Future to Bulldozer architecture?

[Flash831]

Now when Ryzen is released, the Bulldozer line of processors seems to have almost come to an end. Bristol Ridge still sells as OEM only. It might become available through the retail channel, even if that is unclear at the moment.

Stoney Ridge seem to get one last spin however, as a "Stoney Ridge Refresh" are slated for 2017 AND 2018.

Source: https://videocardz.com/67362/amd-pinnacle-ridge-to-feature-up-to-8-zen2-cores

I know NostaSeronx is implying that AMD still has some plans for future evolvement of the BD architecture on 22FDX for instance.

I suppose that now with Ryzen out and Vega on the horizon (hehe), new money will pour into AMD. Would it be feasible to continue working on the BD architecture? I mean the architecture probably has some use cases where it is better suited than a more "typical" core.

For instance, a module is smaller than two separate cores. So I suppose it might fit where performance is not a priority, but rather making the chip as cheap as possible to produce (like Stoney Ridge).

Do you see any future to the Bulldozer architecture?

 

[AMDisTheBEST]

Why would it have future? My FX 8320 runs like an old horse even though I purchased it back in August 2015. Bulldozer derivative is still good for mid level gaming and handles all non gaming related tasks well. I will be keeping FX 8320 for as long as I can because it is a very reliable chip but no way will I buy it or any of its iterative for a new PC build/upgrade. It is now made completely obsolete by amd ryzen

 

[SarahKerrigan]

I strongly suspect future R&D costs, including those involved in porting to a new process, outweigh any usefulness the BD family still has. I'd guess Stony Ridge Refresh will be a minor tweak like Bristol Ridge was, NOT a new design, on the existing process.

It seems likely that AMD will use a Zen derivative for that niche after 2018, although I could see a parameterizable low-end Zen variant or even a cat successor for the licensable-core market.

 

[CrazyElf]

I think AMD is done with Bulldozer.

Zen derivatives will dominate ... .and force Intel to compete at all segments.

 

[inf64]

Bulldozer is gone, let is rest in peace. Zen derivatives are the future, perf./watt is like 2x+ times better.

 

[Abwx]

perf./watt is like 2x+ times better.

Thanks to 14nm, otherwise a shrinked BR with a smaller GPU would be more than adequate for low cost laptops and would be competitive with Zen perf/watt wise....

 

[DrMrLordX]

We'll never know what Bristol Ridge would've been like on 14nm LPP. Might've been interesting. That small l2 cache would still be a liability, as would the entire cache architecture really.

 

[NostaSeronx]

It should be noted that a lot of things can be done to Bulldozer-derived successors.

- Move from hand-custom synthesized macros to fully-synthesized macros. (The Bulldozer architecture is using hand-custom macros from 0.35 µm Alpha 7(21264).)
- Separate from the Alpha legacy and use more efficient, while recently generated, general purpose cores. Move to a more performance/power optimal floating point unit, rather than the full balls to the wall performance of current FPU.
- Use a partitioned cache system which is mostly unified.
- etc.

 

[unseenmorbidity]

Probably a dumpster.

 

[Flash831]

It should be noted that a lot of things can be done to Bulldozer-derived successors.

- Move from hand-custom synthesized macros to fully-synthesized macros. (The Bulldozer architecture is using hand-custom macros from 0.35 µm Alpha 7(21264).)
- Separate from the Alpha legacy and use more efficient, while recently generated, general purpose cores. Move to a more performance/power optimal floating point unit, rather than the full balls to the wall performance of current FPU.
- Use a partitioned cache system which is mostly unified.
- etc.

Nosta, in what scenario do you see a future BD derivate be used? I suspect BD and ZEN could complement each other, depending on use case.

 

[imported_jjj]

There is no reason to use BD as it's big, power hungry and low perf. X86 only has relevance in PC and server but even there its relevance is fading. In all other segments, standard ARM cores would be many times better than BD.

What AMD should do is release the ARM version of Zen next year and use the same socket as the x86 based products. Desktop and server since desktop has a strategic importance in supporting adoption.
Likely they won't do that just yet and focus on x86, unless they need the ARM version for semi-custom and then it becomes cheaper to do it.

 

[VirtualLarry]

or even a cat successor for the licensable-core market.

Are the "cat" cores well and truly dead? There's a thread asking whether the APU in the new Xbox "Scorpio" is going to use Jaguar or Zen core derivatives.

I'm just thinking, if AMD is planning on using Jaguar (or successor "cat" cores), in the new XBox, on 14nm, then why not spin up a 14nm "cat" core APU for the consumer / OEM / DIY market? Then again, they may consider that so niche, that they wouldn't bother. (Like Bristol Ridge is currently.)

I wouldn't mind seeing some new AM1 CPUs on 14nm rather than 28nm; certainly, they should be able to clock higher within the same TDP envelope. TBH, AM1 wasn't a horrible platform - it had SATA6G and USB3.0. Only, the CPUs released for it clocked kind of low, and had around Core2 IPC, so they didn't have stellar ST benchmarks.

If they could re-spin on 14nm, and get up to 3.0Ghz clocks (or higher?), then they might have another winner on their hands. Then again, I don't know how well AM1 did on the marketplace. I don't think that they are even making boards for it anymore. (Maybe an AM1B spec? Backwards-compatible with AM1 APUs, but also handling 14nm re-spin APUs?)

 

[NTMBK]

I know NostaSeronx is implying that AMD still has some plans for future evolvement of the BD architecture on 22FDX for instance.

Seronx is notorious for flat out making shit up. Don't listen to him. Here's a few examples:

http://semiaccurate.com/forums/showthread.php?p=258678#post258678
http://semiaccurate.com/forums/showthread.php?p=212021#post212021
http://semiaccurate.com/forums/showthread.php?p=204593&post204593
http://semiaccurate.com/forums/showthread.php?p=203360#post203360
http://forums.anandtech.com/showpost.php?p=34050572&postcount=5
http://forums.anandtech.com/showpost.php?p=34050816&postcount=9
http://forums.anandtech.com/showpost.php?p=34051032&postcount=11
http://forums.anandtech.com/showpost.php?p=34051100&postcount=13
http://forums.anandtech.com/showpost.php?p=34275041&postcount=10

All of the evidence indicates that Bulldozer is basically dead. There might be a Stoney Ridge or Bristol Ridge refresh for the ultra low budget markets, but I don't expect any new designs (and certainly not on <28nm).

 

[NTMBK]

Are the "cat" cores well and truly dead? There's a thread asking whether the APU in the new Xbox "Scorpio" is going to use Jaguar or Zen core derivatives.

I'm just thinking, if AMD is planning on using Jaguar (or successor "cat" cores), in the new XBox, on 14nm, then why not spin up a 14nm "cat" core APU for the consumer / OEM / DIY market? Then again, they may consider that so niche, that they wouldn't bother. (Like Bristol Ridge is currently.)

I wouldn't mind seeing some new AM1 CPUs on 14nm rather than 28nm; certainly, they should be able to clock higher within the same TDP envelope. TBH, AM1 wasn't a horrible platform - it had SATA6G and USB3.0. Only, the CPUs released for it clocked kind of low, and had around Core2 IPC, so they didn't have stellar ST benchmarks.

If they could re-spin on 14nm, and get up to 3.0Ghz clocks (or higher?), then they might have another winner on their hands. Then again, I don't know how well AM1 did on the marketplace. I don't think that they are even making boards for it anymore. (Maybe an AM1B spec? Backwards-compatible with AM1 APUs, but also handling 14nm re-spin APUs?)

They already made 16/14nm ports of the Jaguar core for the Xbox One S and the PS4 Pro (I think the PS4 Slim, as well?). But I guess the costs of taping out another 14nm design just aren't worth it for such a small, low-end market.

I think they'd be better off making a 2x Zen APU for low powered laptops.

 

[NostaSeronx]

Nosta, in what scenario do you see a future BD derivate be used?

Where Bristol Ridge/Brown Falcon and Stoney Ridge/Prairie Falcon is at currently.

I suspect BD and ZEN could complement each other, depending on use case.

It should be noted it will be related in the Bulldozer family, but not related to the current Bulldozer core lineup. Much like how Bobcat is 14h and Jaguar is 16h, but both are considered Cat cores. The trend of Bobcat(Brazos/Brazos 2.0/Brazos-T) -> Jaguar is also seen with Excavator(Carrizo/Bristol Ridge/Stoney Ridge) -> ???.

Since, the FinFET basket is filled with Zen. The next gen BD lineup is more or less guaranteed to go for FDSOI.

Areas to watch for future development:
- China
- Colorado & Mexico
- Boston

All of which are ULP-related in SoC development and related to Cat/Excavator dev. Where Zen has largely been developed from the High Performance sectors that dealt with Bulldozer/Piledriver.

 

[VirtualLarry]

They already made 16/14nm ports of the Jaguar core for the Xbox One S and the PS4 Pro (I think the PS4 Slim, as well?). But I guess the costs of taping out another 14nm design just aren't worth it for such a small, low-end market.

I think they'd be better off making a 2x Zen APU for low powered laptops.

That's too bad. I would pay up to $60 for a 3.0Ghz Core2Quad-IPC-equivalent APU, that ran at 25W or under. (And probably could OC to 3.5Ghz. It would be like a Core2 revival all over again, except with SATA6G and USB3.0 and PCI-E 2.0/3.0.)

 

[Insert_Nickname]

Are the "cat" cores well and truly dead? There's a thread asking whether the APU in the new Xbox "Scorpio" is going to use Jaguar or Zen core derivatives.

I'm just thinking, if AMD is planning on using Jaguar (or successor "cat" cores), in the new XBox, on 14nm, then why not spin up a 14nm "cat" core APU for the consumer / OEM / DIY market? Then again, they may consider that so niche, that they wouldn't bother. (Like Bristol Ridge is currently.)

Outside semi-custom, yes, the Cat cores are dead. AMD stated pretty clearly Zen is the future from top to bottom. It'll be interesting to see what a Zen APU can do in a 4W TDP...

There was even a rumour at some point that Scorpio would use a cut-down Zen variant. I don't what became of it, since I haven't kept up with future Xbox plans.

I wouldn't mind seeing some new AM1 CPUs on 14nm rather than 28nm; certainly, they should be able to clock higher within the same TDP envelope. TBH, AM1 wasn't a horrible platform - it had SATA6G and USB3.0. Only, the CPUs released for it clocked kind of low, and had around Core2 IPC, so they didn't have stellar ST benchmarks.

If they could re-spin on 14nm, and get up to 3.0Ghz clocks (or higher?), then they might have another winner on their hands. Then again, I don't know how well AM1 did on the marketplace. I don't think that they are even making boards for it anymore. (Maybe an AM1B spec? Backwards-compatible with AM1 APUs, but also handling 14nm re-spin APUs?)

One of the purposes of the AM4 platform was unification. AM4 will be used from top to bottom on the desktop, though I doubt the cheaper varieties can support full-on OC'd Summit Ridge.

 

[bononos]

Whats going to be the future with HSA for AMD?

 

[guachi]

At least Bulldozer is fully compatible with Windows 7.

I guess that's a plus for some.

 

[Azuma Hazuki]

If you want cheap multithread performance this is a kind of golden old age for certain chips. I've seen the FX 8320e, my personal favorite Vishera ever, for as low as $120. That's i5 total performance for the cost of an i3. For certain workloads this is a no-brainer.

 

[Abwx]

The FX8300 is even cheaper and is in principle better than the 8320E, as to be compared to I5s they simply cant as those latter see their throughput collapsing in multitasking once there s an FP task running concurently with an Integer code base app, wich is not the case with the FX..

 

[MentalIlness]

My FX 8320 runs like an old horse even though I purchased it back in August 2015

Same here. My FX 8120 from Feb 2012 Still rockin along. And constantly OC'd.

 

[coffeemonster]

It should be noted that a lot of things can be done to Bulldozer-derived successors.

- Move from hand-custom synthesized macros to fully-synthesized macros. (The Bulldozer architecture is using hand-custom macros from 0.35 µm Alpha 7(21264).)
- Separate from the Alpha legacy and use more efficient, while recently generated, general purpose cores. Move to a more performance/power optimal floating point unit, rather than the full balls to the wall performance of current FPU.
- Use a partitioned cache system which is mostly unified.
- etc.

There was a guy on AMD reddit recently that come into a Bristol Ridge thread talking about how AMD should have kept developing jaguar. That it's considerably faster at the same clock and scales better than Excavator. something something AMD lying to cover up these supposed facts. I dunno, what do you make of it? nonsense?
https://www.reddit.com/r/Amd/comments/61b947/asus_fx550iuwsfx_gaming_laptop_amd_fx_9830p_rx/dfdid5r/

 

[whm1974]

BD and the Cat cores will be waste of time and resources that AMD can't afford. They should release Zen based APUs for the laptop and cheap desktop market.

 

[Abwx]

There was a guy on AMD reddit recently that come into a Bristol Ridge thread talking about how AMD should have kept developing jaguar. That it's considerably faster at the same clock and scales better than Excavator. something something AMD lying to cover up these supposed facts. I dunno, what do you make of it? nonsense?
https://www.reddit.com/r/Amd/comments/61b947/asus_fx550iuwsfx_gaming_laptop_amd_fx_9830p_rx/dfdid5r/

At best it s plain ignorance, at worst it s viral marketing to downplay BR as this SoC made some strides on the mobile market and allowed AMD to grab some marketshare in the mobile segment last year..