Should AMD release an updated 8 core FX (improved cores, smaller process)?

[superstition]

Should AMD not keep its brand loyalists waiting for so long for Zen, now rumored to be pushed until Q1 2017 according to some sources — by releasing a 28nm 8 core FX with the improved design from its FM2 APU line?

Zen is also going to require a new motherboard and RAM. For people with 16 GB of RAM and a high-end AM3+ motherboard like a Sabertooth or a Crosshair it may not be tempting enough by the time it's released.

DDR4 is only really impressive for servers given the latency price for its lower power consumption. Sure, AM3+ is old and it would be nice to have a smaller process node for the motherboards but it's also a lot cheaper (free) for people who have one and who are tired of 2012 CPU tech.

Or, should such a chip be released on FM2?

Thoughts from the topic that inspired me to create the poll:

It'd be interesting to see a Steamroller (and Excavator) part that used the originally planed manufacturing process that could reach ~5GHz, up to eight cores, and 8MB of L3 cache like the FX CPU's have.

You will never see that on the FM2 platform, everything AMD is putting money into now is budget, mobile, onboard GPU and power savings. The FX chip stuff is going the way of the dinosaur. They gambled on multicore years ago and realized that was a horrible investment.

1) They most likely want as big of a gap as possible between 8 core FX and Zen, to add more excitement to the launch and get more people to open their wallets for new motherboards and such.

2) Their current CPU people are probably not the same people who favored the CMT SOI design philosophy. So, instead of making FX look better, they'd rather just have it look worse by leaving it with 2012 design and 32nm all the way up until Zen finally is released.

The problem with these strategies is that Intel is selling stuff at a greater rate during all that time and people may not be willing to buy new motherboards and RAM once Zen finally does ship — if it's not really better than what Intel is offering.

There is a small market for AM3+, at least, 8 core FX chips with an improved process node and the improvements to the chip design that have been rolled out on FM2.

I think it might add even more brand loyalty to give AM3+ one more minor upgrade option (28nm FX 8 core with the improved core design) since Zen is going to take forever. It's not like DDR4 is really an advantage for anything but power consumption which is only an issue for servers. Such an 8 core FX would also make AMD look better after the 9590.

Whether or not it's a good idea to roll out a better 8 core FX on FM2 or not I can't say. I doubt it, though, because the boards are probably not made for that level of heat and power consumption. But, I don't know much about FM2 boards.

 

[superstition]

This was supposed to be a poll but the poll options didn't stick, apparently. I guess it's due to NoScript (more trouble than it's worth).

update: got the poll working

 

[ShintaiDK]

Ask yourself how many they even sell and can sell.

AMD already answered with no, there isn't any ROI in it.

 

[superstition]

Ask yourself how many they even sell and can sell.

How viable is 2012 tech in 2015, 2016?

Can AMD afford to keep asking people to wait and then turn around and ask them to replace their RAM and their motherboards?

Can AMD afford to have the 9590 be its top processor without losing even more mindshare?

 

[dark zero]

No, AMD actually is dead and needs to stay dead with their fake Quad Cores who are Dual.Cores in fact.

 

[ShintaiDK]

How viable is 2012 tech in 2015, 2016?

Can AMD afford to keep asking people to wait and then turn around and ask them to replace their RAM and their motherboards?

Why would anyone buy an updated FX when its performance is still horrible behind?

 

[dark zero]

Let the dead stay dead and if is AMD, spit on their dirty gravestone.

 

[superstition]

Why would anyone buy an updated FX when its performance is still horrible behind? And on an ancient outdated platform as well.

1) People already have invested significant money into their current motherboards, some of them even have water cooling to go with their Sabertooth or Crosshair setups. Are they going to want to wait even longer for Zen and then have to replace their motherboards and RAM?

2) The FX thrives on high clocks. Being stuck on 32nm hurts given the current competition. Being able to bump up the stock base clock helps with marketing and helps to diminish the gap in individual core performance. Moving to a mature 28nm should help to improve the performance of these chips, especially in comparison with the performance per watt of the 9590. If AMD could improve its turbo mode some that could also make a big difference in narrowing the gap.

3) The performance is not "horrible" considering that these are 2012 32nm chips that were not designed for single-core performance in the first place. If you look at tasks that specifically load all the cores they do pretty nicely considering the low cost.

4) DX12, at least if one goes by Ashes, is making CPU performance less important. An i3 gained a lot of ground in Ashes with a 390X under DX12. So, gamers can potentially worry less about not having a super-fast CPU but they can take advantage of the extra MT performance in programs like Handbrake and Blender.

 

[BigDaveX]

At this point it's just too late. It'd just be throwing engineering and manufacturing resources that could be used for Zen at a product that would make zero impact on the market.

If they'd been introducing a Socket AM4-based product of some description, that I could have actually gotten behind, since it'd take some pressure off of Zen - which has the daunting task of introducing a new architecture, socket and memory type all in one go, something I don't think even Intel's tried in recent memory - but realistically it's probably too late for that as well.

 

[dark zero]

AMD has not money and their designs are not compatible with FX.

Wondering why they didn't stopped with their FX chips until now...

 

[superstition]

At this point it's just too late. It'd just be throwing engineering and manufacturing resources that could be used for Zen at a product that would make zero impact on the market.

It seems like 28nm has plenty of maturity and capacity. AMD already has plenty of stuff on it. How hard would it be to port the 8 core chips? All AMD needs to offer are two parts: One low watt and one full throttle.

 

[Fjodor2001]

Better to spend the R&D resources on Zen @ 14 nm, scheduled for release in 2016 anyway.

 

[Phynaz]

Better to spend the R&D resources on Zen @ 14 nm, scheduled for release in 2017 anyway.

Fixed that for you.

But I agree. BD and it's children are failures, there's no point in throwing more money and effort at them.

 

[superstition]

Better to spend the R&D resources on Zen @ 14 nm, scheduled for release in 2016 anyway.

I hope, for their sake, that it is 2016 and not something like Q2 2017.

 

[Phynaz]

I hope, for their sake, that it is 2016 and not something like Q2 2017.

"we remain on track for “Zen” sampling in 2016 with first full year of revenue in 2017"

-Drew Prairie, AMD.

Zen isn't 2016 anymore.

 

[superstition]

Zen isn't 2016 anymore.

That's what I thought which is one of the reasons why I made this topic.

It seems like Zen keeps slipping into the future and will probably end up like Skylake on the desktop... hardly the ultra-revolutionary thing it was hyped that it would be.

 

[erunion]

2) The FX thrives on high clocks. Being stuck on 32nm hurts given the current competition. Being able to bump up the stock base clock helps with marketing and helps to diminish the gap in individual core performance. Moving to a mature 28nm should help to improve the performance of these chips
.

This is false. GF 32nm is higher clocking than GF28nm. Switching to 28nm would LOWER clocks, as seen already seen in APUs.

 

[Fjodor2001]

"we remain on track for “Zen” sampling in 2016 with first full year of revenue in 2017"

-Drew Prairie, AMD.

Zen isn't 2016 anymore.

"with the first full year of revenue in 2017"

http://www.fool.com/investing/gener...advanced-micro-devices-incs-zen-based-ch.aspx

I.e. full year for 2017. Does it matter if it's released on 2016-12-31 or 2017-01-01?

Anyway, same policy as Intel followed for Broadwell, which had actual launch in 2016, even though sampled in extremely small volumes in 2015Q4.

 

[AtenRa]

At 20nm SOI or 14nm FF I would sure would like an 8-Core Excavator, not at 28nm though.

 

[superstition]

This is false. GF 32nm is higher clocking than GF28nm. Switching to 28nm would LOWER clocks, as seen already seen in APUs.

Is that the node or AMD's choice of transisitors (low-leakage vs. high)?

Or is this because only 32nm used SOI?

At 20nm SOI or 14nm FF I would sure would like an 8-Core Excavator, not at 28nm though.

But are any foundries available with capacity in those nodes? Plus, such large shrinkage would entail a lot of R&D, would it not?

For instance:

Cliff Hou, vice president for design technology at TSMC was more conservative. He estimated engineers working in the 10nm node will face more than 5,000 design rules compared to 4,000 at 16nm and less than 2,000 in the 28nm node.

 

[Fjodor2001]

At 20nm SOI or 14nm FF I would sure would like an 8-Core Excavator, not at 28nm though.

But what's the point, when Zen is supposed to be released in ~1 year from now?

 

[Phynaz]

"with the first full year of revenue in 2017"

http://www.fool.com/investing/gener...advanced-micro-devices-incs-zen-based-ch.aspx

I.e. full year for 2017. Does it matter if it's released on 2016-12-31 or 2017-01-01?

Anyway, same policy as Intel followed for Broadwell, which had actual launch in 2016, even though sampled in extremely small volumes in 2015Q4.

As IDC said, if Zen were going to ship in 2016 AMD would say so.

You should look up what a logical fallacy is, so you stop writing about Intel in every attempt to defend AMD.

 

[SlowSpyder]

I'd love to see it, I think Excavator with improved (over Piledriver) L3 would be a nice part. But I doubt they'd make money from it, probably just not enough people with AM3+ boards out there anymore.

 

[AtenRa]

But are any foundries available with capacity in those nodes? Plus, such large shrinkage would entail a lot of R&D, would it not?

For instance:

I was trying to say that no matter how good the design is, without a good fabrication process you will not see any advantages going from 32nm SOI to 28nm bulk.

So actually there is no real reason to have an 8-core Excavator at 28nm and that is why AMD hasnt made one.

 

[AtenRa]

But what's the point, when Zen is supposed to be released in ~1 year from now?

I was talking hypothetically, if we could have a 20nm SOI or 14nm FF 8-Core Excavator today i would get one eyes closed.