AMD Zen-based 8-core Desktop CPU Arrives in 2016, on Socket FM3

[bubbleawsome]

Funny thing is, in well multithreaded games the fx chips are still competing with my i5. DX12 could really add to this.

 

[ensabrenoir]

......you know if Amd sold their ideas to Intel for implementation...... hmmmm.....only in a more perfect world...........

 

[GhostRyder]

Well this is interesting mostly because it brings up a couple of points/interesting ideas/questions.

1: Whats the TDP of this chip
2: What kind of changes can we plan for FM3 (Power Phase design, power draw, feature set, etc)
3: Are they really totally killing AM3+ and are they now killing AM1?

I mean if they are making FM3 the main board for them that is perfectly fine, however currently FM2+ is a limited platform in many areas and it just begs the question how much more serious can they make the platform overall. Its a bit odd honestly but probably necessary since they are getting quite tight on their budget constrains and hard to manage multiple types of chipsets. Ill be following it very fondly and curiously but I am interested in what this could all mean for the platform!

 

[MikeMurphy]

Well this is interesting mostly because it brings up a couple of points/interesting ideas/questions.

1: Whats the TDP of this chip
2: What kind of changes can we plan for FM3 (Power Phase design, power draw, feature set, etc)
3: Are they really totally killing AM3+ and are they now killing AM1?

I mean if they are making FM3 the main board for them that is perfectly fine, however currently FM2+ is a limited platform in many areas and it just begs the question how much more serious can they make the platform overall. Its a bit odd honestly but probably necessary since they are getting quite tight on their budget constrains and hard to manage multiple types of chipsets. Ill be following it very fondly and curiously but I am interested in what this could all mean for the platform!

1-Hot
2-PCIe 3.0, usb3.1
3-AM3 is gone. AM1 is for totally different product range and will stay.

AMD will do a good job on socket FM3.

 

[Jorge]

No surprises here, just as previously reported. Eight core desktop CPUs with sixteen threads (not sixteen cores and thirty-two threads), will be very powerful desktop CPUs with a max TDP of 95w. That's right 95w max for the Zen core desktop CPUs. Zen based APUs obviously will have a range of TDPs based on the core count and graphics used. Socket FM3 will be for the discrete CPUs and FM2+ for the APUs. Both will allow extensive multi-power plane ops. Zen based products will offer all of the current tech options including DDR3/4, PCIe 3.1, USB 3.0 and more.

 

[happita]

No surprises here, just as previously reported. Eight core desktop CPUs with sixteen threads (not sixteen cores and thirty-two threads), will be very powerful desktop CPUs with a max TDP of 95w. That's right 95w max for the Zen core desktop CPUs.

That's a pretty bold claim. Source or it didn't happen.

 

[alwayssts]

Well this is interesting mostly because it brings up a couple of points/interesting ideas/questions.

1: Whats the TDP of this chip
2: What kind of changes can we plan for FM3 (Power Phase design, power draw, feature set, etc)
3: Are they really totally killing AM3+ and are they now killing AM1?

I mean if they are making FM3 the main board for them that is perfectly fine, however currently FM2+ is a limited platform in many areas and it just begs the question how much more serious can they make the platform overall. Its a bit odd honestly but probably necessary since they are getting quite tight on their budget constrains and hard to manage multiple types of chipsets. Ill be following it very fondly and curiously but I am interested in what this could all mean for the platform!

All good questions.

I really get the feeling Zen is going to be essentially Broadwell without terrible scaling issues; or essentially clock similar or even higher compared to skylake but without the arch/ipc improvements (and ditching the built-in graphics should allow them some space to clock cpus higher in a similar tdp). One would think this would be something like 2x65w (140w) chips on a package, but who knows.

In a world where Broadwell didn't *appear* to be an enthusiast IPC/single-threaded disaster (made up for on the desktop through gpu upgrades almost nobody cares about), it would be interesting to see 8-core parts battle it out at similar clocks (4.2ghz?). As it sits though, one has to wonder if wonder if Intel will bring something like 2-4 extra cores at a lower clock to the enthusiast fight, while AMD could bring a higher core clock on the 'midrange' desktop to play against Skylake.

I could be totally wrong, but I envision them doing something like set tdp aims for both chips and mcm packages (ex: 35w, 65w, 95w, CPU; 30w, 60-65w, 70-75w etc gpu) and then mixing/matching it with either a single cpu chip, two cpu chips, or one cpu + greenland gpu clocked in different configurations (ex: 65w/95w/140w etc)...but it's also conceivable they could be arbitrary and very much dictated by whatever it takes to match/beat Intel (or even a low-end nvidia gpu) in whatever metric a certain product is aimed (or is the most efficient mix hence showing them in their best light).

It surely will be interesting to see how this all develops...but I think such a slotted approach (especially from a binning perspective) would make sense, as well as could explain why one such apu is rumored to have an extravagant tdp...It could essentially be the 'top bin' of both cpu + gpu chips (ex: 95w cpu + 75w gpu).

 

[newtekie1]

I'm actually happy they are unifying onto one socket. It was time for AM3+ to die. The 3 chip method is old, and AMD3+ is old. To try and shoehorn a new architecture to fit into AM3+ would have been too limiting. A new platform for a new architecture is the best way to do it if you want the best product.

 

[ShurikN]

That's a pretty bold claim. Source or it didn't happen.

On 14nm I doubt it'll go over 90W. If it does, they failed somewhere... hard.

 

[btarunr]

The first leak (or whatever it was) spoke about SMT, and not having SMT sounds pretty daft in this age.

Yeah, Core i5-4690K peasants will never know what SMT is.

 

[btarunr]

On 14nm I doubt it'll go over 90W. If it does, they failed somewhere... hard.

Intel, with all its 14 nm divine galactic master-race knowledge, is having to rate its Skylake quad-core chips at 95W. I don't think AMD will do better.

 

[Dent1]

Funny thing is, in well multithreaded games the fx chips are still competing with my i5. DX12 could really add to this.

Wasted energy, nobody here will read that, comprehend that, or acknowledge that.

In fact if I didn't reply everyone would be happy to let that comment get buried.

 

[GhostRyder]

1-Hot
2-PCIe 3.0, usb3.1
3-AM3 is gone. AM1 is for totally different product range and will stay.

AMD will do a good job on socket FM3.

Well hot does not really show much on that as hot is relevant to the chip mostly. It could be as low as 95watt or as high as 200watt but I was guessing in the middle depending on clocks (Maybe again 125watt). As for PCIE 3.0, FM2+ already has that though 3.1 and hopefully m.2 will be part of it. Though I was mostly wondering the chipset and board features in general because most FM2+ chips are not exactly high end and these chips seem aimed at the higher range of performance (8 core 16 thread). With the AM3 debate yea it sounds so but AM1 has been left out of a lot of stuff which can be worrying because its a great product line in my book and I hope they continue it.

All good questions.

I really get the feeling Zen is going to be essentially Broadwell without terrible scaling issues; or essentially clock similar or even higher compared to skylake but without the arch/ipc improvements (and ditching the built-in graphics should allow them some space to clock cpus higher in a similar tdp). One would think this would be something like 2x65w (140w) chips on a package, but who knows.

In a world where Broadwell didn't *appear* to be an enthusiast IPC/single-threaded disaster (made up for on the desktop through gpu upgrades almost nobody cares about), it would be interesting to see 8-core parts battle it out at similar clocks (4.2ghz?). As it sits though, one has to wonder if wonder if Intel will bring something like 2-4 extra cores at a lower clock to the enthusiast fight, while AMD could bring a higher core clock on the 'midrange' desktop to play against Skylake.

I could be totally wrong, but I envision them doing something like set tdp aims for both chips and mcm packages (ex: 35w, 65w, 95w, CPU; 30w, 60-65w, 70-75w etc gpu) and then mixing/matching it with either a single cpu chip, two cpu chips, or one cpu + greenland gpu clocked in different configurations (ex: 65w/95w/140w etc)...but it's also conceivable they could be arbitrary and very much dictated by whatever it takes to match/beat Intel (or even a low-end nvidia gpu) in whatever metric a certain product is aimed (or is the most efficient mix hence showing them in their best light).

It surely will be interesting to see how this all develops...but I think such a slotted approach (especially from a binning perspective) would make sense, as well as could explain why one such apu is rumored to have an extravagant tdp...It could essentially be the 'top bin' of both cpu + gpu chips (ex: 95w cpu + 75w gpu).

Yea, no matter what its going to be interesting either way. I guess we will need more details mostly because this is kinda a shocking development overall and it leaves a lot of questions with very little answers. I just hope this at least offers 25%+ IPC improvements overall and this can compete effectively at least on some levels more than straight up budget.

I'm actually happy they are unifying onto one socket. It was time for AM3+ to die. The 3 chip method is old, and AMD3+ is old. To try and shoehorn a new architecture to fit into AM3+ would have been too limiting. A new platform for a new architecture is the best way to do it if you want the best product.

Yea that's true I agree, I just wonder how the FM3 platform will be improved to incorporate this.

 

[natr0n]

This might be worth an upgrade.

Very Excite.

 

[ThE_MaD_ShOt]

Sounds like I see my next main rig.

 

[BiggieShady]

That's a pretty bold claim. Source or it didn't happen.

I have never seen Jorge reply to a reply of his post ... it's always post and run ... and it always reads like a marketing material.

 

[suraswami]

Hopefully Motherboard manufacturers don't F it up!!

 

[ShurikN]

Hopefully Motherboard manufacturers don't F it up!!

And why would they?!
They will have one socket to work with. The rest is up to AMD

 

[MikeMurphy]

Well hot does not really show much on that as hot is relevant to the chip mostly. It could be as low as 95watt or as high as 200watt but I was guessing in the middle depending on clocks (Maybe again 125watt). As for PCIE 3.0, FM2+ already has that though 3.1 and hopefully m.2 will be part of it. Though I was mostly wondering the chipset and board features in general because most FM2+ chips are not exactly high end and these chips seem aimed at the higher range of performance (8 core 16 thread). With the AM3 debate yea it sounds so but AM1 has been left out of a lot of stuff which can be worrying because its a great product line in my book and I hope they continue it.

The AMD IPC is poor compared to Intel's. AMD compensates for poor IPC by boosting clocks. That generates quite a bit of heat, hence the new chips will run hot.

AM1 hasn't been left out of anything. It's an ultra-low cost platform with features sufficient for its purpose. It's not a luxury or performance platform.

 

[TheinsanegamerN]

And why would they?!
They will have one socket to work with. The rest is up to AMD

The same reason it took three freakin years of a manufacturer to make a micro atx am3+ board with a 900 series chipset, when microatx am3 boards had existed before? because manufacturers still make some am3+ and fm2+ motherboards WITHOUT vrm cooling? they constantly screw up, cutting every corner, while comparable intel boards are flush with features and massive vrm counts and heatsinks.

It is quite a legitimate concern that the platform will ultimately be tripped up by not having any decent micro atx boards. or mini itx, for that matter. many people don't want or need full atx, and limiting to full atx is leaving any micro or mini itx game rig to intel.

 

[suraswami]

The same reason it took three freakin years of a manufacturer to make a micro atx am3+ board with a 900 series chipset, when microatx am3 boards had existed before? because manufacturers still make some am3+ and fm2+ motherboards WITHOUT vrm cooling? they constantly screw up, cutting every corner, while comparable intel boards are flush with features and massive vrm counts and heatsinks.

It is quite a legitimate concern that the platform will ultimately be tripped up by not having any decent micro atx boards. or mini itx, for that matter. many people don't want or need full atx, and limiting to full atx is leaving any micro or mini itx game rig to intel.

To add to that, even with ATX variants, it took about may be 10 revisions to get the formula correct to run those FX properly. There are people out there who are willing to pay extra for quality boards.

And not sure if would work, but AMD should look into quality of these boards and run some kind of control on them.

The AMD IPC is poor compared to Intel's. AMD compensates for poor IPC by boosting clocks. That generates quite a bit of heat, hence the new chips will run hot.

AM1 hasn't been left out of anything. It's an ultra-low cost platform with features sufficient for its purpose. It's not a luxury or performance platform.

You mean to say Intel don't have boost clock feature?

 

[MikeMurphy]

You mean to say Intel don't have boost clock feature?

No. Please read.

 

[suraswami]

No. Please read.

Don't understand. Intel chips run hot too. Only thing AMD is lagging behind is power efficiency (of course with the argument of IPC). I read somewhere watts rating of AMD is different than Intel.

Correct me if I am wrong.

 

[mastrdrver]

No way Summit Ridge and Bristol Ridge will be swappable on the same board. It appears Bristol has the southbridge on chip where as Summit lacks it.

Leaks say the hottest Summit Ridge will be 95w (saw a leak but can't remember where it was). Some are expecting at least a 40% improvement over Excavator in IPC. It will not beat Skylake, but it should be close enough.

Memory subsystem is going to see massive gains since AMD is changing to an inclusive cache design.

 

[Aquinus]

The AMD IPC is poor compared to Intel's. AMD compensates for poor IPC by boosting clocks. That generates quite a bit of heat, hence the new chips will run hot.

You're forgetting one thing. They're ditching this mess of a modular and super long pipelined CPU. AMD ran into the same issues Intel did with Netburst, which was that pipeline got too long and that even with high clock speeds, branch miss-prediction became a real problem when it came to stalling the pipeline. On top of that, AMD's L2 cache was way too big and slow compared to Intel's smaller L2.

With that said, we know that AMD has reduced the size of the L2 cache per core and this can only help improve latencies. It is also now dedicated per core (or SMT pair, like Intel's CPUs.) If we also assume that AMD have overhauled their core (which they have,) and that any improvement should be significant. I also suspect that 14nm will play a roll in lowering power consumption. All in all, I'm expecting a measurable improvement with Summit Ridge.

It appears Bristol has the southbridge on chip where as Summit lacks it.

I'm having a little bit of trouble wrapping my head around how that would work. I suspect that FM3 won't be an SoC socket. However it will be like all the others with the PCI-E root complex on the CPU.

Side note: I would hate to see HyperTransport disappear, but then again, why use it when you have PCI-E 3.0.