How do you think AMD will approach the lower market segment with Zen for AM4 desktop?

[cbn]

Beyond 8C/16T, 6C/12T and 4C/8T for Summit Ridge and 4C/8T for Raven Ridge, what do you think AMD will release for the lower market segments?

4C/4T?

3C/6T?

2C/4T?

Also, if you want to speculate on motherboards please do so. I am particularly interested in what the Small form factor "chipset-less" X300/B300/A300 series might bring?

 

[cbn]

As far as the release of parts go, here is what I think is likely to happen:

Summit Ridge: 8C/16T and 6C/12T released initially in Feb. 2017 (with 4C/8T released sometime after.....perhaps around the time Raven Ridge for mobile launches)

Raven Ridge (released a few months after Summit Ridge?): Best processor dies (4C/8T with full and 75% iGPU) focused on BGA mobile....with the more strongly harvested parts (and leftover full dies that didn't get used for mobile ) for AM4 released sometime afterwards.

So 8C/16T and 6C/12T in Feb. 2017 with 4C/8T Summit Ridge around the time Raven Ridge for mobile launches. This followed by various harvested (or not) Raven Ridge parts for AM4.

 

[Abwx]

Also, if you want to speculate on motherboards please do so. I am particularly interested in what the Small form factor "chipset-less" X300/B300/A300 series might bring?

If that was to be chipset less there wouldnt be three naming scheme since there s only one kind of integrated PCH in the CPUs, more probably that this is A320 cut down versions with fewer USB2/3, no Raid and eventually less PCI lanes.

 

[cbn]

If that was to be chipset less there wouldnt be three naming scheme since there s only one kind of integrated PCH in the CPUs, more probably that this is A320 cut down versions with fewer USB2/3, no Raid and eventually less PCI lanes.

I'm thinking Summit Ridge will have different SoC I/O than Bristol Ridge....so I am betting the boards themselves will have different levels of I/O as well.

So, for example, a person could use a Bristol Ridge in a X300....but it might be some of the board's I/O is unusable as it is optimized for Summit Ridge?

P.S. I am speculating on X300 being designed for Summit Ridge I/O based on the X370 being used for the high end Summit Ridge AM4 chipset boards.

EDIT: For my own particular AM4 build-up I am hoping to build a very space efficient/thermally efficient/(relatively) energy efficient SFF desktop machine. This with a fully or semi-modular PSU and a single M.2 SSD on the front on the motherboard*. With the minimal storage and modular PSU I would only need the following cables : 24 pin ATX, 4 or 8 pin CPU, 6/8 pin PCIe power connector(s) for video card.....which should help the airflow inside a small case. Case fan (s) will be powered off motherboard header(s). Video card will use a blower design and receive cold air directly from outside the case (via vents in the enclosure).

*Intel LGA 1151 Mini-ITX have the M.2 SSD on the back of the motherboard.

 

[Phynaz]

I have no idea, except that it will be at least a year from now before the low end parts are introduced. I.E. Athlon and Sempron.

 

[Abwx]

I'm thinking Summit Ridge will have different SoC I/O than Bristol Ridge....so I am betting the boards themselves will have different levels of I/O as well.

So, for example, a person could use a Bristol Ridge in a X300....but it might be some of the board's I/O is unusable as it is optimized for Summit Ridge?

P.S. I am speculating on X300 being designed for Summit Ridge I/O based on the X370 being used for the high end Summit Ridge AM4 chipset boards.

.

If we look at Summit Ridge I/O it should work even with the minimalistic chipset, this latter is necessary since some functionalities like LAN are not integrated in the FCH in both Summit and Bristol Ridges..

 

[superstition]

AMD would do well to offer a K i5 competitor right from the start (and, of course, some part that does its best to compete with i7 6700K), even if it has limited availability. So that means something more cut down than 8C/16T.

I also wonder what the viability of 8C 8T would be as well as 6C 6T.

The key, for targeting mainstream enthusiasts, will be being able to keep up clockspeed and IPC with just enough threads to power a gaming experience without stutter. The more cores/more threads marketing angle is going to be less important for most buyers than the overall result. If a 6 thread part is the most optimal then maybe something like 6C 6T might be seen. I doubt a tricore is going to be the most optimal.

Summit Ridge should be soldered and unlocked.

In terms of the low end that's Bristol Ridge, right?

Maybe in the future a Zen tricore. It would be a very cool marketing angle if AMD could work with Nintendo to release a coupon for a PC Zelda game as a promotion with a tricore processor. Triforce on the box, and as the chip's name, would get a lot of interest. It would need to have a recommended controller, though, since Zelda games are action titles and a keyboard doesn't cut it. Alternatively, Nintendo could make a FFVII-type RPG with the Zelda franchise, one aimed more at older gamers (less kiddie).

 

[VirtualLarry]

Maybe in the future a Zen tricore. It would be a very cool marketing angle if AMD could work with Nintendo to release a coupon for a PC Zelda game as a promotion with a tricore processor. Triforce on the box,

I like that idea. But has Nintendo ever done any PC games?

 

[superstition]

I like that idea. But has Nintendo ever done any PC games?

I'm sure they could find a studio to contract with. I'd be happy to be an executive producer. : )

It would be nice to have a deep exploration-heavy fairly non-linear Zelda turn-based RPG game that's a bit more mature-oriented than a Final Fantasy but not in terms of blood and guts (mainly just sophistication). The PC games industry is too dominated by angst and violence. While Zelda games are always pretty heavy on combat, there is an opportunity to do more than just hack and slash given the pretty much unlimited storage capacity of current PC systems. There is room in the market for something more wistful and dreamlike, like the first Chrono Trigger. Link isn't your dime-a-dozen angry/scarred "heroic male lead".

There would be a lot of publicity to be gained with the chan crowd in particular, as well as old-timers like me who grew up playing the NES game.

 

[DrMrLordX]

Beyond 8C/16T, 6C/12T and 4C/8T for Summit Ridge and 4C/8T for Raven Ridge, what do you think AMD will release for the lower market segments

Undoubtedly, something which does not meet with your approval.

I like that idea. But has Nintendo ever done any PC games?

I don't think they have, and with their lurch towards the mobile segment, I would not expect it any time soon, if ever.

 

[cbn]

Regarding the situation of desktop (65W socket AM4) vs. mobile (35W BGA) for Raven Ridge APU here is my reasoning on the following core and iGPU configurations (assuming the Raven Ridge memory controller will handle dual channel DDR4 3200):

1.) Quad core Raven Ridge with 100% iGPU: Should definitely be 35W BGA only (re: even at 35W a full size Raven Ridge iGPU would be bottlenecked by dual channel DDR4 3200)

2.) Quad core Raven Ridge with 75% iGPU: Should ideally be 35W BGA only.

3.) Triple core Raven Ridge with 75% iGPU: Not sure, Could be 35W BGA or 65W AM4.

4.) Dual core Raven Ridge with 75% iGPU: Should definitely be 65W AM4 only (re: even at 35W BGA a 2C/4T with large iGPU might not have enough CPU frequency (with the iGPU active) to keep stuttering at bay in all gaming scenarios)

 

[cbn]

Just thinking about this more......Perhaps if the Zen APUs for AM4 desktop were confined to triple and dual core only the motherboard BIOS could be universally configured to not throttle the CPU when the iGPU is active. This would have the benefit of helping the dual and triple CPUs work better with either DX12 multi-adapter or Hybrid crossfire.

 

[cbn]

If we look at Summit Ridge I/O it should work even with the minimalistic chipset, this latter is necessary since some functionalities like LAN are not integrated in the FCH in both Summit and Bristol Ridges..

If looking at Summit Ridge diagram it does list the option for 2 General purpose PCIe 3.0 lanes "under processor direct storage" . This could be used for LAN and Wifi (with PCIe 3.0 x 2 left over for M.2 SSD)

 

[cbn]

Regarding the current dGPU market situation where the total number of cards sold is decreasing but the price per card is increasing, I do wonder how much of that is due to improving iGPUs vs. relative lack of CPU processing power at X price point.

For example, taking at look at the following Intel processors (which make up a large part of the Windows Installation base):

Intel Sandy Bridge Core i3 2100 (3.1 Ghz, $117 to $120, Q1 2011)

http://ark.intel.com/products/53422/Intel-Core-i3-2100-Processor-3M-Cache-3_10-GHz

Intel Skylake Core i3 6100 (3.7 Ghz, $117, Q3 2015)

http://ark.intel.com/products/90729/Intel-Core-i3-6100-Processor-3M-Cache-3_70-GHz

We can see that clockspeed increased 600 Mhz over the span of 4.5 years. During this time IPC increased 25% according to Anandtech:

http://www.anandtech.com/show/9483/intel-skylake-review-6700k-6600k-ddr4-ddr3-ipc-6th-generation/23

Sandy Bridge, Your Time Is Up.

A large number of users invested into Intel based platforms during the Core 2 Quad, Nehalem and Sandy Bridge releases. Sandy Bridge was notable because it inferred a large performance gain at stock speeds, and with a good processor anyone could reach 4.7 GHz and even higher using a good high end cooler. With that, Intel has had a problem enticing these users to upgrade because their performance has been constantly matched by Ivy Bridge, Haswell and Broadwell – for every 5% IPC increase from the CPU, an average 200 MHz was lost on the good overclock and they would have to find a good overclocking CPU again. There was no great reason, apart from chipset functionality to upgrade.

That changes with Skylake.

From a clock-to-clock performance perspective, Skylake gives an average ~25% better performance in CPU based benchmarks

Combining the jump in clockspeed of 600 Mhz (3.1 Ghz to 3.7 Ghz) with the 25% increase in IPC the $117 tier of Core i3 has increased single thread only 50% in the span of those 4.5 years.

Meanwhile GPU performance (across all price points) has increased much more than 50% in the same time span.

Therefore it definitely seems to me we have a situation where the GPU compute power at X price point is outpacing the lower CPU compute power at X price point. The result of this would most likely result in a decrease in dGPU sales as the pool of CPUs necessary to feed the newest generation of dGPUs shrinks every year.

With this mentioned, I do wonder what AMD will be able to offer at the $117 price point in order to boost the pool of CPUs capable of keeping up withe the ever improving range of midrange dGPUs?

 

[cbn]

Looking at the die shot of Summit Ridge below, I even wonder if AMD could use chop lines (and a new mask set) to create a new smaller quad core die? This using the bottom half of the die.

It would only have single channel memory though, but if it is actually feasible it would be a way to the increase the pool of 4C/8T and 3C/6T Zen for the ever improving range of mid-level and better dGPUs.

 

[TheELF]

Therefore it definitely seems to me we have a situation where the GPU compute power at X price point is outpacing the lower CPU compute power at X price point. The result of this would most likely result in a decrease in dGPU sales as the pool of CPUs necessary to feed the newest generation of dGPUs shrinks every year.

Are you serious?
Take a look at the gaming market,almost every big budged game is made for the consoles,that's a celeron level CPU with a midrange GPU.
There is no need for huge CPU compute power because there simply aren't any games out there that could use it,unless you are trying for 144Hz but that's clearly not the tier you are talking about.
And Dx12 will,when/if it finally takes off,lower the CPU needs even more,a i3-6100 is able to drive even high end cards in most of the available games.

 

[cbn]

Are you serious?
Take a look at the gaming market,almost every big budged game is made for the consoles,that's a celeron level CPU with a midrange GPU.
There is no need for huge CPU compute power because there simply aren't any games out there that could use it,unless you are trying for 144Hz but that's clearly not the tier you are talking about.
And Dx12 will,when/if it finally takes off,lower the CPU needs even more,a i3-6100 is able to drive even high end cards in most of the available games.

Consoles have better optimization than PCs so even if they have relatively weak octocore Jaguar CPUs it is hard to directly compare to PC dual core.

However, with that mentioned I do agree that a dual core can drive a good many games.....but I doubt it would be able to do so without stuttering if the GPU were big enough.

Take, for example, these results from AtenRA using Battlefield 4 64 player using Mantle (a low level graphics API) and HD7950 @ 1000 Mhz:

Even a humble FX-8150 (that's Bulldozer, not Vishera) beats the i3-6300 (which is 100 Mhz faster than i3-6100) in terms of smoothness (and frame rate). Granted, the i3-6300 is using DDR3 1600 (not DDR4 2133 or faster)....but I still feel it gives us an idea of where the trend is going. With GPUs faster than HD7950 @ 1000 Mhz (eg, GTX 1060 or RX 480) I would expect the stuttering/smoothness problem to only get worse.

P.S. Back when the i3-2100 was released a midrange Nvidia GPU would have been the GTX 460, but by the time i3-6100 was released the same mid-level GPU would have been a GTX 960. That is pretty large increase in GPU relative to the 50% increase in single thread going from i3-2100 to i3-6100.

 

[TheELF]

Take a left towards the thread that talks about AMD DRIVER CPU OVERHEAD,after that go and search about bad console porting,and then look at a few sites that show how much slow mem bottlenecks the skylake i3s.

Consoles have better optimization than PCs so even if they have relatively weak octocore Jaguar CPUs it is hard to directly compare to PC dual core.

If your original point was that we need more/better optimization to run mid/highrange cards then I'm totally with you.

 

[TheELF]

That is pretty large increase in GPU relative to the 50% increase in single thread going from i3-2100 to i3-6100.

Yeah and most people think that 4k is the new res they have to play at...