AMD CPU speculation... and expert conjecture

[randomizer]

We have had several requests for a sticky on AMD's yet to be released Steamroller 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.

Our previous sticky concerning Piledriver can be found here:

http://www.tomshardware.com/forum/343697-28-piledriver-rumours-expert-conjecture

Enjoy ...

 

[de5_Roy]

some steamroller related links.

AMD's Steamroller Detailed: 3rd Generation Bulldozer Core
http://www.anandtech.com/show/6201/amd-details-its-3rd-gen-steamroller-architecture

amd's cpu roadmap (from fx8150 review)
http://www.tomshardware.com/gallery/8-amd-roadmap,0101-310523-0-2-3-1-jpg-.html
AMD Talks Steamroller: 15% Improvement Over Piledriver
http://www.tomshardware.com/news/AMD-Steamroller-Piledriver-Kaveri-processors,17217.html
AMD Explains Advantages of High Density (Thin) Libraries
http://www.tomshardware.com/news/Steamroller-High_Density_Libraries-hot-chips-cpu-gpu,17218.html
AMD's Vision of the Future is All About ''Surround Computing''
http://www.tomshardware.com/news/amd-hot-chips-Surround-Computing-steamroller,17221.html

A brief look at AMD’s Steamroller core
http://semiaccurate.com/2012/09/06/a-brief-look-at-amds-steamroller-core/
More details about some new AMD cores
http://semiaccurate.com/2012/02/02/more-details-about-some-new-amd-cores/

AMD Pushes Steamroller and Excavator Forward, Bullish about Performance Increases
http://vr-zone.com/articles/amd-pushes-steamroller-and-excavator-forward-bullish-about-performance-increases/17088.html

AMD Hot Chips 2012
http://www.rage3d.com/articles/amd_hot_chips_2012/

AMD CTO reveals first Steamroller details
http://techreport.com/review/23485/amd-cto-reveals-first-steamroller-details

 

[mayankleoboy1]

point to note :

AMD is claiming a 30% gain (and i believe them) .
But a 30% gain over what ? Over Bulldozer/Piledriver.
So SR might just be competitive with Sandy's in single thread performance. And by then Haswell with 10% gain over IB will come to spoil the party. That is, if AMD can deliver SR in time......

 

[weaselman]

I do hope AMD Pull a rabbit out of there hat with this revision of the architectural silicone,
It will be interesting, Its always good to have competition in any sort of market.
we all know AMD has been a bit behind in the last few years. I can only hope this little baby in some ways will keep closer on par with Intel this time.
But we will see once the test or engineering samples are ready for testing.

 

[Chad Boga]

Well the big question now is when will it be released?

Everyone thought it would be out in 2013, yet AMD's latest roadmap shows that it won't be any earlier than 2014.

Interesting that Chuckles thinks that Steamroller being released is "likely", rather than a certainty.

From a previous post

From Donanim Haber

AMD's 2013 Roadmap

* Steamroller is completely MIA.
* Piledriver for AM3+ continues throughout 2013
* New 3rd Gen APU "Richland" in 2013 based on Piledriver and Radeon Core 2.0 (?). No word on what process it's being built on.
* Kabini/Jaguar appears to be on track

 

[fazers_on_stun]

point to note :

AMD is claiming a 30% gain (and i believe them) .
But a 30% gain over what ? Over Bulldozer/Piledriver.
So SR might just be competitive with Sandy's in single thread performance. And by then Haswell with 10% gain over IB will come to spoil the party. That is, if AMD can deliver SR in time......

According to the THG article linked to by De5_roy, as THG was told by Mark Papermaster CTO of AMD, some 2 months ago:

We are getting our first look at the "Steamroller”, which is the core for the "Kaveri" APU, among others. AMD is expecting to see a 15 percent improvement in performance per watt over the "Piledriver" core. The improvements are seen through design-level improvements rather than process-level improvements.

"Performance per watt" is not the same as raw max performance (basically IPC*clock frequency*cores or modules). Also, since Steamy is supposed to be on GF's 28nm node, then it is strange that no improvements are obtained from the die shrink.. Maybe the fact that GF is switching to strained silicon has something to do with it.

 

[mayankleoboy1]

^
I was building on the analysis of CharlieD and IIRC Anand.
It is semi wishful thinking, i accept.

 

[mayankleoboy1]

Well the big question now is when will it be released?

Everyone thought it would be out in 2013, yet AMD's latest roadmap shows that it won't be any earlier than 2014.

2014 ??

SR wont be competitive with anything , this late.
It just migh get beaten by ARM's next gen 64 bit arch

2014 is broadwell, IIRC. How the hell does AMD expect to be competitive with Intel's 14nm ? Even if intel has piss-poor design, thay can just throw moar transistors at it, till they succeed. 14nm FINFET is too big a lead over 28nm

 

[Chad Boga]

2014 ??

SR wont be competitive with anything , this late.

That might be why Chuckles was unsure if it would come out or not.

It just migh get beaten by ARM's next gen 64 bit arch

Perhaps on performance/watt, but not on outright performance.

2014 is broadwell, IIRC. How the hell does AMD expect to be competitive with Intel's 14nm ? Even if intel has piss-poor design, thay can just throw moar transistors at it, till they succeed. 14nm FINFET is too big a lead over 28nm

Well some(i.e. one ) have speculated that it might be impossible for Intel to release a performance oriented, billion+ transistor CPU for the desktop, perhaps that offers a glimmer of hope.

 

[bawchicawawa]

point to note :

AMD is claiming a 30% gain (and i believe them) .
But a 30% gain over what ? Over Bulldozer/Piledriver.
So SR might just be competitive with Sandy's in single thread performance. And by then Haswell with 10% gain over IB will come to spoil the party. That is, if AMD can deliver SR in time......

I think they said that during bulldozer(Correct me if I'm wrong), which is actually a 32.5% gain.

 

[kettu]

Apparently Steamroller will be compatible with socket am3+: http://www.theinquirer.net/inquirer/news/2208525/amd-sticks-with-socket-am3-for-steamroller]

I don't know how reliable that information is but good news if it's true.

 

[abbadon_34]

Sticking with the same socket is great for existing AMD users, i just hope it doesn't hold them back. I'd rather have an amazing CPU with a new socket then a mediocre CPU on the same. Same as all those moving to AMD to intel.

Wouldn't it be funny though if the new CPU was so great that the people who switched from AMD to Intel had to dust of their old AM3+ motherboards?

In another reality maybe

 

[mayankleoboy1]

i would count myself lucky if the socket is the only thing that is bad in Steamroller.

 

[sarinaide]

Even if AM3+ is the socket they need a new chipset, they are very far behind in that front and AM3+ as it is now is not a good platform.

Wasting money on flogging AM3+, I really don't understand this decision at all.

On Steamroller architecture, tightening up cache latencies, lowering branch mispredicts, improvements to IMC should see fair enough gains. If I was to be a betting man I think the requisite parts will be maybe 3-5% slower than SB/IB (FX8~i7/i5 and FX6 and FX4 should sit in between a i3 and i5 or match the mainsteam i5's) Where this puts them after Haswell maybe around 7-10% slower.

Kaveri APU's should be quite impressive though.

 

[abbadon_34]

reading the link posted, semiaccurate and vrzone give me hope, though i know it's just PR. I think their real achievement will in 2014-2015 (so they say) with a unified socket and graphics, like they're doing with Dual Graphics. They're been teasing that for years. That would finally make the ATI purchase come into bloom. It would become THE gaming CPU (unless intel response in kind). Combining whatever AMD CPU/APU you buy crossfiring whatever graphics card to you buy? Install advantage over Intel and Nvidia.

 

[bawchicawawa]

Even if AM3+ is the socket they need a new chipset, they are very far behind in that front and AM3+ as it is now is not a good platform.

Wasting money on flogging AM3+, I really don't understand this decision at all.

On Steamroller architecture, tightening up cache latencies, lowering branch mispredicts, improvements to IMC should see fair enough gains. If I was to be a betting man I think the requisite parts will be maybe 3-5% slower than SB/IB (FX8~i7/i5 and FX6 and FX4 should sit in between a i3 and i5 or match the mainsteam i5's) Where this puts them after Haswell maybe around 7-10% slower.

Kaveri APU's should be quite impressive though.

I'm generally curious.
Why is the am3+ platform outdated? Could you guys give me some specifics?

 

[de5_Roy]

looks like interlagos thread got destickied as well. so here it is -

AMD launches 16-core Abu Dhabi Opterons
Like Vishera before it, this one underwhelms
http://semiaccurate.com/2012/11/04/amd-launches-16-core-abu-dhabi-opterons/

 

[gamerk316]

reading the link posted, semiaccurate and vrzone give me hope, though i know it's just PR. I think their real achievement will in 2014-2015 (so they say) with a unified socket and graphics, like they're doing with Dual Graphics. They're been teasing that for years. That would finally make the ATI purchase come into bloom. It would become THE gaming CPU (unless intel response in kind). Combining whatever AMD CPU/APU you buy crossfiring whatever graphics card to you buy? Install advantage over Intel and Nvidia.

Not sold. Too many latency problems whenever you move to a dual GPU conflig, and more headaches then its worth in my mind.

 

[sarinaide]

To early to speculate on latency, that could very well be remedied somewhat by release.

They talk of improved and increased parallelism, its interesting.

 

[ProVisionOman]

I wish if they just dump AM3+ and Modules all together and go for the ultimate gamer setup which can be tied up with an AMD GPU. That would make it AMAZING just like abbadon_34 said. That would kill Intel and nVidia unless these two overpriced companies get together and form an ultimate gaming setup for 1000$ CPU and GPU only lol.

 

[-Fran-]

Not sold. Too many latency problems whenever you move to a dual GPU conflig, and more headaches then its worth in my mind.

Latency problems for a dual GPU is just about balancing loads per GPU. That's the trick behind it.

Once you're balancing loads taking into account some delays/latency within the sub-systems that feed and carry data, you can achieve very interesting stuff. Like I said in the IB thread, I've been quite impressed with the MVP software from Lucid, but its still infant. They need to polish a TON to make it really viable, but the gains are real and very nice.

I don't see why AMD can't do that when they control both parts of the spectrum, whereas Intel doesn't (discrete graphics). Lucid is currently tied to Z68 and Z77 chipsets AFAIK, but AMD could come up with something to partner up, I don't know.

Also, I don't know (or care) much about Steamroller per sé nowadays, but for the HSA initiative. I think what I'm seeing with Lucid on the Intel platform will come to us from HSA in the near future. At least, I hope so

Cheers!

 

[lilcinw]

Also, I don't know (or care) much about Steamroller per sé nowadays, but for the HSA initiative.
Cheers!

I agree. AMD has stated that they are no longer in the x86 performance race and I take that to mean that they can get more out of HSA than they would be able to from classic 'IPC' improvements going forward. If they can actually accomplish that on a reasonable timeline is the question.

I am very interested in how well the FirePro branded APUs will be received and I wonder if an Opteron branded APU is going to show up in the extended roadmap.

I would like to see SR do well enough to extend the viability of my AM3+ platform but if the release is not until 2014 that seems unlikely. Upgrading my 4100 to an 8320 for $170 was an easy decision, upgrading again may not be.

 

[kettu]

Even if AM3+ is the socket they need a new chipset, they are very far behind in that front and AM3+ as it is now is not a good platform.

Wasting money on flogging AM3+, I really don't understand this decision at all.

I don't think they're wasting money with AM3+, quite the contrary. All the validation work is allready done. Besides chipsets aren't that important anymore. Most of their important functionality either are or will be integrated on the CPU. I don't see any major flaws with the platform and I'm not expecting a platform update before DDR4 memory is ready for the consumer market.

On Steamroller architecture, tightening up cache latencies, lowering branch mispredicts, improvements to IMC should see fair enough gains. If I was to be a betting man I think the requisite parts will be maybe 3-5% slower than SB/IB (FX8~i7/i5 and FX6 and FX4 should sit in between a i3 and i5 or match the mainsteam i5's) Where this puts them after Haswell maybe around 7-10% slower.

I agree, those things should bring decent gains. I think the most important improvement will come from dedicated decode units for each integer core. On performance side of things I'd expect 8 core Steamroller to be significantly faster than SB/IB in highly parallel workloads. Afterall piledriver is pretty evenly matched allready. For lightly threaded workloads they might still be a little behind SB/IB.

 

[gamerk316]

Latency problems for a dual GPU is just about balancing loads per GPU. That's the trick behind it.

Once you're balancing loads taking into account some delays/latency within the sub-systems that feed and carry data, you can achieve very interesting stuff. Like I said in the IB thread, I've been quite impressed with the MVP software from Lucid, but its still infant. They need to polish a TON to make it really viable, but the gains are real and very nice.

I don't see why AMD can't do that when they control both parts of the spectrum, whereas Intel doesn't (discrete graphics). Lucid is currently tied to Z68 and Z77 chipsets AFAIK, but AMD could come up with something to partner up, I don't know.

Also, I don't know (or care) much about Steamroller per sé nowadays, but for the HSA initiative. I think what I'm seeing with Lucid on the Intel platform will come to us from HSA in the near future. At least, I hope so

Cheers!

Except they can't do this even with standard CF. Neither can NVIDIA with SLI. Alternate Frame Rendering is simply going to cause significant latency, by design (specifically, the delay caused by copying the buffer to the "primary" GPU).

http://techreport.com/review/21516/inside-the-second-a-new-look-at-game-benchmarking/5

So you'd need a totally different approach to multi-GPU to fix the latency issue. Using the APU opens up other problems as well (specifically: Where is the APU storing its frame data? If main memory, what happens if the data is moved to the HDD before its sent to the GPU?)

So yeah, I'll stick with a single, mid-top tier GPU, thanks. Hybrid solutions are good from a power management perspective, but not much else in my mind. And yes, I know I'm in the minority on this.

 

[noob2222]

That might be why Chuckles was unsure if it would come out or not.

It just migh get beaten by ARM's next gen 64 bit arch

Perhaps on performance/watt, but not on outright performance.

2014 is broadwell, IIRC. How the hell does AMD expect to be competitive with Intel's 14nm ? Even if intel has piss-poor design, thay can just throw moar transistors at it, till they succeed. 14nm FINFET is too big a lead over 28nm

Well some(i.e. one ) have speculated that it might be impossible for Intel to release a performance oriented, billion+ transistor CPU for the desktop, perhaps that offers a glimmer of hope.

man, don't ever speculate anything against Intel, you will face the wrath of chad for thinking his god-sent company might run into some issues, but its ok to bash the satanic company AMD all you want, they have scrapped their entire cpu division and are working on something unholy to chad.

we all know according to Intel's superior 22nm that Broadwell will be a 6 ghz part with overclocking to 9 ghz on air cooling because its going to be that much superior than anything that has ever been invented before. no other foundry in the world will ever produce a 14nm part because only Intel can.

There, does that make you feel better by praising everything Intel can possibly think of?

Even if AM3+ is the socket they need a new chipset, they are very far behind in that front and AM3+ as it is now is not a good platform.

Wasting money on flogging AM3+, I really don't understand this decision at all.

I don't understand this thinking. Whats so bad about am3+? most of the 990fx boards have 6+ sata 6gb ports, 4+ usb 3 ports, 42 pci-e lanes, CF & SLI support, all for around $150. What does an equivalen Intel board cost? Most of the intel boards force you to an x8/x8 pci-e configuration with just 2 cards, unless you move to spending $250+. http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=100007627%20600093977%20600315497%20600239779&IsNodeId=1&name=4

About the only difference is pci-e 3.0 ... wich nets you maybe 1% ... http://www.techpowerup.com/reviews/Intel/Ivy_Bridge_PCI-Express_Scaling/23.html

so ... just how bad is it not having pci-e 3.0?