AMD May Have Ryzen 16-Core Version and New Chipset in the Works for June

[OneCool]

The quad-core only part launch is still planned, and we might see something like that there.

Might.

I'm not holding my breath at this point

 

[Aenra]

I like to read a lot and as most of you having come across my posts can guess, i'm also not afraid of asking. Both lead to learning
Except end of the day, unlike most here, i am a.. 'consumer' i guess? 'Casual'? This isn't exactly a 24/7 occupational hobby for me. I read on what i feel i need to know; not on everything.
Why am i saying all this?

Because when AMD tells me Ryzen + AM4 mobos + 4 years, i say O.K., Ryzen + AM4 mobos + 4 years.
Now imagine if i had already bought a mobo. Only to hear of this now. Me, i'd be livid. I don't like Intel's "tricks", but at the very least, i always know which way the wind's blowing with them.
This is.. new.

(the above is phrased such due to the type of responses i'm expecting, such as "but this is an HEDT". I get that. I also get that this being a science and not magic, this has been in the works for a long time now. Ergo if the rumor -does- stand, they knew this was coming long before they promised Ryzen + am4 mobos + 4 years)

I'm leaving aside the people who have no choice (budget limited) and as such buy from AMD because that's what they can afford -they don't really support, do they? They're customers out of necessity-.
I'm focusing on the people who actually support them. Those that do, they do for reasons relating to mentality, practices, ethics and the age-old desire to side with the little guy.

Following Intel's 'way' the second the money starts flowing in is not the way to go if you want these supporters. I don't appreciate this at all.
Although yes, happy to know this may be a possibility.

 

[Kanan]

I like to read a lot and as most of you having come across my posts can guess, i'm also not afraid of asking. Both lead to learning
Except end of the day, unlike most here, i am a.. 'consumer' i guess? 'Casual'? This isn't exactly a 24/7 occupational hobby for me. I read on what i feel i need to know; not on everything.
Why am i saying all this?

Because when AMD tells me Ryzen + AM4 mobos + 4 years, i say O.K., Ryzen + AM4 mobos + 4 years.
Now imagine if i had already bought a mobo. Only to hear of this now. Me, i'd be livid. I don't like Intel's "tricks", but at the very least, i always know which way the wind's blowing with them.
This is.. new.

(the above is phrased such due to the type of responses i'm expecting, such as "but this is an HEDT". I get that. I also get that this being a science and not magic, this has been in the works for a long time now. Ergo if the rumor -does- stand, they knew this was coming long before they promised Ryzen + am4 mobos + 4 years)

I'm leaving aside the people who have no choice (budget limited) and as such buy from AMD because that's what they can afford -they don't really support, do they? They're customers out of necessity-.
I'm focusing on the people who actually support them. Those that do, they do for reasons relating to mentality, practices, ethics and the age-old desire to side with the little guy.

Following Intel's 'way' the second the money starts flowing in is not the way to go if you want these supporters. I don't appreciate this at all.
Although yes, happy to know this may be a possibility.

AM4 isn't suited for more than 8 core Ryzen's it's simply not big enough, it's a typical socket suited for the mainstream parts and it is perfect the way it is compared to the consumer line of Core cpus of Intel.
Go look up the new Naples CPUs and you will quickly realize they are twice as big in size, this is because they pack up to 32 cores. This one will have 16 or less cores and 16 or more of them will be disabled, essentially a market segment to make money with (Naples server CPU) garbage that has defective units in it, like the Core i7 LGA2011 line. So it's absolutely fine and has nothing to do with their AM4 stuff - only those who need more than 8 core's OR need more PCI-E lanes for whatever reason (Multi GPU, NVMe, whatever) would want to buy this. It will have double or more PCI-E lanes compared to the mainstream Ryzen line. Anyway, this is just talking rumours, let's see if its true. But I'm pretty sure it is. Reminds me of the Opteron line of Athlon CPUs that were introduced alongside Athlon 64 on a different socket, just that this time it will be made for prosumers and not for a niche server market.

 

[eidairaman1]

Big Ryzen.

May jump to it.

 

[OneCool]

Big Ryzen.

May jump to it.

Same but I'm tired of the ups and downs so far.
I guess it's a waiting game for me right now.

 

[Prima.Vera]

Please don't tell me is going to be a 2.6Ghz CPU...

 

[Kanan]

Please don't tell me is going to be a 2.6Ghz CPU...

2.6 GHz + overclockable = no problem at all. 16 core's means lower clocks, it's not possible to have high clocks and high core count at the same time, not at a set TDP like 140 at least.

 

[Super XP]

I like to read a lot and as most of you having come across my posts can guess, i'm also not afraid of asking. Both lead to learning
Except end of the day, unlike most here, i am a.. 'consumer' i guess? 'Casual'? This isn't exactly a 24/7 occupational hobby for me. I read on what i feel i need to know; not on everything.
Why am i saying all this?

Because when AMD tells me Ryzen + AM4 mobos + 4 years, i say O.K., Ryzen + AM4 mobos + 4 years.
Now imagine if i had already bought a mobo. Only to hear of this now. Me, i'd be livid. I don't like Intel's "tricks", but at the very least, i always know which way the wind's blowing with them.
This is.. new.

(the above is phrased such due to the type of responses i'm expecting, such as "but this is an HEDT". I get that. I also get that this being a science and not magic, this has been in the works for a long time now. Ergo if the rumor -does- stand, they knew this was coming long before they promised Ryzen + am4 mobos + 4 years)

I'm leaving aside the people who have no choice (budget limited) and as such buy from AMD because that's what they can afford -they don't really support, do they? They're customers out of necessity-.
I'm focusing on the people who actually support them. Those that do, they do for reasons relating to mentality, practices, ethics and the age-old desire to side with the little guy.

Following Intel's 'way' the second the money starts flowing in is not the way to go if you want these supporters. I don't appreciate this at all.
Although yes, happy to know this may be a possibility.

Anybody buying Ryzen and a Socket AM4 motherboard today, doesn't make this platform obsolete. If this rumour is true, AM4 will serve as the Low, Mid & High end, along with Mobile. AMD clearly stated that Socket AM4 will be preserved & supported for many years to come. And a future Socket AM4+ to be announced sometime in 2019 to accommodate ZEN 3.

Unless people build PC's every single year, this shouldn't be a deterrent to supporting today's AM4 & Ryzen. Like most, I usually build a new PC from the ground up, every 5-7 years, with minor little upgrades here and there, throughout that time frame.

Intel on the other hand has a nasty habit of making your last years Intel CPU/Mobo as obsolete as possible. AMD is not like that. Let this new AMD Platform serve the Enthusiasts. That are willing to spend $1,000's on a Ryzen and another $500 on a Mobo/chipset.

 

[JMccovery]

Go look up the new Naples CPUs and you will quickly realize they are twice as big in size, this is because they pack up to 32 cores. This one will have 16 or less cores and 16 or more of them will be disabled, essentially a market segment to make money with (Naples server CPU) garbage that has defective units in it, like the Core i7 LGA2011 line. So it's absolutely fine and has nothing to do with their AM4 stuff - only those who need more than 8 core's OR need more PCI-E lanes for whatever reason (Multi GPU, NVMe, whatever) would want to buy this. It will have double or more PCI-E lanes compared to the mainstream Ryzen line. Anyway, this is just talking rumours, let's see if its true. But I'm pretty sure it is. Reminds me of the Opteron line of Athlon CPUs that were introduced alongside Athlon 64 on a different socket, just that this time it will be made for prosumers and not for a niche server market.

Since Naples is a quad-die MCM CPU made of 4 Summit Ridge/Zeppelin dies, and not a monolithic 32-core chip, would there be much of a point in one with half it's cores disabled (which probably wouldn't net a clock increase), over just making a dual-die MCM?

I understand that yeilds may not always be perfect, but disabling all those cores is a lot of silicon to waste.

 

[Kanan]

Since Naples is a quad-die MCM CPU made of 4 Summit Ridge/Zeppelin dies, and not a monolithic 32-core chip, would there be much of a point in one with half it's cores disabled (which probably wouldn't net a clock increase), over just making a dual-die MCM?

I understand that yeilds may not always be perfect, but disabling all those cores is a lot of silicon to waste.

I have two theories about Ryzen HEDT:

a) it's 32 core (8xCCX each 4 cores, half or more disabled), quite unlikely as you pointed out, a very very big chip even compared to LGA2011 CPU's

b) it's a 16 core with 4x CCX, each 4 cores and the defective core's disabled leading to 12 and 8 core Ryzen HEDT as well. Quite likely.

The problem with b) is, if you want to game with it and you're not going for the full chip, which will probably be very expensive, it will be 3/3/3/3 (usable cores) or 2/2/2/2 cores. Whereas 3/3/3/3 is still quite usable even with the crappy Windows scheduler that's senselessly using 2 CCX even if the usage is at 4 cores or lower, the 2/2/2/2 8 core part will be just a mess for gaming, because constant CCX traffic is not avoidable then - the Windows scheduler problems just add to it, but even a perfect scheduler wouldn't be able to cope up for the bandwidth/latency issues with inter-CCX communication then. That said, even the 4/4/4/4 full 16 core chip wouldn't be better for gaming than Ryzen 7, no game needs more than 8 cores anyway.

 

[Super XP]

I have two theories about Ryzen HEDT:

a) it's 32 core (8xCCX each 4 cores, half or more disabled), quite unlikely as you pointed out, a very very big chip even compared to LGA2011 CPU's

b) it's a 16 core with 4x CCX, each 4 cores and the defective core's disabled leading to 12 and 8 core Ryzen HEDT as well. Quite likely.

The problem with b) is, if you want to game with it and you're not going for the full chip, which will probably be very expensive, it will be 4/2/4/2 (usable cores) or 2/2/2/2 cores. Whereas 4/2/4/2 is still quite usable even with the crappy Windows scheduler that's senselessly using 2 CCX even if the usage is at 4 cores or lower, the 2/2/2/2 8 core part will be just a mess for gaming, because constant CCX traffic is not avoidable then - the Windows scheduler problems just add to it, but even a perfect scheduler wouldn't be able to cope up for the bandwidth/latency issues with inter-CCX communication then. That said, even the 4/4/4/4 full 16 core chip wouldn't be better for gaming than Ryzen 7, no game needs more than 8 cores anyway.

I don't recall AMD every mentioning about releasing Ryzen with disabled cores. I am under the impression that AMD will not release CPU's with disabled cores. Because of how there Cache structure is setup. Correct me if I am wrong, but I suppose they will have to release more CPU's to determine if they have locked/disabled cores or not.

 

[Kanan]

I don't recall AMD every mentioning about releasing Ryzen with disabled cores. I am under the impression that AMD will not release CPU's with disabled cores. Because of how there Cache structure is setup. Correct me if I am wrong, but I suppose they will have to release more CPU's to determine if they have locked/disabled cores or not.

The Ryzen 5 and 3 will be all the same (Ryzen 7) chips, just with less core's usable and on Ryzen 3 SMT deactivated as well. AMD confirmed it's 3/3 and 2/2 activated CPUs each CCX, also the whole reason why it's inferior compared to Ryzen 7 at gaming - less CPUs in one CCX = more CCX communication = worse latency = bad for gaming.

 

[OneCool]

I don't recall AMD every mentioning about releasing Ryzen with disabled cores. I am under the impression that AMD will not release CPU's with disabled cores. Because of how there Cache structure is setup. Correct me if I am wrong, but I suppose they will have to release more CPU's to determine if they have locked/disabled cores or not.

It's what everyone is assuming ..Myself included. Also the most cost effective for AMD.

Always been the case really.

 

[Super XP]

It's what everyone is assuming ..Myself included. Also the most cost effective for AMD.

Always been the case really.

Interesting.
You would think AMD would be very careful to ensure nothing hinders Ryzen's performance. Putting aside the Bios Updates needed and the various Windows 10 optimization updates needed.

These Ryzen 3 and 5 CPU's are suppose to be direct competitors to i5's and i3's. And are suppose to OC better too.

 

[theGryphon]

Why would they do 2+2+2+2 for an 8-core part? It defies logic. 4+4+0+0 would be the way to go, especially given the cache structure, not only from performance pov but also power savings. Similarly, 4+4+4+0 for 12 core.

 

[Super XP]

Why would they do 2+2+2+2 for an 8-core part? It defies logic. 4+4+0+0 would be the way to go, especially given the cache structure, not only from performance pov but also power savings. Similarly, 4+4+4+0 for 12 core.

Agreed.
Because it would mess up it's cache latency and speed.

 

[Kanan]

Why would they do 2+2+2+2 for an 8-core part? It defies logic. 4+4+0+0 would be the way to go, especially given the cache structure, not only from performance pov but also power savings. Similarly, 4+4+4+0 for 12 core.

No, the logic behind it is quite right. At first I thought the same, but then AMD revealed Ryzen 5 are 3+3 and 2+2 cores. And it made sense, because that way you can release a CPU with a defective CPU (or more) in each CCX - the same logic probably holds true for the bigger parts with 4 CCX and more. It's simply about yield and maximum profit. The other option would've been to go for a completely free approach, but that would mean Ryzen would vary in performance and AMD said that's not going to happen "no silly games".

So when dealing with a four-core or six-core CPU, and the base core design has eight-cores, how does AMD cut them up? It is possible for AMD to offer a 4+0, 3+1 or 2+2 design for its quad-core parts, or 4+2 and 3+3 variants for its hexacore parts, similar to the way that Intel cuts up its integrated graphics for GT1 variants. The downside with this way is that performance might differ between the variants, making it difficult to manage. The upside is that more CPUs with defects can be used.

We have confirmation from AMD that there are no silly games going to be played with Ryzen 5. The six-core parts will be a strict 3+3 combination, while the four-core parts will use 2+2. This will be true across all CPUs, ensuring a consistent performance throughout.

http://www.anandtech.com/show/11202/amd-announces-ryzen-5-april-11th

 

[theGryphon]

For Ryzen 5 parts, where prices/margins are low, it makes sense from business perspective to do 3+3 and 2+2. It's also a segment that strictly prioritizes price before performance.

However, for these new HEDT parts, the story is totally different. Margins are higher and the market is performance-centric. So, for everybody's (including AMD) sake, I hope they don't go 2+2+2+2 or 3+3+3+3. This is unless the architecture requires a balanced distribution across the included CCX's...

 

[Kanan]

If you want to use all L3 caches, all CCX have to be activated too. If not, you can disable them I guess, but I'm not entirely sure if it's possible either.

It has nothing to do with HEDT (in my books Ryzen 7 is HEDT anyway, just with fewer PCI-E lanes) or not, it's about maximum profit and nothing else. I already explained why.

 

[theGryphon]

If you want to use all L3 caches, all CCX have to be activated too. If not, you can disable them I guess, but I'm not entirely sure if it's possible either.

It has nothing to do with HEDT or not, it's about maximum profit and nothing else. I already explained why.

Nobody is discussing the greedy business sense of going 2+2+2+2 over 4+4+0+0, and they will probably do that. I merely expressed my hope

In any case, I don't think I saw a scientific review comparing 2+2 vs 4+0 in terms of performance and power usage. We are very reasonably assuming that the latter should be better, but by how much?

 

[Kanan]

Nobody is discussing the greedy business sense of going 2+2+2+2 over 4+4+0+0, and they will probably do that. I merely expressed my hope

In any case, I don't think I saw a scientific review comparing 2+2 vs 4+0 in terms of performance and power usage. We are very reasonably assuming that the latter should be better, but by how much?

I saw someone doing it and the difference was quite high, ~10-15 fps in a high fps scenario of over 120 fps in a modern game. Now just to find that source and exact info again ....

edit:
I found this at least, not really the source I was speaking of but better than nothing:
http://www.phoronix.com/scan.php?page=article&item=amd-ryzen-cores&num=2

 

[Super XP]

I saw someone doing it and the difference was quite high, ~10-15 fps in a high fps scenario of over 120 fps in a modern game. Now just to find that source and exact info again ....

edit:
I found this at least, not really the source I was speaking of but better than nothing:
http://www.phoronix.com/scan.php?page=article&item=amd-ryzen-cores&num=2

Those are some strange benchmark results. In a couple scenarios, you would thing some setups in theory would do better than others, but that is not the case in the link you provided. Interesting. Probably has to do with Motherboard Bios's that need updates.

 

[coolernoob]

I bet reviews will devote 90% of testing only for gaming on 1080p with top end GPUs (TPU included) and dudes with system like: i7-7700K and gtx 1060 Gaming X saying something like: "see - intel gets like 200 frames but this only 185 frames (on gtx 1080 TI watercooled on 1080p) - not good - so FAYL amd, again"

 

[Kanan]

Those are some strange benchmark results. In a couple scenarios, you would thing some setups in theory would do better than others, but that is not the case in the link you provided. Interesting. Probably has to do with Motherboard Bios's that need updates.

True, nvm, Ill try to find my source again later.

 

[PowerPC]

Meanwhile, an i3 can be easily overclocked to 5,3 Ghz... although this is only much better for gaming, but still.