Would you buy a Steamroller/Excavator based AMD FX CPU if it existed today?

[DrMrLordX]

To the best of my knowledge, we won't see any Steamroller chips for FM2+ with more than 3M/6T, and probably never more than 2M/4T in reality. Not sure if the same design limitations exist with Excavator/Carrizo.

Despite the weakness of the AM3+ platform, I think the new low-leakage FX parts will breathe new life into the platform along the order of 6-12 months. The perf/watt on them is fundamentally better than the bog standard 8350 that people keep pointing to whenever discussing the subject of the FX line.

 

[sm625]

IF excavator were 15% faster than kaveri at javascript, AND it overclocked to 4.8GHz, AND the L2 bandwidth was within 20% of haswell, then yeah probably.

 

[Enigmoid]

Not buying or considering anything until I see a review.

 

[SlowSpyder]

And this is what i've been saying all along. Current FX lineup is not as bad as some say or think, the biggest handicap is that ancient AM3+ tech they're bundled with.

I would buy if they released a updated chipset but i would never touch an FX chip again with this AM3+ crap they have now. My Intel setup is so much better and that's not because of the cpu alone, the entire package counts.

The platform is dated, but many boards still come with USB 3 and SATA 6Gbps. Somewhat surprisingly MSI released a new AM3+ board not all that long ago that has some decent features. Only up to four USB 3.0 connectors, though.

 

[Bradtech519]

The platform is dated, but many boards still come with USB 3 and SATA 6Gbps. Somewhat surprisingly MSI released a new AM3+ board not all that long ago that has some decent features. Only up to four USB 3.0 connectors, though.

I upgraded last year to my AM3+/FX8350 rig. I've been pleased with my board. It also has USB 3 & SATA 3. No PCIe 3.0 like you can find on some FM2+ boards though. Some people have had issues with this board mainly with Vishera CPUs. Usually when voltages are left to auto. I've set all mine manually and it's stable now. It's one of those "it is what it is" deals. I paid no where near retail prices is the only reason I went with this rig($225 for CPU, Mobo, RAM, Cooler). I can't say I'd pay retail prices for the CPU & Motherboard if I didn't get this deal at the price range I got it. Coming in having to pay retail I don't think having PCIe 3.0, USB3, & SATA3 would have been what made me do so. The performance of Intel CPUs (Core I5 Haswells) over AMD in the same price range is what it would be. All the extra features are nice for future proofing.

 

[Spungo]

I'll buy another high end AMD if they can get their power consumption down. Right now it's a bit of a problem. They can almost match the speed of Intel, but the power consumption is way off.
http://www.anandtech.com/bench/CPU/1026

I wouldn't buy a 220W chip that needs water cooling, so I'll skip the 9 series and look at the next highest AMD offers, which is the FX-8350. That's a great chip and it's 125W give or take. Look up a few notches. Intel i7 4765T is about the same speed but only 35W give or take. That's the change AMD needs. Get the same performance but cut the power consumption by 1/2 or 2/3.

I care about power because my computers run folding 24/7.

 

[Atreidin]

Excavator, yes, steamroller, nah. Would be fun to poke at and see what it can do.

 

[MLSCrow]

Honestly, if AMD did make a quad module Steamroller or Excavator (obviously more Excavator than SR) FX chip, made it on an FM2+ board, with DDR4 support, with it reaching clock ranges from 4-5GHz, I feel that it would be plenty for just about anything, gaming included and if 15% more efficient, it'd even clip at the heals of Ivy, which, isn't too shabby imo. It would also cost the same as an i5, and you could save that Hyperthreading money and put it toward a better GPU or other part. The more I think about it, the more I'm convinced I'd buy one. If only AMD would make it.

 

[Abwx]

http://www.anandtech.com/bench/CPU/1026

I wouldn't buy a 220W chip that needs water cooling, so I'll skip the 9 series and look at the next highest AMD offers, which is the FX-8350. That's a great chip and it's 125W give or take. Look up a few notches. Intel i7 4765T is about the same speed but only 35W give or take. That's the change AMD needs. Get the same performance but cut the power consumption by 1/2 or 2/3.

I care about power because my computers run folding 24/7.

One can easily come to such flawed consclusions when using as exemple a rigged FP bench, let s see what a real FP bench says :

http://www.anandtech.com/bench/CPU/502

 

[OVerLoRDI]

I'd be interested in the APUs.

Give me a high TDP APU with decent CPU and solid GPU performance? I'd build myself a little Mini-ITX unit and be done with it.

 

[AtenRa]

SteamRoller and Excavator were doomed the moment GloFo decided to ditch IBMs 20nm Gate First SOI and go Gate last FF. Add to that AMDs decision to go Fusion chips only and Server iGPU-less designs were doomed to oblivion.

32nm SOI SKUs can barely manage to compete today only because of lower prices but when first 14nm Intel desktop SKUs will became available i strongly believe that even lower prices will not be enough to recommend any 32nm in 2015.

 

[Spungo]

One can easily come to such flawed consclusions when using as exemple a rigged FP bench, let s see what a real FP bench says :

http://www.anandtech.com/bench/CPU/502

Pretty much every other test puts the 125W AMD FX-8350 against the 35W Intel i7, and the i7 wins a lot of the time.

That's the problem with 32nm technology. The power consumption is much higher for the same performance.

 

[Spawne32]

SteamRoller and Excavator were doomed the moment GloFo decided to ditch IBMs 20nm Gate First SOI and go Gate last FF. Add to that AMDs decision to go Fusion chips only and Server iGPU-less designs were doomed to oblivion.

32nm SOI SKUs can barely manage to compete today only because of lower prices but when first 14nm Intel desktop SKUs will became available i strongly believe that even lower prices will not be enough to recommend any 32nm in 2015.

How is gate-last a bad thing?

 

[AtenRa]

How is gate-last a bad thing?

I didnt say it is a bad thing, GloFo ditched IBMs process and aimed for a new Gate Last FF process. That way AMD was left with 32nm and 28nm untill 2016+.
If GloFo would use IBMs 20nm Gate First SOI, AMD would have a SteamRoller 20nm product in 2014 as it was the original plan.
With 28nm Planar, a Quad Module 8 Core SKU would not be better than current 32nm SOI so they didnt even tried to make one.

28nm Planar was only barely acceptable for the APUs due to iGPU high density logic. And 28nm HDL will only have a huge affect in Low Power SKUs for Mobile. I don’t expect a huge improvement in Desktop SKUs over Kaveri with Carrizo.

 

[Spungo]

That way AMD was left with 32nm and 28nm untill 2016+.

Good lord.

It would be nice if Intel would make chips for AMD and Nvidia. 22nm or 14nm for a video card would be awesome.

 

[Idontcare]

SteamRoller and Excavator were doomed the moment GloFo decided to ditch IBMs 20nm Gate First SOI and go Gate last FF.

You frame that situation as if GloFo had a choice. They had no choice.

Their customer's told them that they either get onboard with developing and offering a 20nm node that competed with what Samsung and TSMC were developing for 20nm or the customers were going to walk away and go sign contracts with TSMC and Samsung.

And AMD wasn't about to subsidize or foot the R&D bill for developing yet-another specialized sub-node at GloFo that only AMD would be using.

The way you should really frame the 20nm debacle is to note that it was IBM who decided to pursue a node development course that essentially was very self-serving to IBM's business needs to the exclusion of their own fab-club process development partners.

GloFo couldn't count on IBM to care enough about its customers to pursue the development of a 20nm node that delivered what GloFo had been guided by its customers as needing in order to compete with 20nm orders that were going to TSMC and Samsung.

And look at where that bullheadedness got IBM (they got their 20nm SOI for their Power chips, but at the cost of alienating themselves and losing enough business that they have to sell their fabs to GloFo) and GloFo (who had to license Samsung's internally and independently developed 14nm and 20nm process).

At the very least we should credit GloFo for trying to get the customers what they wanted. But their fate wasn't in their control, it was in IBM's control, and it turns out IBM couldn't even be good stewards of their own fate let alone that of their fab-club partners.

 

[AtenRa]

You frame that situation as if GloFo had a choice. They had no choice.

Their customer's told them that they either get onboard with developing and offering a 20nm node that competed with what Samsung and TSMC were developing for 20nm or the customers were going to walk away and go sign contracts with TSMC and Samsung.

And AMD wasn't about to subsidize or foot the R&D bill for developing yet-another specialized sub-node at GloFo that only AMD would be using.

The way you should really frame the 20nm debacle is to note that it was IBM who decided to pursue a node development course that essentially was very self-serving to IBM's business needs to the exclusion of their own fab-club process development partners.

GloFo couldn't count on IBM to care enough about its customers to pursue the development of a 20nm node that delivered what GloFo had been guided by its customers as needing in order to compete with 20nm orders that were going to TSMC and Samsung.

And look at where that bullheadedness got IBM (they got their 20nm SOI for their Power chips, but at the cost of alienating themselves and losing enough business that they have to sell their fabs to GloFo) and GloFo (who had to license Samsung's internally and independently developed 14nm and 20nm process).

At the very least we should credit GloFo for trying to get the customers what they wanted. But their fate wasn't in their control, it was in IBM's control, and it turns out IBM couldn't even be good stewards of their own fate let alone that of their fab-club partners.

I dont disagree with anything you just said, but i was only talking about Quad Module or 5x Module SteamRoller and Excavator SKUs at 20nm that were canceled due to GloFo pursuing the process they needed to compete against TSMC and the others.
It was that moment that both SteamRoller and Excavator Server parts were doomed and AMD had no choice but to change the entire roadmap up to 2016 for their entire portfolio.

 

[NostaSeronx]

Should point out that the 28-nm SHP node is from Freescale.

http://i.imgur.com/ioKXTsg.png

Those for Advanced 28-nm FDSOI;
Freescale
GlobalFoundries
Samsung

http://i.imgur.com/yrzjA0v.jpg

 

[Lepton87]

I didnt say it is a bad thing, GloFo ditched IBMs process and aimed for a new Gate Last FF process. That way AMD was left with 32nm and 28nm untill 2016+.
If GloFo would use IBMs 20nm Gate First SOI, AMD would have a SteamRoller 20nm product in 2014 as it was the original plan.
With 28nm Planar, a Quad Module 8 Core SKU would not be better than current 32nm SOI so they didnt even tried to make one.

28nm Planar was only barely acceptable for the APUs due to iGPU high density logic. And 28nm HDL will only have a huge affect in Low Power SKUs for Mobile. I don’t expect a huge improvement in Desktop SKUs over Kaveri with Carrizo.

Are you sure you mean 20nm and not 22nm SOI the same process IBM Power 8 CPUs are using? The process is already developed but thanks to Global Foundries AMD can't use high performance nodes and thus it just can't compete on the high-end and even mainstream. The question remain if it weren't for the GF would IBM even license the process to AMD? I'm not so sure they would license their most cutting edge process.

 

[DrMrLordX]

I wouldn't buy a 220W chip that needs water cooling, so I'll skip the 9 series and look at the next highest AMD offers, which is the FX-8350.
.

This here illustrates an attitude that needs to change. AMD updated their FX line. Unless you insist on running FX chips @ stock settings, you can do much more with an 8320E or 8370E than you can an 8350, in terms of perf/watt. It is probable that the 8310 is similarly capable.

The 8350 isn't the sweet spot for the FX anymore.

 

[Vesku]

An Excavator 4 module FX with adjustable TDP and well binned would be a great last hurrah for my AM3+ box that is currently running an Athlon II X3 450. I might just settle for a FX 6300 if I see a good sale but I'd have to think about it as opposed to a decently priced Excavator 4 module.

 

[Ajay]

No - because AMD cannot produce a quad module CPU right now that will have the price/perf or perf/watt that I'm looking for - they are still stuck with 28nm HPM. 20nm could be available to them soon, but it's only LPM. Sucks being AMD right now, but big risks can bring big rewards or utter failure.

 

[Idontcare]

I dont disagree with anything you just said, but i was only talking about Quad Module or 5x Module SteamRoller and Excavator SKUs at 20nm that were canceled due to GloFo pursuing the process they needed to compete against TSMC and the others.
It was that moment that both SteamRoller and Excavator Server parts were doomed and AMD had no choice but to change the entire roadmap up to 2016 for their entire portfolio.

It wasn't a decision made in a vacuum to AMD's interests though.

It was a business decision reached by both parties once GloFo had enough data to determine how much it would cost to develop the process tech that AMD wanted, and AMD told GloFo they were not interested in paying for it if it was going to cost that much.

Ultimately it was AMD who made the business decision to cancel their own SKUs and for GloFo to cancel the development of the high performance 20nm process.

No different than what happened to 45nm and 32nm at TSMC (the customers told TSMC not to spend R&D money developing those nodes, not the other way around).

In process tech, IDM and foundries alike, it is not a "build it and they will come" business. You only start building it (start developing the node) if at the same time you have customers who are willing to sign contracts (4 yrs out) or letters of intent to have their chips fabbed on that node.

If there isn't enough seed-interest at time T-4yrs, then come time T the node will have never been funded for development in the first place. But it is the customers who ultimately direct the foundries in terms of what to build, when to build it, and how fancy (expensive, performance, etc) to make the node.

It really isn't correct to lay blame of AMD's product SKU situation at the feet of their foundry when it was AMD themselves who made the business decision to direct GloFo to go cheap on its 20nm process flow, knowing well in advance that such a decision had the unavoidable consequence of undermining the viability and competitiveness of AMD's planned CPU designs.

 

[Abwx]

No - because AMD cannot produce a quad module CPU right now that will have the price/perf or perf/watt that I'm looking for - they are still stuck with 28nm HPM. 20nm could be available to them soon, but it's only LPM. Sucks being AMD right now, but big risks can bring big rewards or utter failure.

8370E has better perf/watt than a 4670K in many apps, but as pointed numerous time by :

The perf/watt on them is fundamentally better than the bog standard 8350 that people keep pointing to whenever discussing the subject of the FX line.

This here illustrates an attitude that needs to change. AMD updated their FX line. Unless you insist on running FX chips @ stock settings, you can do much more with an 8320E or 8370E than you can an 8350, in terms of perf/watt. It is probable that the 8310 is similarly capable.

The 8350 isn't the sweet spot for the FX anymore.

......urban legends are still used to sustain flawed argumentation and continualy misleading realities about the FX line.

 

[mrmt]

It was a business decision reached by both parties once GloFo had enough data to determine how much it would cost to develop the process tech that AMD wanted, and AMD told GloFo they were not interested in paying for it if it was going to cost that much.

Ultimately it was AMD who made the business decision to cancel their own SKUs and for GloFo to cancel the development of the high performance 20nm process.

Steamroller and Excavator FX killed dead the moment AMD decided to cut losses and pull out of the server market. A 315mm^2 wouldn't be viable as a desktop-only product. My guess is that the APU line would also be dead if not for the WSA quota they had to fulfill with GLF.