AMD CPU speculation... and expert conjecture

[Cazalan]

Steamroller is dead on arrival come january 14th.

Unless they have plans for releasing more powerful products in the future.

How is it DOA if Intels Broadwell has been delayed to Q3-2014?

Idk, about broadwell, but I heard from a insider that they're on path for 14nm coming 2014 (they're starting production).

Yes they started production but had yield issues. We're not expecting to see systems with those parts until the 2nd half of 2014 now.

 

[de5_Roy]

not directly related to SR, but it can have great impact on amd and gaming in the future:
Valve joins the Linux Foundation
http://techreport.com/news/25729/valve-joins-the-linux-foundation

OH MAI GAWD!!! I'm litterally laughing my butt off! I thought this whole GTX 780 Ti GHz thing was a joke from The Q6660 Inside... turns out, they're doing a GHz edition...

http://www.eteknix.com/gigabyte-gtx-780-ti-ghz-edition-specs-released/

kudos to amd for popularizing "ghz edition" branding.

Spoiler

Steamroller is dead on arrival come january 14th.

Unless they have plans for releasing more powerful products in the future.

How is it DOA if Intels Broadwell has been delayed to Q3-2014?

Idk, about broadwell, but I heard from a insider that they're on path for 14nm coming 2014 (they're starting production).

Yes they started production but had yield issues. We're not expecting to see systems with those parts until the 2nd half of 2014 now.

rumors are suggesting that broadwell dt might be bigger mess than haswell. the so-called k-series BRW socs will have iris pro gfx, lot like 14nm versions of core i7 4770R and i5 4670R. i expect them to be way out of budget range.

the reason i'm mentioning this is amd should not let this opportunity slip. they missed a chance to champion kaveri before, around the time haswell came out and induced universal groaning and yawning. they missed another chance to champion kabini and temash who knows why (at least jaguar uarch itself got some publicity and single-handedly boosted amd's x86 marketshare\o/).
intel isn't sitting still and will try to pit their own 14nm atom ulp socs vs mullins and puma.

 

[Cazalan]

Would AMD have stuck with 32nm SOI had they known GF would fail to produce another SOI node?

Sure, 28nm sounds like a good thing, but think about it this way. SOI offers lower heat and lower power. so does 28nm, but apparently not at the same rate. This leads to increasing ipc by x% and dropping the clock speed by y%. Is 32nm soi better than 28nm bulk?

Would be a good question if we get an AMA for the APUs. Could be the design tools were simply better for the bulk nodes as they have more customers using them. They're using million+ dollar software tools to design these things, but they're not perfect. TSMC on bulk as well so they get some reuse on design parameters/constraints between flows. GCN was done on bulk so they get time savings there by not having to redesign it for SOI. It all comes down to costs. Time is money and they let go about 30% of their staff while taking on new custom APUs and ARM designs.

 

[noob2222]

^^ that makes a lot of sense seeing as all their other "Radeon" designs are already 28nm. Just a quick copy and paste so to speak and since the majority of the die space is the gpu on kaveri, PS4, XB1, beema and mullins.

 

[juanrga]

@tracker45: continue dreaming...

@tourist: latency and bandwidth. Yes, Kaveri uses DDR3 memory. It was mentioned many times here and even speeds were given. The next quote is from this same page:

Kaveri with 2133MHz DDR3 (and possibility to use faster modules 2400, 2600...)

 

[juanrga]

Why AMD chose 28nm bulk for all the new products was explained before in this thread. In any case good to see people here rediscovering what was said about 100 pages ago.

 

[noob2222]

^^ referring back to your BSN(or wherever that was) article from the guy that got fired over a year ago, just before GF announced FD-SOI?

GF is the one that forced bulk only, not AMD's decision. 28nm GCN wouldn't be that hard to retro to a 28nm SOI process, but going to 32nm as calazan said would be time and cost. 28nm FD-SOI vanished with ST-Ericsson. Great, you win that arguement by default, not the reasons you gave.

 

[de5_Roy]

fear not d.i.y. amd lovers
"AMD sees signs of life in system builder market"
http://www.fudzilla.com/home/item/33318-amd-sees-signs-of-life-in-system-builder-market
...as long as you're in teh europes.

li'l bonus:
http://www.techpowerup.com/195425/asus-shows-off-radeon-r9-290x-directcu-ii-graphics-card.html
if kaveri is hiding a ninja dma bus inside it's u.n.b.... 😗......

 

[gamerk316]

After this generation, I expect we move toward Ray Casting.

That's prohibitively expensive but I can see us heading that way in a few generations. We'll need an order of magnitude increase in Vector processing power at the minimum before we can get real time rendering at anything resembling decent resolutions. But when it happens it's going to be beautiful.

I was talking generation as being 7-8 years, or in other words, capability limited by the PS4/XB1 GPU's. Assuming dGPU's keep gaining ~15% generation over generation, simplified Ray Casting algorithms should be runnable on consumer hardware about the time this generation starts coming to a close.

And yes, its going to be beautiful, since all those really hard light-based computations we currently do are basically free as a result of Ray Casting.

 

[8350rocks]

What do you do game(more FP heavy) or other things?

Like I said, I don't know what is biased towards integer performance in the Desktop world, hence the question.

Cheers!

Just to answer this question...

Compiling, rendering, encryption, and compression are all integer heavy tasks.

Hence the 8350 competes very strongly in those benchmarks.

 

[gamerk316]

@palladin9479: There are at least two companies, Intel and Nvidia, that disagree with your "APU's can't replace dedicated processors due to economics slamming into physics" mantra.

Intel and NVIDIA are market leaders, and thus have the necessary cash flow to support throwing money around to keep their performance edge. AMD doesn't, as we've noted many, many times. Different economic realities are in place here; what makes sense for one doesn't for the other.

 

[tracker45]

They have to release at least one normal Steamroller CPU.

 

[etayorius]

They have to release at least one normal Steamroller CPU.

Sadly there is not even a slight hint about that in any of AMD roadmaps... i wish they do, but most probably not...

 

[elemein]

They have to release at least one normal Steamroller CPU.

They have to release at least one normal Steamroller CPU.

Sadly there is not even a slight hint about that in any of AMD roadmaps... i wish they do, but most probably not...

They may have something like the Athlon; but that isnt really "normal" as it's missing the L3....

 

[tracker45]

If the prices drop on the 8350, people should just buy that instead of Steamroller.

The 8350 already competes with the 2500k.

 

[jdwii]

I feel that this thread is barely worth commenting on... It used to exciting and factual. Now it's just drama and insults. And worst of all, it's against other fans. We're arguing over speculation here... Not even actual benchmarks... I have no idea what is going on with you guys, but I'm not taking part in it. At least Hafijur picked a team and sided with it. lol.

If the prices drop on the 8350, people should just buy that instead of Steamroller.

The 8350 already competes with the 2500k.

I see this processor aging quite well but i don't see the platform aging well. With newer APU's coming out that really should beat the 8350fx in any program that uses 4 cores or less i can't recommend a dying platform. For me i like to upgrade if you already have a AM3+ board then i guess it makes sense if you're still on phenom but if you feel your processor might make until your ready to buy a new board+CPU(300$ for both maybe more for ram) and then wait.

 

[tracker45]

If FM2+ is backwards compatible with am3 CPUs, people should get the 8350.

 

[logainofhades]

We have AM3+ for a least another year, a Vishera refresh, and I don't expect much from Kaveri.

 

[de5_Roy]

If FM2+ is backwards compatible with am3 CPUs, people should get the 8350.

socket fm2+ is compatible with neither am3 cpus nor am3+ compatible cpus. fx8350 can't fit on an fm2+ motherboard.

 

[blackkstar]

You guys are realizing that for CPUs 2014 is basically going to be Kaveri and Vishera and a 14nm die shrink from Intel, where Intel has said there's no IPC increase, right? We might see Haswell-E at the end of 2014 but that's not going to be significant unless you're willing to drop a lot of money on a hex core as you can already buy Haswells.

Our best bet is to see how well Kaveri performs against Haswell and then decide if we think it'd be worthwhile for AMD to introduce a new high end platform around the time of Haswell-E and DDR4. I'm just running through some reviews and tacking on 20% for all the FX 8350 scores and comparing it to Haswell and it actually looks somewhat promising.

If AMD could deliver Excavator by that time and it was another 20% faster than Steamroller, that'd make 44% increase over Piledriver. Meaning Excavator enthusiast platform would be competing against Haswell-E. 44% increase in performance with 5 or 6 modules would actually crush Haswell-E. 50% more cores and 44% faster per core.

Then again, in regards to looking at the glass half empty or half full, I'd think that I'm probably looking at it like it's an olympic sized swimming pool full of the most delicious beer I can fathom with hot bitches surrounding me going "it's all yours blackstar-sama~~~~~"

But I suppose at the very least it's showing that it's possible to happen. Not saying it will of course. But if AMD sits on an architecture that could actually beat Haswell-E and then just messes around with APU HSA only I'm going to be so mad I'll probably just go buy some non-x86 computers and install Gentoo on them. Maybe by then I can sell off this gaming crap and buy some Power7 rigs for a reasonable rate.

 

[juanrga]

^^ referring back to your BSN(or wherever that was) article from the guy that got fired over a year ago, just before GF announced FD-SOI?

GF SOI plans were announced at 2011 (28nm was SOI)

http://www.nordichardware.com/Science-Technology/globalfoundries-discuss-future-nodes-committed-to-soi.html

AMD decision to ditch SOI by bulk was announced a year latter by Mark Papermaster

http://www.fudzilla.com/home/item/27572-amd%E2%80%99s-2013-to-rely-on-bulk-28nm-process

This same Mark Papermaster

http://www.amd.com/us/aboutamd/corporate-information/executives/Pages/mark-papermaster.aspx

The same Mark Papermaster who gave us the famous HotChip talk about Steamroller.

Glofo has now shown its unability to get ready the 28nm SOI process. A pair of months ago an enthusiast SOI poster here tried to convince all us that GF 28nm SOI was ready and that Kaveri delay was for the transition to SOI. It was all fake. AMD decision to migrate to bulk was a good one.

 

[juanrga]

@palladin9479: There are at least two companies, Intel and Nvidia, that disagree with your "APU's can't replace dedicated processors due to economics slamming into physics" mantra.

Intel and NVIDIA are market leaders, and thus have the necessary cash flow to support throwing money around to keep their performance edge. AMD doesn't, as we've noted many, many times. Different economic realities are in place here; what makes sense for one doesn't for the other.

Scientists and engineers working at AMD, Intel, and Nvidia are well aware that APUs are more powerful than a traditional dCPU + dGPU configuration. But whereas Intel and Nvidia have the money to push the research to develop the ultra-high-performance APU product, I am said that AMD has its 2011 plan parked due to lack of funds. The original AMD concept is a 8-core APU, with shared L3 cache with the iGPU, offering a total performance of ~10TF in a thermal envelope of 200-250W. The Nvidia design is also 8-core but offers >20TF.

 

[etayorius]

AMD willl probably release a faster and upgraded version of Kaveri in late 2014, and Excavator APU by the First Half of 2015, they done it before after Trinity with Richland and i see no reason for AMD releasing a 10% faster Steamroller APU.

 

[noob2222]

^^ referring back to your BSN(or wherever that was) article from the guy that got fired over a year ago, just before GF announced FD-SOI?

GF SOI plans were announced at 2011 (28nm was SOI)

http://www.nordichardware.com/Science-Technology/globalfoundries-discuss-future-nodes-committed-to-soi.html

AMD decision to ditch SOI by bulk was announced a year latter by Mark Papermaster

http://www.fudzilla.com/home/item/27572-amd%E2%80%99s-2013-to-rely-on-bulk-28nm-process

This same Mark Papermaster

http://www.amd.com/us/aboutamd/corporate-information/executives/Pages/mark-papermaster.aspx

The same Mark Papermaster who gave us the famous HotChip talk about Steamroller.

Glofo has now shown its unability to get ready the 28nm SOI process. A pair of months ago an enthusiast SOI poster here tried to convince all us that GF 28nm SOI was ready and that Kaveri delay was for the transition to SOI. It was all fake. AMD decision to migrate to bulk was a good one.

your first article NEVER ONCE MENTIONED 28NM SOI. I talks about 20nm SHP wich GF recently canned.

Your second article if you believe everything you read ...

Papermaster said that the company's cooperation with ARM is mainly to satisfy client demand for comprehensive functions and to allow quick product development, and is not founded to target any specific competitor.

only time will tell wether or not bulk was a good decision, a slow-down of 11% on the cpu and 15% on the GPU. It was a cheap decision as it allowed for fast implementation of GCN into all of AMD's APU products.

GF had no 28nm SOI until ST-Ericsson contracted them for their new phone chip that they wanted to produce with 28nm FD-SOI, not 20nm.

http://www.advancedsubstratenews.com/2013/01/st-ericssons-28nm-fd-soi-smartphonetablet-chip-at-vegas-a-great-start-to-2013/

For the folks designing smartphones and tablets (and ultimately for the end-user), that port to FD-SOI gets the NovaThor L8580:
•CPUs running 35% faster and GPU and multimedia accelerators running 20% faster
...
Wondering what’s next? The 14nm FD-SOI node is already in development, the ARM Cortex-A15‘s on the radar, and the FD-SOI roadmap is already defined up the 10nm node.

If you really don't think AMD would have taken advantage of that, your delusional. AMD would be stupid not to. Remember all those GF presentations on how great FD-SOI is going to be?

What actually came next was ST-E breaking up and GF saying fk it to SOI. This was after your papermaker was told there is no soi at 28nm, but would be implemented at 20nm by GF. Now it looks like GF isn't doing any SOI past 32nm.

. The ST-E chip was fully functional in February and operating at 2.8-3.0 ghz

 

[de5_Roy]

AMD willl probably release a faster and upgraded version of Kaveri in late 2014, and Excavator APU by the First Half of 2015, they done it before after Trinity with Richland and i see no reason for AMD releasing a 10% faster Steamroller APU.

quite possible. the reason richland came out is because amd had nothing to show for new products at that time. in the begining, richland was the codename for lower binned dual core trinity apus.

i personally would like to see amd take a stab at kaveri 1.5 with gddr5 dual imc in laptops.