posts: 2184
Joined: Wed May 29, 2002 8:22 pm
Location: Northern Colorado

Re: So is the whole APU strategy of AMD a bust?

Postposted on Mon Feb 24, 2014 10:09 am

morphine wrote:
just brew it! wrote:I think AMD may have also been blindsided a bit by the massive shift to mobile

How one is blindsided by a massive cargo ship is beyond me :)


ARM is hardly a massive cargo ship.
The three pillars of my digital life: AMD FX-8350, Google Nexus 7 (Qualcomm Snapdragon S4 Pro), Intel Core i5-2450M
ronch
Gerbil Elite
 
Posts: 663
Joined: Mon Apr 06, 2009 7:55 am
Location: C:\Program Files\

Re: So is the whole APU strategy of AMD a bust?

Postposted on Mon Feb 24, 2014 10:12 am

I was making an analogy to the whole market, not just ARM. Whoever didn't see the shift to mobile must have been hibernating for the past few years.
There is a fixed amount of intelligence on the planet, and the population keeps growing :(
morphine
Grand Admiral Gerbil
Silver subscriber
 
 
Posts: 10023
Joined: Fri Dec 27, 2002 8:51 pm
Location: Portugal (that's next to Spain)

Re: So is the whole APU strategy of AMD a bust?

Postposted on Mon Feb 24, 2014 10:20 am

morphine wrote:I was making an analogy to the whole market, not just ARM. Whoever didn't see the shift to mobile must have been hibernating for the past few years.


Well, you gotta admit, ARM is at the forefront of this mobile craze. Ok, they're not a massive cargo ship. .. More like a squadron of tiny lifeboats. :-)
The three pillars of my digital life: AMD FX-8350, Google Nexus 7 (Qualcomm Snapdragon S4 Pro), Intel Core i5-2450M
ronch
Gerbil Elite
 
Posts: 663
Joined: Mon Apr 06, 2009 7:55 am
Location: C:\Program Files\

Re: So is the whole APU strategy of AMD a bust?

Postposted on Mon Feb 24, 2014 10:59 am

morphine wrote:
just brew it! wrote:I think AMD may have also been blindsided a bit by the massive shift to mobile

How one is blindsided by a massive cargo ship is beyond me :)

This is one of the biggest cases of technological miscalls in recent history. This is one of Hector Ruiz's biggest mistakes, not paying enough attention and focus on cpu and gpu ultra portable low power segments, .
TaBoVilla
Gerbil Team Leader
Silver subscriber
 
 
Posts: 251
Joined: Sun Mar 27, 2005 2:37 am
Location: Madrid, Spain

Re: So is the whole APU strategy of AMD a bust?

Postposted on Mon Feb 24, 2014 5:45 pm

DPete27 wrote:I think many people commenting here are looking at the term "APU" too optimistically. IMO, "APU's" were implemented to save platform costs by bringing the integrated graphics on-die (see SoC) first and foremost.


My viewpoint goes both ways, optimistically and efficiency. Part of bringing down the costs is to cut out parts of the CPU cores that the GPU can handle. This decreases the die area needed, while theoretically boosting processor performance. Unless I completely misunderstood the presentations that was the end goal of the APU strategy. Unfortunately it's still quite a ways off.
Kougar
Gerbil XP
 
Posts: 360
Joined: Tue Dec 02, 2008 2:12 am
Location: Texas

Re: So is the whole APU strategy of AMD a bust?

Postposted on Wed Mar 19, 2014 12:07 pm

For Intel the APU strategy has worked. AMD not so much.

As a consumer having a one chip solution for a desktop computer has provided very little benefit to me. For one, it is not a one chip solution. Motherboards still have an I/O chip and cost about the same as they did when the IGC were built into the chipset. A budget system still consists of a 100 dollar mother board and a 150 dollar processor, just as it has been for 20 years. The difference is that It is all one brand where in the past you could mix and match motherboards and processors. Intel wins here. They have eliminated the competition for its chipsets.

Where it does work for the consumer is the portable gaming laptop market. That is a laptop you don't have to plug into the wall to play games. Yes, you could have built a gaming laptop back in the CPU/GPU days, but it only runs for one hour while on batteries. In order for joe six pack to play games on his laptop you had to have Apple lead the way. Apple had to prove it was a market before Intel or AMD thought that it was good idea. AMD ended up the looser here despite having a product that is better suited to the market than Intel. The problem being they both lost to ARM based solutions. Now Intel looks to be ready for the future battle between ARM and x86 and AMD is not.
Dagwood
Gerbil XP
 
Posts: 497
Joined: Tue Oct 24, 2006 12:10 am

Re: So is the whole APU strategy of AMD a bust?

Postposted on Wed Mar 19, 2014 12:18 pm

I think that AMD's APU strategy in video game consoles has been pretty successful at least from a technological perspective.

The Kabini + next-generation Kabini parts from AMD are actually quite interesting since they occupy a price/performance niche where Intel basically doesn't compete. I like the idea of socketed Kabini quite a bit too.

Part of AMD's issues come with execution and the fact that they don't control their own fabs anymore. The TSMC-fabbed parts (consoles + Kabini) seem to be doing OK, but GloFo has not been kind to AMD on the product execution front.

All in all I wouldn't throw out the whole APU concept, but AMD needs to be careful in keeping expectations realistic about what the APUs are supposed to be doing and realize where they aren't really competitive (they ain't CPU powerhouses to be sure).
4770K @ 4.7 GHz; 32GB DDR3-2133; GTX-770; 512GB 840 Pro (2x); Fractal Define XL-R2; NZXT Kraken-X60
--Many thanks to the TR Forum for advice in getting it built.
chuckula
Gerbil Elite
Gold subscriber
 
 
Posts: 568
Joined: Wed Jan 23, 2008 9:18 pm
Location: Probably where I don't belong.

Re: So is the whole APU strategy of AMD a bust?

Postposted on Wed Mar 19, 2014 12:28 pm

Dagwood wrote:A budget system still consists of a 100 dollar mother board and a 150 dollar processor, just as it has been for 20 years.

I'd say that is really more low-mid-range once you get to those kinds of prices. You can put together the core of a budget system for substantially less than that, and the graphics performance will still blow any of the older IGPs out of the water.
(this space intentionally left blank)
just brew it!
Administrator
Gold subscriber
 
 
Posts: 37843
Joined: Tue Aug 20, 2002 10:51 pm
Location: Somewhere, having a beer

Re: So is the whole APU strategy of AMD a bust?

Postposted on Wed Mar 19, 2014 1:13 pm

just brew it! wrote:I'd say that is really more low-mid-range once you get to those kinds of prices.

Yeah... I'd say a budget/office use/non-gaming system would be a $70 processor on a <$70 mobo. To bring this back into the scope of this thread, "back in the day" it wasn't uncommon to see even budget systems with some sort of discrete GPU to handle day-to-day tasks. Nowadays, IGPs are good enough for nearly all consumer use and even some light gaming.
i5-3570K, ASRock Z77 Pro4-m, Asus GTX660 TOP, 120 GB Vertex 3 Max IOPS, 2 TB Samsung EcoGreen F4, 8GB G-Skill @1.25V, Silverstone PS07B
DPete27
Gerbil Jedi
Silver subscriber
 
 
Posts: 1679
Joined: Wed Jan 26, 2011 12:50 pm
Location: Madison, Wisconsin


Return to Processors

Who is online

Users browsing this forum: No registered users and 4 guests