Discussion: AMD Ryzen

[elbert]

This isn't the ZEN cpu's to compare against the i7-5960X. The quad and octal are consumer level cpu. Understand AMD has already announce a ZEN 24 and 32 core server dual proc that leaks suggest 2.7Ghz with boosts of 2.9Ghz for the es. Now this means AMD already has a 12 and 16 core CPU for the enthusiasts.

Those 24 and 32 core server CPUs are quad die configurations. The maximum number of cores per die is eight. The highest Zen CPU for consumer is an eight core 16 thread CPU. This is the maximum for enthusiasts. Mainstream users will be served by quad-cores.

Actually the max number is 4 if you notice cache stacks in 4 core increments. Not that it matters but ive yet to see either company slap 4 seperate silicon together as you suggest.

I predicted years ago that AMD would use quad-die SoCs of 8-core each.

And it was confirmed recently.

Again we aren't looking at Snowy Owl batch 8, 12, 16 core nor the Zeppelin ZP cluster 16, 24, 32 core.

 

[juanrga]

Unless the ZEN 8 cores are 250~300. Again the 16 core ZEN is what should be compared to the i7 8 and 10 core. The 12 core ZEN against the i7 6 cores. These are the enthusiast CPU's.

As explained to you a pair of posts above there is no 12-core and 16-core Zen for desktop. The top desktop Zen is 8-core CPU.

You do understand that only talking about summit ridge.

The second slide is the desktop roadmap. As mentioned above the higher core counts, 16-core, 24-core, 32-core are for the server platfform.

 

[juanrga]

I was under the impression that AM4 was going to be dual channel only, and it wouldn't make much sense to me to have a 32 thread CPU fed by dual channel RAM.

Right, AM4 is dual-channel. The server sockets have higher number of channels. 16-core --> quad-channel. 32-core --> 8-channel.

 

[elbert]

I was under the impression that AM4 was going to be dual channel only, and it wouldn't make much sense to me to have a 32 thread CPU fed by dual channel RAM.

Right, AM4 is dual-channel. The server sockets have higher number of channels. 16-core --> quad-channel. 32-core --> 8-channel.

The Snow Owl is quad channel which makes the 3rd socket SP4 BGA package that is pin compatible with SP4R2. This one I hear is backwards compatible to the the AM4 CPU's.

 

[ComputerSecurityGuy]

Yeah. And piling on the servers, 8 Zen cores at 95W is a pretty attractive position. Broadwell and Haswell are both at 140W for similar clocks. Even if Zen was about Ivy Bridge performance wise, it would still make a decent splash.

 

[juanrga]

Traditionally AMD and Intel measure TDP differently. The 125W rated AMD chips actually dissipated a lot more power than Intel 140W. Therefore, we must wait before claiming that 95W AMD chip will be less power hungry than 140W chip from competence. Moreover, the 140W rating on Intel chips includes the power hungry 256bit SIMD units, whereas AMD Zen only has 128bit units.

 

[juanrga]

I was under the impression that AM4 was going to be dual channel only, and it wouldn't make much sense to me to have a 32 thread CPU fed by dual channel RAM.

Right, AM4 is dual-channel. The server sockets have higher number of channels. 16-core --> quad-channel. 32-core --> 8-channel.

The Snow Owl is quad channel which makes the 3rd socket SP4 BGA package that is pin compatible with SP4R2. This one I hear is backwards compatible to the the AM4 CPU's.

The AM4 CPUs are derived from server dies that don't pass qualification. AMD will offer 8-core configurations for servers and that implies backward pin compatibility. But that doesn't mean that it works in the opposite direction with AMD offering 12 and 16 core for desktop. The 8-core Zen is the top product for desktop enthusiasts, both consumer and PRO lines. Check the desktop roadmap above.

 

[elbert]

I was under the impression that AM4 was going to be dual channel only, and it wouldn't make much sense to me to have a 32 thread CPU fed by dual channel RAM.

Right, AM4 is dual-channel. The server sockets have higher number of channels. 16-core --> quad-channel. 32-core --> 8-channel.

The Snow Owl is quad channel which makes the 3rd socket SP4 BGA package that is pin compatible with SP4R2. This one I hear is backwards compatible to the the AM4 CPU's.

The AM4 CPUs are derived from server dies that don't pass qualification. AMD will offer 8-core configurations for servers and that implies backward pin compatibility. But that doesn't mean that it works in the opposite direction with AMD offering 12 and 16 core for desktop. The 8-core Zen is the top product for desktop enthusiasts, both consumer and PRO lines. Check the desktop roadmap above.

No that means the 24 and 32 cores server cpu's will not work on either socket. The 12 and 16 cores will not work on the AM4 motherboard but the 4 and 8 cores will drop in compatible to the enthusiast motherboards. Think of the SP4 BGA as the equal to the 2011v3. The AM4 thus the 1151 and the final socket for the 24 and 32 cores. The only CPU's with leaked confermation of being MCM 24 and 32 cores.

AMD is thought to be using a Multi-Chip Module (MCM) configuration, which would also make sense — it can connect two 16-core chips on the same physical silicon using a high-bandwidth interconnect.

Finally half of the 24 and 32 cores PCI-E lanes(64) are dedicated to 2 and 4 way systems. The motherboard and 12 and 16 cores for enthusiases have a launch of 2017 h2. IE the servers will launch first.

 

[Cazalan]

For reasons I've explained MANY times in the past: Most software doesn't scale, and the software that does tends to scale to infinity, and are being processed by GPUs instead. There's a reason why performance for non-benchmark software peaks somewhere between an i3 and i5, and why per-core performance matters more then core count.

Honestly, Zen would be better off if they cut the number of cores in half and gained 500 MHz. The 16% Clockrate boost is worth FAR more then a doubling of core count.

If AMD does not sell 4/6 core versions then consumers likely will do just that. Buy the 8 core and disable some cores in the BIOS. Freeing up TDP for higher boost clocks. Parts like that end up clocking better anyway because you have more die area for thermal transfer.

 

[juanrga]

For reasons I've explained MANY times in the past: Most software doesn't scale, and the software that does tends to scale to infinity, and are being processed by GPUs instead. There's a reason why performance for non-benchmark software peaks somewhere between an i3 and i5, and why per-core performance matters more then core count.

Honestly, Zen would be better off if they cut the number of cores in half and gained 500 MHz. The 16% Clockrate boost is worth FAR more then a doubling of core count.

If AMD does not sell 4/6 core versions then consumers likely will do just that. Buy the 8 core and disable some cores in the BIOS. Freeing up TDP for higher boost clocks. Parts like that end up clocking better anyway because you have more die area for thermal transfer.

In the first place, the known Zen die has a core-granularity of four. There are four cores in each cluster and it seems that individual cores cannot be disabled inside each cluster. There are only 4-core and 8-core engineering samples, not 6-cores and the above AMD roadmap for desktop CPU shows only 8-core and 4-core. Moreover the next AMD roadmap for mobile only shows 4-core APUs, not 2-cores.

In the second place, disabling cores doesn't imply we can break the physical laws of silicon, thermodynamics, or microarchitecture limitations. We cannot take a 3.5GHz i5 disable two cores and get a boost to 6GHz on air for instance. We have more die area but the heat generated continues being concentrated on a small die area (the production of heat is not distributed across the whole die). Disabling one cluster on the 8-core die Zen could give us about 350MHz more (my estimation).

 

[taseas]

I still want to see when the new AM4 motherboards will come to the stores. Also, what happended to the "Summit Ridge" ? Wasn't it planed to be released alongside AM4 platforms in Q4 2016 ? Is there anything official about a possible delay ? I am waiting for a high-end AM4 motherboard to pair it with an 8C16T Zen CPU and (at least) a Samsung 950 pro M2 SSD. If I go with a good intel board/cpu right now I will exceed by far the 800€ price tag and of course intel i7 prices will drop significantly if AMD ZEN is at least on par with i7-5930k performance.

 

[Ecky]

I still want to see when the new AM4 motherboards will come to the stores. Also, what happended to the "Summit Ridge" ? Wasn't it planed to be released alongside AM4 platforms in Q4 2016 ? Is there anything official about a possible delay ? I am waiting for a high-end AM4 motherboard to pair it with an 8C16T Zen CPU and (at least) a Samsung 950 pro M2 SSD. If I go with a good intel board/cpu right now I will exceed by far the 800€ price tag and of course intel i7 prices will drop significantly if AMD ZEN is at least on par with i7-5930k performance.

If AMD's Zen is on part with 5930K performance, it will not be significantly cheaper than a 5930K, considering AMD is on the edge of bankruptcy - especially within the first few months after launch. Intel sets prices, not AMD, because they have greater marketshare. Chances are good that initially, AMD's chips will actually sell for a premium due to early adopters, then drop down to only very slightly cheaper than their Intel equivalents.

 

[juanrga]

I still want to see when the new AM4 motherboards will come to the stores. Also, what happended to the "Summit Ridge" ? Wasn't it planed to be released alongside AM4 platforms in Q4 2016 ? Is there anything official about a possible delay ? I am waiting for a high-end AM4 motherboard to pair it with an 8C16T Zen CPU and (at least) a Samsung 950 pro M2 SSD. If I go with a good intel board/cpu right now I will exceed by far the 800€ price tag and of course intel i7 prices will drop significantly if AMD ZEN is at least on par with i7-5930k performance.

Full slides from July 15 event

I always said paper launch in 2016 and real launch in 2017. Now Zen slides have "2017" in huge font.

I don't get which is the origin of the belief that Zen will be cheap and "Intel i7 prices will drop significantly". In fact, many enthusiast got disappointed when Intel launched Broadwell-E series. Many were expecting a price drop, but Intel increases prices...

I expect 8C/16T Zen to be somewhat priced between 6C and 8C Broadwell.

 

[ComputerSecurityGuy]

Yeah. Plausibly, Zen will be with the 6C/12T version at the $350ish point where it competes against the i7-6700K, and the 8C/16T probably $700-800. Plausibly $600. Any quad core (defective/disabled cores) would probably be at the $250 or so price point Intel is now leaving wide open since the Skylake Xeons need a C series chipset. But honestly, people, don't expect AMD to price lower than needed or Intel to drop prices unless AMD is really cutting into their margins.

In short, I'm forecasting a return of the Athlon II/Nehalem times.

 

[cdrkf]

I still want to see when the new AM4 motherboards will come to the stores. Also, what happended to the "Summit Ridge" ? Wasn't it planed to be released alongside AM4 platforms in Q4 2016 ? Is there anything official about a possible delay ? I am waiting for a high-end AM4 motherboard to pair it with an 8C16T Zen CPU and (at least) a Samsung 950 pro M2 SSD. If I go with a good intel board/cpu right now I will exceed by far the 800€ price tag and of course intel i7 prices will drop significantly if AMD ZEN is at least on par with i7-5930k performance.

Full slides from July 15 event

I always said paper launch in 2016 and real launch in 2017. Now Zen slides have "2017" in huge font.

I don't get which is the origin of the belief that Zen will be cheap and "Intel i7 prices will drop significantly". In fact, many enthusiast got disappointed when Intel launched Broadwell-E series. Many were expecting a price drop, but Intel increases prices...

I expect 8C/16T Zen to be somewhat priced between 6C and 8C Broadwell.

Well if Zen is fast enough it will be priced against Intel accordingly. If it's slow, it will also be priced accordingly. The 4 core / 8 thread APU is where the value options will be for gamers, assuming sufficient IPC and clock speeds to compete against i5 / i7 quad core parts.

I mean to put it another way, Zen is either fairly slow and cheap, or fairly fast and equally expensive. What the fans want is fast and cheap but there's no way AMD can afford to go that route (as much as it might help boost market share), at least with a large die product like Summit Ridge.

I think where there is an opportunity for a value option will be an APU with the graphics disabled (essentially an updated Athlon for use with a discrete gpu). A quad core, 8 thread Athlon using a salvaged apu die would be something that could be sold for less imo, but that comes later down the road.

 

[ComputerSecurityGuy]

I'm actually thinking the six core Zen is going to be the real consumer hit. Priced around the i7-6700K, it could slice away a large number of people who use i7 chips- editors and streamers most notably. Think the Phenom II X6 vs the i7-920.

 

[gamerk316]

Given that each "module" is four cores, I don't see why so many are speculating on six core parts. Looks like 4/8 to me.

 

[cdrkf]

Given that each "module" is four cores, I don't see why so many are speculating on six core parts. Looks like 4/8 to me.

Why are we assuming that they cannot disable a core within a module? If Summit Ridge has 8 cores, it's quite possible that one / two could be faulty on a die, which would make sense to use for a 6 core part.

I mean AMD were able to disable individual cores within a monolithic quad core with Phenom to offer the various 'x3' models so I don't see a Zen x6 being that far fetched, or is there some technical reason they would have to disable a whole quad core block or nothing? I mean in theory they could probably do 5, 6, 7 core variants if they wanted but I doubt that would make sense due to lack of differentiation.

 

[taseas]

Given that each "module" is four cores, I don't see why so many are speculating on six core parts. Looks like 4/8 to me.

The Bulldozer architrcture is gone for good. There is no module thing now with ZEN. Each one of the 8 cores has its OWN L2 cache, hence it will be fairly easy to disable one of the cores. Even easier of course is to laser cut them, just like Intel ( Intel's i7 5930k is basicly an i7 5960x with 2 cores disabled) .

 

[juanrga]

Given that each "module" is four cores, I don't see why so many are speculating on six core parts. Looks like 4/8 to me.

Indeed, the die has two separate clusters of four-cores each, with independent L3 regions. This unusual design pointed to a granularity of four cores. And that is exactly what we see. The Engineering Samples are only 95W 8-core and 65W 4-core. And the official roadmaps posted just above show only 4-core and 8-core CPUs and 4-core APUs. There is not 2-core APU neither 6-core APU.

I don't get why people continue speculating about non-existent 6-core.

 

[ComputerSecurityGuy]

Ah. Well, in that case, I still think quad core Zen will be at the $200-250ish price mark, since now that Intel has kind of ruined the Xeon E3 deal it could make a decent splash, even though in clocks and IPC the i7 will be superior.

 

[-Fran-]

Given that each "module" is four cores, I don't see why so many are speculating on six core parts. Looks like 4/8 to me.

Indeed, the die has two separate clusters of four-cores each, with independent L3 regions. This unusual design pointed to a granularity of four cores. And that is exactly what we see. The Engineering Samples are only 95W 8-core and 65W 4-core. And the official roadmaps posted just above show only 4-core and 8-core CPUs and 4-core APUs. There is not 2-core APU neither 6-core APU.

I don't get why people continue speculating about non-existent 6-core.

Take a look at how nVidia disabled stuff in the GTX970.

If memory serves right, AMD never announced the Phenom X3 in its long-term roadmaps nor in the short-term ones? It just suddenly appear? I might be wrong, but I don't remember it being mentioned anywhere until it was actually released. Same for the Athlon II Rana cores.

Cheers!

 

[gamerk316]

Given that each "module" is four cores, I don't see why so many are speculating on six core parts. Looks like 4/8 to me.

Indeed, the die has two separate clusters of four-cores each, with independent L3 regions. This unusual design pointed to a granularity of four cores. And that is exactly what we see. The Engineering Samples are only 95W 8-core and 65W 4-core. And the official roadmaps posted just above show only 4-core and 8-core CPUs and 4-core APUs. There is not 2-core APU neither 6-core APU.

I don't get why people continue speculating about non-existent 6-core.

Take a look at how nVidia disabled stuff in the GTX970.

If memory serves right, AMD never announced the Phenom X3 in its long-term roadmaps nor in the short-term ones? It just suddenly appear? I might be wrong, but I don't remember it being mentioned anywhere until it was actually released. Same for the Athlon II Rana cores.

Cheers!

If you recall, AMD fans made it a point to remind everyone the Phenom series of CPUs were "true quad cores", rather then "two dual cores slapped together". This allowed AMD to easily disable a single core to create the X3 series [and why motherboards eventually allowed that core to be reenabled].

Zen seems to have a granularity for four cores, based on the cache design (See Juans comments above). As a result, it's unlikely AMD can disable just a handful of cores in hardware, which points to four and eight core parts only.

Maybe that's why all the marketing only talks about four and eight core parts?

 

[-Fran-]

If you recall, AMD fans made it a point to remind everyone the Phenom series of CPUs were "true quad cores", rather then "two dual cores slapped together". This allowed AMD to easily disable a single core to create the X3 series [and why motherboards eventually allowed that core to be reenabled].

Zen seems to have a granularity for four cores, based on the cache design (See Juans comments above). As a result, it's unlikely AMD can disable just a handful of cores in hardware, which points to four and eight core parts only.

Maybe that's why all the marketing only talks about four and eight core parts?

My point was to say "marketing != technical" and also to remind you all that no one here *really knows* what can or can't be disabled without a hands on of the actual manufacturing and/or detailed design of the core. Just look at how nVidia fused off stuff in GM104 (and now for the 1070). Depending on how you really design the granularity of each component, you can't be certain from just taking a look at an overall design diagram.

Cheers!

 

[gamerk316]

If you recall, AMD fans made it a point to remind everyone the Phenom series of CPUs were "true quad cores", rather then "two dual cores slapped together". This allowed AMD to easily disable a single core to create the X3 series [and why motherboards eventually allowed that core to be reenabled].

Zen seems to have a granularity for four cores, based on the cache design (See Juans comments above). As a result, it's unlikely AMD can disable just a handful of cores in hardware, which points to four and eight core parts only.

Maybe that's why all the marketing only talks about four and eight core parts?

My point was to say "marketing != technical" and also to remind you all that no one here *really knows* what can or can't be disabled without a hands on of the actual manufacturing and/or detailed design of the core. Just look at how nVidia fused off stuff in GM104 (and now for the 1070). Depending on how you really design the granularity of each component, you can't be certain from just taking a look at an overall design diagram.

Cheers!

True, but both the Marketing and Technical seem to be aligning in this particular case. AMD makes no mention of six-core parts in their release plans, and the design of the chip would indicate core counts will be in groups of four. With the technical and marketing in alignment, I don't understand why people are expecting something that almost certainly won't come to be.