New roadmaps out

N

NaroonGTX

Gawd
Joined
Jun 3, 2013
Messages
710
CtQ53Us.jpg

ARw8deG.jpg


Looks like no Steamroller FX in 2014 after all.
 
yeah that ROADMAP didn't tell me what I wanted to here but at least the FX isn't phased out it seems.
 
I think what's most interesting to note is that their 1P Opteron servers will have APUs rather than pure CPUs. Most likely by 2015 we'll see full APU lineups from AMD.
 
Yeah shitty news indeed. Guess all those rumors about a year ago of AMD conceding the performance CPU segment to Intel was right. Looks like they're going all in on the weaker APU market. I get why because that's where the vast majority of customers are but it's still pretty sad to see them abandoning our segment completely. What's worse is I wonder how this will effect us with Intel having no competition in our segment. AMD hasnt been much competition for them anyway but still an alternative and you still have to spend nearly $100 more to get an overclockable Intel chip vs an overclockable AMD chip. Wonder how things will change with AMD out of the picture completely.
 
I would rather have 22nm AM4 Excavators with quad channel DDR4 support and 6 module processors in 2015.
 
MacLeod said:
Yeah shitty news indeed. Guess all those rumors about a year ago of AMD conceding the performance CPU segment to Intel was right. Looks like they're going all in on the weaker APU market. I get why because that's where the vast majority of customers are but it's still pretty sad to see them abandoning our segment completely. What's worse is I wonder how this will effect us with Intel having no competition in our segment. AMD hasnt been much competition for them anyway but still an alternative and you still have to spend nearly $100 more to get an overclockable Intel chip vs an overclockable AMD chip. Wonder how things will change with AMD out of the picture completely.
Click to expand...

Again... who says an APU can't be a competitive enthusiast CPU? Yes, in its current forms it's aimed at mainstream, but that doesn't mean they can't get it to enthusiast class, especially when the software starts supporting it properly.
 
Tsumi said:
Again... who says an APU can't be a competitive enthusiast CPU? Yes, in its current forms it's aimed at mainstream, but that doesn't mean they can't get it to enthusiast class, especially when the software starts supporting it properly.
Click to expand...

That would be more of a miracle then you think it would be. What I am suspecting is with Mantle being the key to perfomance even with crossfire (APU+GPU) this will just be tolerable for mid end. More then likely FX8350 would still outperform it with the same videocard.
 
Pieter3dnow said:
That would be more of a miracle then you think it would be. What I am suspecting is with Mantle being the key to perfomance even with crossfire (APU+GPU) this will just be tolerable for mid end. More then likely FX8350 would still outperform it with the same videocard.
Click to expand...

You clearly don't understand the real purpose of the iGPU in HSA computing.
 
tsumi said:
Again... who says an APU can't be a competitive enthusiast CPU? Yes, in its current forms it's aimed at mainstream, but that doesn't mean they can't get it to enthusiast class, especially when the software starts supporting it properly.
Click to expand...

Im sure they could do that but I don't think they will. Their roadmap clearly shows Piledriver riding out the top tier segment thru 2014. The APU's are aimed at low power rigs with mainstream performance demands. You're right that they certainly could do something with them but I doubt it. Plus the prices would be higher because you're having to pay for an IGPU you're not gonna use and unless they're as fast as Intel chips, it would be pointless to add $30-40 to the price tag.

No, I don't see AMD giving a shit about the performance CPU segment, at least not thru 2014. Maybe Excavator will be different and they'll have something for us but I've been waiting since 2009 and if I'm told "just wait 1 more year" I'm going to stab myself in the eye with a spoon.
 
MacLeod said:
Im sure they could do that but I don't think they will. Their roadmap clearly shows Piledriver riding out the top tier segment thru 2014. The APU's are aimed at low power rigs with mainstream performance demands. You're right that they certainly could do something with them but I doubt it. Plus the prices would be higher because you're having to pay for an IGPU you're not gonna use and unless they're as fast as Intel chips, it would be pointless to add $30-40 to the price tag.

No, I don't see AMD giving a shit about the performance CPU segment, at least not thru 2014. Maybe Excavator will be different and they'll have something for us but I've been waiting since 2009 and if I'm told "just wait 1 more year" I'm going to stab myself in the eye with a spoon.
Click to expand...

You didn't seem to have read the second half of my post, nor do you seem to understand the purpose of the iGPU in HSA computing either.
 
Tsumi said:
You didn't seem to have read the second half of my post, nor do you seem to understand the purpose of the iGPU in HSA computing either.
Click to expand...

Yeah I'm obviously not as smart as you and maybe you're right that everybody on here will be selling their i7's for these APU's but I'm not putting any money on it.
 
MacLeod said:
Yeah I'm obviously not as smart as you and maybe you're right that everybody on here will be selling their i7's for these APU's but I'm not putting any money on it.
Click to expand...

Now you're just putting words in my mouth.
 
slightly tweaked Steamroller for desktop, performance and server on 20nm is listed for 2015. With desktop chips being released mid-late 2015.

Excavator is supposed to be the follow up sampling 3rd qtr of 2015, with a 20mn release in 2016.
 
Bummer. Was really holding out hope for a new FX line to emerge in the near future.
 
Mackintire said:
slightly tweaked Steamroller for desktop, performance and server on 20nm is listed for 2015. With desktop chips being released mid-late 2015.

Excavator is supposed to be the follow up sampling 3rd qtr of 2015, with a 20mn release in 2016.
Click to expand...

So... Piledriver throughout 2014? :(

And, Steamroller on 20nm in early 2015? Wow...

I wonder how much further can they tweak the Piledriver modules in another refresh? A lot of the improvements are already in Steamroller. It would have been great to see a Steamroller-based FX line in 2014.
 
In a perfect world, you'd see excavator in 2015. But many things have to executed flawlessly for that to occur. AMD's track record for perfect deployments is not exactly inspiring.
 
Well that was uninspiring. I was hoping for a new chip to come out for the desktop. I'm of course making the assumption that an APU isn't in the same league as a FX-9000 series. Well guess it's time to decide what I want to spend.

Thanks for the post!
 
In a perfect world, you'd see excavator in 2015. But many things have to executed flawlessly for that to occur. AMD's track record for perfect deployments is not exactly inspiring.
Click to expand...

The current AMD isn't the same one of yesteryear. Pretty much totally different across the board. Kaveri was pushed back a year, but if it means we get a much more efficient, better-performing part, then so be it. That meant that the EX-based APU got pushed back as well, and now it's scheduled for Q1 2015. I don't see any logical reason for it to somehow miss that date.
 
NaroonGTX said:
Lolwat? Where did you get this from? All sources point toward Carrizo releasing in Q1 2015. Then there's this: http://www.digitimes.com/news/a20130826PD216.html
Click to expand...

"In 2015, for the desktop market, AMD will release Carrizo-based APUs, featuring Excavator architecture with two power consumption specifications: 45W and 65W. The company will also release Nolan to replace Beema."

VR-Zone are the ones who said Carrizo is excavator.
 
Last edited:
I can't take AMD seriously.

I like how they have the X1150 and X2150 for all of 2013. Well, it's mid November and they're still not available. No thx.
 
Mackintire said:
"In 2015, for the desktop market, AMD will release Carrizo-based APUs, featuring Excavator architecture with two power consumption specifications: 45W and 65W. The company will also release Nolan to replace Beema."

Yes, a targeted late 4th QTR release is still a 2015 release.
Click to expand...

You claimed 20nm Steamroller for 2015... which didn't make sense in the least bit.
 
VR-Zone are the ones who said Carrizo is excavator.
Click to expand...

Excavator was only ever announced as being in an APU, and Carrizo being leaked as the codename makes sense. VR-Zone was on the money about just about everything else. Your original post makes no sense, it's like someone was just throwing random things out.

I can't take AMD seriously.

I like how they have the X1150 and X2150 for all of 2013. Well, it's mid November and they're still not available. No thx.
Click to expand...

I'm glad you made this post, because I just realized something. These roadmaps START at 2H 2013 and apparently END at 1H 2014. Richland clearly was not available until Summer 2013, and those Jaguar server parts are apparently still not released yet.

So This means that there might be a Kaveri refresh in 2H 2014. Still, I think Warsaw is meant for all of 2014 because I've never heard anything on the contrary.
 
Tsumi said:
You clearly don't understand the real purpose of the iGPU in HSA computing.
Click to expand...

Doesn't come any further then the relevant specs , which for Kaveri is not that much. To put it in perspective R9-290X is a good 5 times faster. That is why I mentioned Mantle .it allows Kaveri to contribute even with a seperate GPU.
 
Tsumi said:
You claimed 20nm Steamroller for 2015... which didn't make sense in the least bit.
Click to expand...

Not for me either, by then the am3+ chipset is so old it needs a walking cane and find vendors that will still update it with drivers for sound and other devices.

Does not make any financial sense either.
 
Pieter3dnow said:
Doesn't come any further then the relevant specs , which for Kaveri is not that much. To put it in perspective R9-290X is a good 5 times faster. That is why I mentioned Mantle .it allows Kaveri to contribute even with a seperate GPU.
Click to expand...

/facepalm

The purpose of the iGPU in HSA computing is to augment and accelerate CPU tasks. Yes, it can be used for graphics rendering, but that is not the goal of HSA.
 
KazeoHin said:
No steamroller FX planned...

vader.jpg
Click to expand...



lol right!!!

I would have liked to see a steam roller FX cpu as wells. Seems like that is not going to happen. Oh well my piledriver @ 5ghz will hum along just fine until there is a suitable replacement.
 
Tsumi said:
/facepalm

The purpose of the iGPU in HSA computing is to augment and accelerate CPU tasks. Yes, it can be used for graphics rendering, but that is not the goal of HSA.
Click to expand...

Yes but goals differ from actual real life performance. this is what you are talking about
http://semiaccurate.com/2013/10/21/amd-makes-gpu-comute-reality-hq/


What do you end up with? In APUs at least a CPU can work on a program for the serial portions and then simply pass it to the GPU in the middle of a running task. Nothing gets interrupted from the user perspective and the latency is minimal. Parallel portions can then run on the GPU and then it can pass the thread back to the CPU when the parallel portions are done. On pure GPU workloads the GPU can request things from the CPU directly, allocate a buffer, run this code on a texture at memory location ABC, and many other things. In theory it is exactly what AMD promised Fusion would do 3-4 years ago when Llano first popped on to the scene.
Click to expand...
.
 
Pieter3dnow said:
Yes but goals differ from actual real life performance. this is what you are talking about
http://semiaccurate.com/2013/10/21/amd-makes-gpu-comute-reality-hq/
Click to expand...

Did Llano have HUMA? Was it even based on the Bulldozer architecture? Are the HSA tools even widely available yet?

You should know by now that expectations and reality are two different things. AMD expected Bulldozer to be ready in 2009, and we all know how that turned out.

The reality is, Llano, Trinity, and Richland did not have HSA. They had the first steps, but Kaveri is the first fully enabled HSA APU. So to base how HSA will work out based on Llano is illogical at best.
 
You do realize that when things need to be implemented in software AMD ends up not getting the support?

Well Charlie was talking about promised hardware features with Llano. That Kaveri is HSA and to supports this means that 3rd party have to program for it to work. Also from what i understand Kaveri is the only chip that will be HSA

http://www.techradar.com/news/mobil...14-mobile-apu-lineup-1199201?src=rss&attr=all
One important note: Neither Beema nor Mullins include HSA features. AMD informed us that for 2014, HSA features are on the top of the performance stack with Kaveri.
Click to expand...

This does not make sense.
 
They've been pushing HSA rather aggressively, releasing dev friendly tools, working with Oracle to implement HSA features in Java, etc. Have you even looked at who is supporting the HSA foundation? AMD isn't the only ones, there's ARM, Samsung, Oracle, and more.

As for Beema/Mullins, there's a multitude of possible reasons.
1. It doesn't use Steamroller cores, so it might have not been as beneficial to implement HSA.
2. Those are used in low power devices without external GPUs, so the iGPU's primary job is going to be rendering graphics anyways.
3. HSA requires additional circuitry and adds to the costs and complexity of the chip.
4. Tied in with the above, it increases power consumption in a space where cutting half a watt means an extra hour+ of battery life.
5. Maybe they can't cram enough GCN cores in there to make HSA useful enough to justify the costs of implementing it.
6. Performance isn't a primary concern of what those are targeted at, efficiency is.

Not to mention, Kaveri and Berlin are essentially the trial runs of HSA. It only makes sense to introduce a feature in one product first to iron out the kinks before implementing it in your other products.
 
Pieter3dnow said:
You do realize that when things need to be implemented in software AMD ends up not getting the support?

Well Charlie was talking about promised hardware features with Llano. That Kaveri is HSA and to supports this means that 3rd party have to program for it to work. Also from what i understand Kaveri is the only chip that will be HSA

http://www.techradar.com/news/mobil...14-mobile-apu-lineup-1199201?src=rss&attr=all


This does not make sense.
Click to expand...
Hence why I'm hoping more developers pick up HSA and hUMA. It does seem pretty beneficial.

I'm also hoping maybe OS developers like Apple and Microsoft, and various Linux-based distros bake it into their operating systems to have native support for it. It's wishful thinking, but would help AMD push HSA forward.
 
The idea of HUMA an HSA is to mix up the flexible, serial nature of the X86 CPU with the sheer power of the parallel floating-point GPU architecture. Instead of having a 'gpu' operation and a 'cpu' operation with a combined 'apu' operation.

This has yet to happen, but with kaveri, it is one step closer.
 
Tsumi said:
Now you're just putting words in my mouth.
Click to expand...

You did a pretty good job of that all by yourself, Chief.

HSA/GPU/blah-blah are nice ideas, but the application base, pretty much non existent at the moment, will take years to be developed. (You do 'realize that,' don't you? :rolleyes: ) In the meantime, AMD needs a stronger x86 offering to hold the fort in the 'enthusiast' end of the mainstream range.

As for the Kaveri APU lineup, I'd gladly exchange a big slug of shading units for another 'Roller module. (Spend the transistor budget on something I can use, please!) Depending on IPC and clocks, 6 'Roller cores + modest GPU resources would at least be relevant, and wouldn't represent a drastic downgrade for Thuban and 'Dozer/'Driver owners. As things stand MANY people see ZERO reason to buy into an FM2+ system.
 
Tsumi said:
/facepalm

The purpose of the iGPU in HSA computing is to augment and accelerate CPU tasks. Yes, it can be used for graphics rendering, but that is not the goal of HSA.
Click to expand...

Show me the apps.
(crickets)
Show me the benchmarks.
(crickets)
Then (and ONLY then) will I sample a small cup of the Sickly Sweet Fruit Flavored Beverage you have been chugging with abandon.
 
Geonerd said:
You did a pretty good job of that all by yourself, Chief.

HSA/GPU/blah-blah are nice ideas, but the application base, pretty much non existent at the moment, will take years to be developed. (You do 'realize that,' don't you? :rolleyes: ) In the meantime, AMD needs a stronger x86 offering to hold the fort in the 'enthusiast' end of the mainstream range.

As for the Kaveri APU lineup, I'd gladly exchange a big slug of shading units for another 'Roller module. (Spend the transistor budget on something I can use, please!) Depending on IPC and clocks, 6 'Roller cores + modest GPU resources would at least be relevant, and wouldn't represent a drastic downgrade for Thuban and 'Dozer/'Driver owners. As things stand MANY people see ZERO reason to buy into an FM2+ system.
Click to expand...

I specifically stated in its current form it's not an enthusiast CPU. I also did not at any point recommend anyone to replace their i7s with FM2+ APUs. I stated that you cannot discount the idea of APUs as being enthusiast CPUs. So yes, you're the one putting words in my mouth because you didn't even bother reading my posts, and I had absolutely nothing to do with that.

Geonerd said:
Show me the apps.
(crickets)
Show me the benchmarks.
(crickets)
Then (and ONLY then) will I sample a small cup of the Sickly Sweet Fruit Flavored Beverage you have been chugging with abandon.
Click to expand...

Maybe you should actually try to read and understand my posts. Then maybe you won't say stupid things like this.
 
Everyone has the same train of thoughts that Geonerd has, show us then we believe but outright blind trust in software solutions by 3rd party is hard to swallow.
 
If the major OS's AND major s/w vendors AND OSS get on board with HSA, and implement it in such a way that it's utilized fully, it could, in theory, *scream* performance.
 
You must log in or register to reply here.
Back
Top