Where is AMD heading?

[Enderman]

well they need to get more efficient and put out less heat, is what I'm saying

yes

all i see is intel doing less performance improvements and more efficiency per watt

amd is just trying to add as much power as they can to compete against intel, while creating a shit ton more heat

intel seems to be taking the right path, and soon will be working down to 14 or 10nm architecture

[Kamina]

Their focus is on the budget market. They try to give the best performance for an entry level processors at a good price.

 

Their enthusiast market still stands for GPU's, not so much for CPU's with their current architecture lacking in IPC. AMD is currently the leading movement of HSA, which is where we should be heading. They're a bit ahead of their time, just like their 8 core CPU but the market for 8 core gaming wasn't ready for it. It was a bad gamble for them.

 

They are going for profit to get out of their debt, and in 2015-18 you should see AMD making profit and possibly using that for better R&D. They won't compete with Intel in places where they know they cant win. They've learned that by now and they need to make changes to become a profitable company again.

[pixeldensity]

i don't blame amd for wanting to invest more in apus atm b/c of the financial status they are in right now. I think they rather hit a more general consumer market base rather than the enthusiast market right now b/c they can make more money from the general consumer market. We have to remember that not only does amd not have the size of R&D funds like intel does, but they are also fighting two battles at the same time (gpu and cpu). So I think that we won't see an enthusiast grade cpu for a while. Maybe they have a team trying to figure out an architecture to one up intel's cpus right now, but maybe that's hopeful thinking.

[Rolex]

I just built a pc for my buddy that had a R7 250x and it ran a Kavari A10 7850k @ 4.5Ghz and 2400mhz ram. I was impressed with the improvements they made. The processor runs about the same as my Home Entertainment Pc which has a phenomII 1065T 6 core. I remember running that pc and getting decent frames out of it with a single GTX 480. His build uses the dual graphics technology AMD APU's have. All together hes able to run assasins creed black flag on medium to high with over 40 FPS.

 

Where im getting at with this is, I believe AMD is heading in the right direction for their APU's adding much more performance aswell as utilizing many improvements to allow the GPU to help with Multitasking in conjuction with the CPU. As for enthusiast platform stuff. FX series is not on my plate. I run Intel in my PC for performance and low power. But overall I believe AMD is heading somewhere. I think they feel that mobile will be their win on gaming and sales. but as for enthusiast platform or performance platform chips. i believe they will continue to grow and build. but will always be behind intel. I feel they need to drop their current FX architecture and build something new. Build on single thread performance and with their 8 core line. They might just get somewhere in line with enthusiast platform intels. But AMD is budget friendly and they know that. I dont think you will see them go to expensive in the future. I think they will keep to what is cheaper and more cost efficent to them. but as for performance in the high end. Intel will always have that. In my opinon. 

 

But for the APU's im really happy to see them improve. They are on to something.

[Guest]

Even given AMD's strained financial status at the moment, the merits of their strategy are rather debatable. On one hand, AMD has never been this low for quite some time; on the other hand, they're still trying to expand into as many different markets as possible. AMD's CPU business is in shambles, GCN is hanging by a thread (that thread being closed-loop Asetek coolers and monstrous compute performance), Radeon memory was and still is a joke, Kaveri failed to impress, Opterons might as well be a thing of the past, and the mobile market is way too competitive and established for AMD. Had they spent the time to sit down and actually spend time to discuss and revise their CPU architectures, AMD wouldn't be where it is today.

 

With the launch of Beema low-power parts, people have been claiming these mobile parts to be AMD's saving grace. I most certainly don't see it that way, and the inevitable adoption of ARM's Cortex-A57/53, Nvidia's Denver cores, and Qualcomm's next in-house Krait successor may very well spell doom for AMD. Intel tried to make an incursion into the smartphone market with Medfield and failed, because Android was already too far into ARM architectures for x86 to make any real impact. Developers didn't see any real motivation to port libraries for the tiny x86 market share when ARM kept supplying ever-powerful cores (from the Cortex-A8 to dual-core in the A9, to quad core in the A15 and now in A57) and manufacturers simply kept on adopting those designs. AMD is making the same mistake that Intel made with Medfield; you can't expect to have any success if the major players in the market are already firmly planted on ARM's side of the field.

 

AMD had better support the enthusiast market. If they turn their back now, on everyone who has been hoping for a real successor to Phenom II since 2008, then they'll have chosen a path that will doom them and one that they can never go back on. APUs cater to a niche market, and are not an effective stopgap to FX. Neither is Beema, for which I see absolutely no purpose. Failing to refocus their attention on their desktop CPU and GPU development will allow their two largest competitors, Intel and AMD respectively to retake market share that will be extremely difficult to impossible for AMD to take back once it is done. Contrary to what AMD believes, ARM is the least of their worries.

[Joshua Ondangan]

That...would be awesome . But for the same price as the i7 4770k.

 

They may be developing stuff behind the scenes...although they're probably not :-(

well I do openly agree that I love intel although more expensive but the heat output is kinda different. It may not matter to the other people with snow on winter but on places like mine where summer (29c-33c room temp) winter (26c-28c room temp).

although I do have the bulldozer  and some other amd stuff I do to run benches whenever customers try to buy some. If only the 8150 bulldozer had the price tag of the i3 2100 back then it would have made a good choice for cheap workstation rig. alas their performance doesn't match their price tag. but the 8320 is kinda decent. built one for a friend and works nice but @ stock settings it was 44c on a bitfenix shinobi xl even with the tt frio adv cooler. for the gist of it i gave my friend d2 gt 5400rpm and temps went down a bit after giving it a mild oc @ 4.5ghz (72c @load prime95) (I filled his case with 5400rpm gentle typhoons) it's kinda cool compared to his room temp of 35c but his right now is one loud mother fucker. he ain't changing it anytime soon since those fans were my gift for him.

[Sharp_3yE]

Basically id have to say, they are going to focus on APU's much more, try to make them more energy efficient and make them faster. They will eventually focus on the mobile world as well I think. Intel is also focusing on mobile as well. They have been developing mobile chips and that is where everything is going. Tablets and phones are a huge market now. They will continue making faster and more efficient desktop processors but will also get into this market.

[Sonefiler]

OMG. MAYBE, MAYBE, since Graphene has just started being mass produced, maybe AMD is gonna make like 18nm CPUs/GPUs!

[Rolex]

@

Sonefiler Graphene will be super expensive. Also Im not sure it will be much use in processors graphene is a carbon material dies are made from silicon.

[KeltonDSMer]

Even given AMD's strained financial status at the moment, the merits of their strategy are rather debatable. On one hand, AMD has never been this low for quite some time

AMD's share price is actually double what it was two years ago.

 

GCN is anything but in shambles, it scales so well they are able to make use of identical architecture in the lowest performance parts all the way up to Hawaii.

 

Radeon RAM is silly, but AMD spends 0$ R&D on because they don't make or design any part of it; it is a rebrand of some other company's parts. It is just ilke SSDs...only a couple NAND makers and a few good controllers. Are Plextor SSDs a joke then? They are just Marvel controllers and Toshiba flash, not even custom firm-ware to set them apart. How about ADATA SSDs? They are litterally the same thing as the crucial M series with Micron NAND and Marvel controllers, just a different "brand".

 

Kaveri was not really a let down in my eyes. There can be no software support for something that does not exist, so HSA had to debut on the hardware side even if there is virtually no software to take advantage of it.

 

AMD and Nvidia are both at the mercy of TSMC for their GPU manufacturing. It is not their designs that determine what size of process node is used to make the chips, but what the foundry is actually able to produce with decent yields. 20nm has been delayed not because of AMD or Nvidia, but because of TSMC. Keep in mind though, in this situation AMD actually has an advantage over Nvidia because they used to be a foundry. AMD is more familiar with actually producing chips, so they are in a better position to weigh pros and cons of a given process and make subtle design tweaks to positively change the yields/performance of that process.

 

I don't think low power x86 is AMD saving grace either, but the Beema launch was not really a new part. Same architecture as last gen, same process node, yet they use far less power than last gen from taking advantage of a manufacturing process that is mature. They hugely reduced leakage and use software to maximize/minimize performance/power consumption to a point where they are very competitive with Intel's offerings. Where do these small low power chips matter though? Besides full windows tablets, I don't think they are very appealing. At least AMDs approach was just making something that already exist much better where as Intel is dumping billions in R&D into a market that is what I believe to be much smaller than what both AMD and Intel think it is.

 

AMD is not afraid of ARM, Intel is. AMD has licensed multiple ARM designs already...ARM is one of the big names in the HSA foundation, along with AMD, Qualcom, Samsung, Texas Instruments, Mediatek, the list goes on and on. Beema even has an ARM core for security functions crammed in there. AMD is going to push ARM into markets it has been unable to penetrate in the past, they already are rolling out ARM/x86 servers.

 

I also agree that AMD must not abandon the HPDT or FX and that current APUs are not a stopgap. AMD needs enthusiest support to succeed in the DT market.

 

Think about how long a design on paper takes to manifest into an actual product that people can buy, well over a year. AMD had to flip their roadmaps from two years ago upside down and only focus on a handful of products. I feel the execution was as good as anyone could ask for and in the mean time have hopefully been very deliberate in how they manage their tiny R&D budget.

 

AMD just days ago rehashed their wafer agreement with GloFo for hundreds of millions more in purchases, over 1billion in total for the year. I know GloFo has had trouble, but they were AMD before the spinoff in 2009. Intel may have the fabs, but are also losing huge money on keeping them open yet not running close to max capacity. I say this only to show AMDs unique position. They understand the fabricaion processes better than any company without their own foundry, yet don't carry the same financial liability to keep the fabs running.

 

Sorry for jumping all over the place with all these topics. @tabascosauz I really enjoyed reading your post. Although I don't agree with everything you mentioned, I appreciate your views and hear what you are saying.

tabascosauz

[Guest]

@KeltonDSMer Thanks for the thoughtful input.

 

What I meant was that Kaveri made extremely minimal advancements in CPU technology and was more like a step backwards or staying stationary for AMD. I realize that with Kaveri's R7 GCN graphics, it's finally possible to do some real gaming on an APU alone, and now FM2+ thin-ITX is actually a very real option, because for a lot of budget gamers out there, there's no need to buy a Athlon II X4 760K and have to pair that with a discrete GPU; Kaveri does it all, though there's a lot of room for improvement.

 

I do agree that AMD is still in quite a good position to give Nvidia a run for its money. Nvidia's lack of experience shows in the roller coaster that is their Tegra lineup, and I think Nvidia really needs to think about the future of its mobile SoCs, because Tegra 3 was the closest taste of success for Nvidia. The green camp has had their ups and downs over the years and they are really neck-to-neck. The showdown between Maxwell and GCN (2.0?) will be an interesting one.

 

I think that AMD has been associated with the word "budget" in a way that the company hadn't expected to. After AMD realized the true extent of SNB/IVB's invincibility at the time, they turned to the budget market because that's the only market that their Bulldozer and Piledriver (to an extent) designs could be sold as. However, I think that their plans have worked against them in that it's not entirely possible to build a high-end all-AMD rig because board manufacturers have been swayed by Intel. Most new board designs/features now debut on Intel boards and if budget/popularity allows, then those technologies will trickle down to AMD boards. I also think that despite PCIe 3.0 x16 not having any real, tangible benefits over PCIe 2.0 x16, AMD should probably take note and up their game in that respect. Even though Piledriver and Steamroller have all but erased the painful memory of Bulldozer, AMD's chipsets and boards really need work. Perhaps AMD will have a breakthrough in the coming months / quarters and produce a design that will win over the important AIB manufacturers once again.

 

But again, budget is the key here. AMD wins hands down. I am seriously considering a 760K + 265 build this summer over a 4670K/4690K + 265 because of the price. With the i5 I will have to resort to re-using a NH-U9B, which will limit me to a 4.1-4.3GHz OC, not that Haswell did very well to begin with. A 760K will allow me to slip in a H55, small SSD, and still save me roughly $150. And this is all in a mITX form factor.

[Guest]

[hawthorne]

I HOPE they improve the cooling and efficiency of their graphics cards to make sure they don't run at 90C and don't sound like a hoover.

[Guest]

I HOPE they improve the cooling and efficiency of their graphics cards to make sure they don't run at 90C and don't sound like a hoover.

 

AMD just needs to learn how to make reference designs and use turbo boost to keep cards at 80°. Maybe they should go to a Nvidia seminar to learn all about NVTTM.

[SSOB]

AMD is slowly killing of the performance division. It will be a buget brand for buget PC's and integrated solutions most likely. Intel is Light years ahead of it's CPU game.

[Cubit481]

i think amd is heading to catch up with intel...

[Humbug]

It doesn't look like they are going to have a real answer for Maxwell and unless the prices drop heavily, I don't see how they are going to stay competitive on the discrete graphics market.

They have stayed competitive in the discrete graphics market for many years and that is not going to change.

For now High end maxwell chips are just some phantom product that people keep fantasising about. In reality both nvidia and AMD are having issues with 20nm transition. If they do release before AMD it will only be a couple of months before AMD responds with their Bermuda architecture. Do you really think AMD who has been improving their GPUs consistently for years is going to suddenly lose the ability to compete with nvidia? It's just that their next gen is not as hyped as maxwell... Nvidia doesn't really overhype maxwell but the press and fanboys do. The only maxwell card released so far is slower than competing AMD cards, with it's only selling point being power consumption...

AMD faces bigger problems on the CPU front.

[Guest]

^I agree. I can't even count all the rumors floating around the web like GTX 750 Ti > GTX 660, 750 Ti = 660 + 192-bit GDDR5, etc. As a result, 750 Ti was a complete disappointment compared to even the humble rebrand that is the R7 265.

 

AMD's next architecture hasn't even been released and people are making conclusions about AMD's ability to strike back on the GPU front. GCN may be old, power-hungry and hot, but there's no reason that Bermuda won't come with AMD's optimizations on the power and temperature front.

[nosirrah123]

less heat? sorry but as far as i know intel cpu´s nowdays run alot hotter then AMD FX cpu´s lol

So why do they need to put out less heat?

??? AMD's processors are HUGELY hot according to everything I've heard.

[TastyRabbit]

I made quite a lot of money off amd's share jump last year. So I think I can add a comment to this. 

 

I don't really see a short term future for AMD currently but in the next 36 months the outlook is  good. Remember back to September-ish (I'm too lazy to look for the figures) when they actually posted profits for the quarter. This quarter the expected earnings are 0.02. Basically, what I'm getting at, is AMD are actually becoming profitable.

 

People who say Beema and Mullins are a bad move, I think are wrong. I think most people understand that the limited nature of tablets is the inability to share stuff across platforms. They use tablets for convenience of size and portability. 90% of users just want to surf the internet etc, and use Office. I think that's why APU's are important. Why I would back AMD over Intel on this is because AMD will do it on a budget. I may be an enthusiast, but most people aren't. I believe market sentiment is leaving high end produces of phones and tablets and moving to value solutions that offer the same abilities. Most people are starting to change from brands, to buying on specification and usability. Its quite easy to see this with high quality Chinese no brands (or Import rebrands) entering our markets from everything from Motorbikes to phones and doing very well. AMD's integration of CPU and GPU will, I would bet (once integration begin's) be likely to be very good and cheap.  

 

What I would love to see from AMD - for myself - is a total redesign of the PC market. I think they could actually pull this off, at massive risk.

 

Redesign of motherboards with a CPU socket and a GPU socket, with shared RAM (DDR3-4) and maybe a DDR5 slot for GPU.

 

I would buy that!!! It would make upgrading so cheap, and you could keep all the same waterblocks - coolers etc.  

 

I should post - I actually use an Intel CPU myself - but I'm not the market norm, nor are you if your reading this!!! 

[the ginger gamer]

I'm not sure, probably an apu future.But i think that if they were smart they would come out with a new fx series, keep improving the apu's, and continuing making their gpu's. However they should work on lower tdp and much cooler components because those are their two biggest issues. Also i don't really see them not making the a newer fx series because as long as there is a  want for budget grade cpu's then they will keep making them. Also if you look at the game debate charts the for the top ten most used cpu's most of them are actually amd such as the fx 6300 which placed second most popular. 

[Humbug]

However they should work on lower tdp

One of the reasons that AMD and Nvidia are always neck and neck in terms of GPUs is because they are on the same process node.

 

But when it comes to CPUs it's a different story as Intel has a fabrication advantage in terms of the process node, I don't see how AMD can overcome that. However much they improve their architecture they are handicapped...

[angrylemon]

Hey forum, 

 

just wondering, where do you think AMD will be heading in the future? I mean, with Maxwell around the corner on the GPU market and Intel having been ahead on the CPU market for quite a while now, can we expect AMD to continue producing enthusiast-grade components in the future, at all? 

 

It doesn't look like they are going to have a real answer for Maxwell and unless the prices drop heavily, I don't see how they are going to stay competitive on the discrete graphics market. And the CPU market? Well, AM3+ has been there for quite a while and the FX-series hasn't seen any significant improvements either.

 

To me it looks like AMD is planning to focus entirely on the APU and mobile market, pretty much excluding enthusiast grade hardware from their budget... 

 

I really hope they don't, but what do you think?

not the right way...

[Guest]

I made quite a lot of money off amd's share jump last year. So I think I can add a comment to this. 

 

I don't really see a short term future for AMD currently but in the next 36 months the outlook is  good. Remember back to September-ish (I'm too lazy to look for the figures) when they actually posted profits for the quarter. This quarter the expected earnings are 0.02. Basically, what I'm getting at, is AMD are actually becoming profitable.

 

People who say Beema and Mullins are a bad move, I think are wrong. I think most people understand that the limited nature of tablets is the inability to share stuff across platforms. They use tablets for convenience of size and portability. 90% of users just want to surf the internet etc, and use Office. I think that's why APU's are important. Why I would back AMD over Intel on this is because AMD will do it on a budget. I may be an enthusiast, but most people aren't. I believe market sentiment is leaving high end produces of phones and tablets and moving to value solutions that offer the same abilities. Most people are starting to change from brands, to buying on specification and usability. Its quite easy to see this with high quality Chinese no brands (or Import rebrands) entering our markets from everything from Motorbikes to phones and doing very well. AMD's integration of CPU and GPU will, I would bet (once integration begin's) be likely to be very good and cheap.  

 

What I would love to see from AMD - for myself - is a total redesign of the PC market. I think they could actually pull this off, at massive risk.

 

Redesign of motherboards with a CPU socket and a GPU socket, with shared RAM (DDR3-4) and maybe a DDR5 slot for GPU.

 

I would buy that!!! It would make upgrading so cheap, and you could keep all the same waterblocks - coolers etc.  

 

I should post - I actually use an Intel CPU myself - but I'm not the market norm, nor are you if your reading this!!! 

 

I see what you're getting at, and I can see why AMD chose to take that path. And AMD has already proven that APUs have one strong characteristic that allows them to provide heightened performance with cheaper/lesser specs on paper: their shared memory. The Xbone lacks what the PS4 has, a shared pool of GDDR5 memory that in theory should completely or partly mitigate the potential bottleneck of the weak Jaguar cores. And in the mobile world, this is essential. Currently, memory bandwidth is pitifully anemic, with Snapdragon 800 barely reaching 12.8 GB/s theoretical bandwidth. However, even with 25.6 GB/s, on par with most dual channel DDR3 desktops, there are other factors to take into account. CPU performance and GPU performance both need a boost, though all the major players are making healthy strides towards better low-power computing. Furthermore, that memory is often overhyped and oversimplified by manufacturers; 4GB of RAM doesn't really mean a thing if it's single channel, low-clocked LPDDR2.

 

Obviously, Mullins has yet to prove itself. And we will hold off our performance conclusions until that happens. But the major issue at hand lies not in AMD's ability to create a SoC, it's in AMD's ability to create an SoC that will win designs. ARM may be far behind x86 in maturity and raw power (hence why desktop enthusiasts like us point at ARM and Apple's claim of "desktop-class performance" in Cyclone/Cortex-A57 and we will simply ROFL all day long), but x86 can't touch ARM in the mobile space. With Medfield it was because Intel chose a sh*tty-ass GPU to go along with a single 32nm Saltwell core, and that design sent the world laughing and Intel's designers back to the drawing table. We're not talking about Windows tablets here. Tablets, specifically iPad and many others such as Galaxy Tab/Note and Asus TF Pad/Infinity, run ARM and as a result, ARM-optimized, ARM-prioritized software that is exceedingly time-consuming and difficult to port to x86. That's why Intel failed and continues to fail to make an impact in the Android space, no matter how great Bay Trail/Silvermont is. AMD has shown us that it can use a small ARM design in its security core in its new ULP APUs; this doesn't mean that ARM suddenly became "friends" with AMD and AMD has a free ride to success. How do you think Samsung will react to this potential competitor to its new Exynos Hexa/Octa designs? How do you think Apple will react to this? What about Qualcomm, which is at the moment struggling to develop an effective high-end 64-bit design in time? Maybe LG, which could possibly manufacture its own rumored "Odin" SoC? There's much to consider in the development, in these companies which could not compete with AMD in the traditional desktop/mobile space, but which all have more experience than ARM in the mobile space.

 

That being said, it's probably best to treat Mullins as a proof-of-concept, but AMD certainly doesn't look to be content with that. AMD plans for it to be an all-out effort; AMD wants success. Which leads me to my last point, AMD needs to stop overhyping its own products. Bulldozer was the victim of the enthusiast community and that didn't count, but was the countdown to the "next big thing" really necessary for Mullins? Oh, so AMD wants to get BACK into the mobile space. So why did you sell Imageon in the first place?

 

EDIT: I read your comment about motherboard redesign and I don't think that will happen. One can't simply buy GDDR5 and stick it in there for the GPU; it's called graphics-DDR5 for a reason. It's not on a stick. Not to mention that the GPU performance will vary wildly with what memory the user has. Plus, DDR2/3/4 all use different pin layouts so it wouldn't be possible unless we are able to take individual NAND packages and use them in a modular sort of way.

[DroPouT1999]

I think they're working on something, just very very secretly otherwise...