Friday, December 1st 2017

AMD Second-generation Ryzen "Pinnacle Ridge" Confirmed to Support AM4

AMD, in an interview with Overclockers UK (OCUK), confirmed that its second-generation Ryzen desktop processors will support the existing AM4 socket, so current Ryzen platform users can seamlessly upgrade to the new processors, with a BIOS update. Most current AM4 socket motherboards will require BIOS updates to support Ryzen "Raven Ridge" desktop APUs, and Ryzen "Pinnacle Ridge" CPUs, as the two require an update to the latest AGESA 1.0.0.7 version. In the interview, AMD representative James Prior confirmed that the company plans to keep AM4 its mainstream-desktop processor socket all the way up to 2020, which means at least another two to three generations of processors for it.

The next generation is "Pinnacle Ridge," which is rumored to be an optical-shrink of the "Summit Ridge" silicon to the 12 nm process, enabling higher clock speeds. The decision to keep AM4 doesn't mean the company's 300-series chipset will be made to stretch over 3 years. The company could release newer chipsets, particularly to address 300-series chipset's main shortcoming, just 6-8 older PCI-Express gen 2.0 general purpose lanes (while Intel chipsets put out up to 24 gen 3.0 lanes).

Source: OCUK (Facebook)
Add your own comment

144 Comments on AMD Second-generation Ryzen "Pinnacle Ridge" Confirmed to Support AM4

#1
trparky
newtekie1 said:
It only provides x8 2.0, not 3.0.
Wait. What? Where are you seeing that AMD is still using PCI Express v2.0?
Posted on Reply
#2
EarthDog
For its chipset lanes. NOT for the communication line between CPU and chipset.

See chipset block diagram somewhere in this thread for details. :)
Posted on Reply
#3
Xzibit
newtekie1 said:
We aren't talking about the link to the chipset being the problem, the problem is the chipset itself. AMD's link to the chipset is PCI-E x4 3.0. However, the chipset itself only provides 8 PCI-E 2.0 lanes for the system to use for connected devices. This is not enough.
Its the same bandwidth to saturate the Cpu-to-chipset. If your adding more then it can transfer its for connectivity options.

Your going to have to pick and choose which connections are priority. You obviously cant have more data being shoved to the Cpu then it can handle at x4 3.0.
Posted on Reply
#4
notb
Xzibit said:
Its been several gens that Intel MB OC on their own or as simple as a click/selection from different board members.
I meant manual adjustments. Auto OC works within manufacturers' limits. PCs are designed for this kind of treatment.
The whole idea of manual OC is to go beyond these limits.
If your not running the latest connections a 1150 is still fine. Heck some here still run 775. Every once in awhile you'll see a member asking for parts in the forum that he cant find because stores don't stock nor keep things after 2 gens if they don't sell, Maybe its different in Poland but here in the states after the new gen is released most chains get rid of last gen stock. Plus most of the stuff online is from 3rd party vendors that get put up close to initial selling price.
The whole idea of shopping for consumer electronics in Poland is different (and not just electronics). We don't use local shops that much. The market is dominated by online stores and auctions, so they can keep stock for a long time (and quickly import something rare).
So it's much like with amazon and ebay in US. The difference is: if you don't like buying online and you'd like to buy something in a local PC store in Poland... you can't. It went bankrupt 10 years ago.

So yeah... 1150 boards are quite easy to get. 775 is much harder, but 2 years ago it wasn't that bad. I've even seen some Asus P5Q and other high-end models available back then (new, sealed boxes). Usually it's the mid-range "business" stuff that stays in shops for longer (like ASUS CSM lineup).
The sillyness is people don't want others to have that longevity of a socket.
Why is that silly?
I am fine with you being able to get a motherboard for many years. But I'm hardly happy about the fact that it probably slows the development cycle and - more importantly - raises the price of my mobo.
Socket 775 was 2006. You bought 4yrs into it and it lasted you 7 years. Let that sink in 11yrs. Yet some how find a way to complain about a socket projected to last from 2017-2020. That's 3 years. Remember you bought into socket 775 4yrs after it was released. :kookoo:
Actually socket 775 was introduced in 2004. Moreover, 1156 came out in 2009, so in 2010 I bought an obsolete, 6-year-old platform. But it was very cheap and I needed just that. I was using a notebook most of the time, anyway. But few years later it became my main PC and managed pretty well.

The fact that I used it for 7 years has nothing to do with longevity. I've doubled RAM at some point and replaced disks, but all other parts were from the original build.
I didn't need a faster desktop and that one didn't want to break down.

The point is: it doesn't matter how long-lasting a platform is. I simply buy whatever matches my needs at particular moment and I use it for as long as possible. This is a typical, most common consumer behavior and it strongly favors a quicker replacement cycle.
Posted on Reply
#5
newtekie1
Semi-Retired Folder
trparky said:
Wait. What? Where are you seeing that AMD is still using PCI Express v2.0?
Everywhere that lists the specs for AMD's chipsets. If you want, look a few posts up, I post AMD's slides that show it as well as block diagrams that also show it.

Xzibit said:
Its the same bandwidth to saturate the Cpu-to-chipset. If your adding more then it can transfer its for connectivity options.

Your going to have to pick and choose which connections are priority. You obviously cant have more data being shoved to the Cpu then it can handle at x4 3.0.
Of course you can't, but that is what the chipset is for. That is why Intel puts out 24! PCI-E 3.0 lanes from their chipset, as well as all the SATA connections, all with a PCI-E x4 3.0 link back to the CPU. It works just fine.

AMD just made their chipset too shitty, that is what it comes down to. The chipset is where they need to improve, not the CPU and its socket.
Posted on Reply
#6
Xzibit
newtekie1 said:

Of course you can't, but that is what the chipset is for. That is why Intel puts out 24! PCI-E 3.0 lanes from their chipset, as well as all the SATA connections, all with a PCI-E x4 3.0 link back to the CPU. It works just fine.

AMD just made their chipset too shitty, that is what it comes down to. The chipset is where they need to improve, not the CPU and its socket.
If you understand how it works. Have you read your manual lately?

Look for the * and the shared bandwidth between devices. If the manual is comprehensive it will let you know what turns off or what downgrades to what when certain slots are used especially on the higher boards.
Posted on Reply
#7
newtekie1
Semi-Retired Folder
Xzibit said:
If you understand how it works. Have you read your manual lately?

Look for the * and the shared bandwidth between devices. If the manual is comprehensive it will let you know what turns off or what downgrades to what when certain slots are used especially on the higher boards.
The 24 PCI-E lanes is 24 PCI-E lanes, they are never disabled. They are shared between things like the M.2 slots and such, but it is still providing 24 functional PCI-E 3.0 lanes that can all be used at the same time.

The Z370 Taichi for example, has 3 M.2 slots that all run at PCI-E x4 3.0, as well as two PCI-E x1 3.0 slots, a PCI-E M.2 x2 slot for the wirless card. These can all be used at the same time. And there are even more devices than that running off the PCI-E lanes from the chipset. The sound card runs off a PCI-E lane, as well as the second 1Gb/s network adapter, and the AsMedia USB3.1 controller. All running off the PCI-E lanes provided by the chipset, with no lane sharing.

The only thing shared is the SATA ports with the M.2 slots. If SATA based M.2 drives are used, then some of the normal SATA ports are disabled, because the M.2 slots share SATA connections with the standard ports. But that is a limitation of the number of SATA ports on the chipset, not PCI-E lanes. None of the PCI-E lanes on the Z370 Taichi are shared, because the Z370 chipset offers 24 fully functional PCI-E 3.0 lanes.

So, yes, I know exactly how it works, and there is no bandwidth sharing between devices.
Posted on Reply
#8
EarthDog
The SATA ports disable on this board when m.2 devices are in use.
*M2_1, SATA3_0 and SATA3_1 share lanes. If either one of them is in use, the others will be disabled.
M2_2, SATA3_4 and SATA3_5 share lanes. If either one of them is in use, the others will be disabled.
If M2_3 is occupied by a SATA-type M.2 device, SATA3_3 will be disabled.
Boards with that much connectivity have to share the bandwidth.
Posted on Reply
#9
newtekie1
Semi-Retired Folder
EarthDog said:
The SATA ports disable on this board when m.2 devices are in use.



Boards with that much connectivity have to share the bandwidth.
I said that. They aren't sharing PCI-E bandwidth though, they are sharing SATA connections from the chipset, there are only 6 of those to go around from the chipset.

But we aren't talking about SATA ports! How is this hard to understand?

Intel's chipset provides 24 PCI-E 3.0 Lanes! They aren't shared, some aren't disabled. The Z370 chipset provides 24 fully functional PCI-E 3.0 lanes!

AMD's chipset only provides 8 PCI-E 2.0 lanes. Do you all see where the problem is? Is it that hard to figure out?
Posted on Reply
#10
EarthDog
To be more clear, you are specifically talking CPU connected PCIe Lanes (16 in Coffee Lake). Because the lanes attached to the chipset, are also PCIe lanes as you know and have said. So do understand my confusion. When I see this:
newtekie1 said:
All running of the PCI-E lanes provided by the chipset, with no lane sharing.
...do understand why I brought up SATA because some will disable when M.2 is used... and they use the PCIe lanes provided by the chipset (24).
Posted on Reply
#11
newtekie1
Semi-Retired Folder
Ok, lets recap and then I'm done, because I feel like I'm talking to brick walls here.
  • The original claim, by TheLostSwede, was that AMD can't improve the AM4 platform without changing the socket.[list]
  • I pointed out that this is false because they use the same x4 link to the chipset that Intel does, so there is no reason AMD couldn't add more PCI-E lanes and SATA ports to their chipset without the need for a socket change and the problem is AMD didn't put enough lanes on the chipset.
  • Then TheLostSwede tried to say everyone was talking about the CPU lanes.
    • I pointed out that Ryzen has 4 more CPU lanes than Z370, so that is also wrong if that was what he was talking about.
    • I also pointed out that everyone up until his post was talking about chipset lanes. Btarunr even says AMD chipset shortcomings.
  • Then there is a big discussion about the actual number of lanes for each platform, with btarunr posting an inaccurate chart showing AMD actually having more lanes than they really do.
    • I cleared that up.
    • Intel has 20 PCI-E 3.0 lanes provided by the CPU, with 4 used for the link to the chipset. Then the chipset provides another 24 PCI-E 3.0 lanes.
    • AMD has 24 PCI-E 3.0 lanes provided by the CPU, with 4 used for the link to the chipset. Then the chipset provides another 8 PCI-E 2.0 lanes.
  • Then for some reason Xzibit chimed in and said that you can't actually use all 24 PCI-E lanes on the Intel chipset at the same time, that they are for some reason shared and some lanes are disabled when others are in use.
    • This is completely false. All 24 lanes provided by the chipset are able to be used at the same time.
    • The only thing that is shared is SATA ports, as the Intel chipset only provides 6 SATA ports. So if a SATA M.2 is inserted, it disabled SATA ports. This is not in any way related to the number of PCI-E lanes provided by the chipset.
    [/list]In conclusion, and I'm only going to say this this last time, if people don't get it, I guess too bad; AMD's problem is their shitty weak chipset. They need to increase the number of PCI-E lanes it provides, as well as make them PCI-E 3.0. This can all be done by updating the chipset alone, it does not require a new socket or even CPU. The number of lanes coming from the CPU is adequate, and the PCI-E x4 link between the CPU and the chipset is more than sufficient to handle a chipset with a lot more PCI-E lanes than AMD's X370 currently provides.
  • Posted on Reply
    #12
    Mussels
    Moderprator
    AMD actually told me this on facebook, they've been pretty public about the 2020 goal for AM4.

    Its why i went ryzen - when a better clocking chip comes out, this goes to my wife and i get an upgrade - no intel shenanigans required.
    Posted on Reply
    #13
    Vlada011
    Customers like to here such things.
    At least bigger part of customers.
    They don't like when their chipset and special socket is outdated after 12 months.
    But with Intel chipset and socket are same, they don't want to allow support even when motherboard BIOS update could fix everything. Than new chipset on same socket mean new motherboard if you want to upgrade CPU only.
    That's game when Intel make favor to motherboard manufacturers and they make favor to Intels.
    How? Agressive advertasing of extremely nice motherboards force you on upgrade for 10% improvement.
    X99>X299. And people upgrade even if no reason.
    They would not upgrade for 10% better CPU on same motherboard, but if you present new generation with new name, some little details they are hypnotized. And Intel allow to motherboards sell two motherboard instead to customers use 2 CPU on same mobo.
    Because of that AM4 will be "working horse" until 2020 and X99 will be mine working horse until 2020 and DDR5. No reason to buy 400$+ motherboards if you can't use two generation of CPU and models with more cores than first generation.
    Posted on Reply
    #14
    GoldenX
    I don't understand why some people say this is bad, that it slows development. AMD said AM4 last until 2020, not that Z370 lasts until 2020, they can and more certainly will release newer better chipsets. You as the user have the option of keeping the motherboard, or change it for a new one.
    Posted on Reply
    #15
    newtekie1
    Semi-Retired Folder
    GoldenX said:
    I don't understand why some people say this is bad, that it slows development. AMD said AM4 last until 2020, not that Z370 lasts until 2020, they can and more certainly will release newer better chipsets. You as the user have the option of keeping the motherboard, or change it for a new one.
    Agreed, sticking with the same socket isn't a problem. I think a lot of people seem to think the socket is AMD's weak link, but the chipset is, and that can be updated without changing the socket.
    Posted on Reply
    #16
    notb
    GoldenX said:
    I don't understand why some people say this is bad, that it slows development. AMD said AM4 last until 2020, not that Z370 lasts until 2020, they can and more certainly will release newer better chipsets. You as the user have the option of keeping the motherboard, or change it for a new one.
    First of all: changing sockets frequently raises Intel's earning! (isn't that obvious? :-))
    Second: new socket means new motherboards. All vendors will propose something, with all the latest tech they can include. Does this stimulate mobo evolution? Maybe yes, maybe not. It sure doesn't hurt.
    Look at AMD AM3+ lineup from before Ryzen came out. Sure, there were some new motherboards, but quite a lot of "current" models were really long in the tooth.
    Posted on Reply
    #17
    GoldenX
    notb said:
    First of all: changing sockets frequently raises Intel's earning! (isn't that obvious? :))
    Second: new socket means new motherboards. All vendors will propose something, with all the latest tech they can include. Does this stimulate mobo evolution? Maybe yes, maybe not. It sure doesn't hurt.
    Look at AMD AM3+ lineup from before Ryzen came out. Sure, there were some new motherboards, but quite a lot of "current" models were really long in the tooth.
    We hat tons of chipsets for socket 775, even from different vendors like VIA, ATI or Nvidia, starting from DDR1 and AGP all the way to DDR3 and PCI-E 2.0, it's processor performance that limits innovation, not the socket. FX was DOA, there was no incentive to make good motherboards for it. With Ryzen giving good performance, core count and for example M.2 to the masses, we can expect better products in the future.
    I've said it before, look at the A320 chipset and compare it to the B250 or H110 chipset, suddenly a lot better.

    Of course Intel makes more money by making their platforms a closed garden, they are one step away of being Apple, I'm sure we would love that. In fact they hinted at that with Broadwell.
    Posted on Reply
    #18
    johnnyfiive
    efikkan said:
    And why would anyone upgrade a one year old system with a new one? Even if they sell the old CPU, it's still a waste of money. Intel chooses to bring new platform features over prioritizing those 0.1% of buyers who want to upgrade to every new iteration. In reality nearly everyone keeps motherboard, CPU and RAM "bundled together" throughout the lifespan of a system. Graphics cards, SSDs, HDDs, etc. are on the other hand easy to swap independently.


    Why are you sugar-coating it?
    Vega is the largest failure in many years for AMD, and there is no reason to buy it for gaming. So when a product is inferior, the fans keep focusing on theoretical specs over actual performance…
    I gotta disagree with ya there. If you're a fan of FreeSync and are familiar with how AMD does their drivers (performance gains over time), then the Vega cards are pretty good cards (At their normal pricing). Both Vega 56 and 64 are good buys at $399 and $499, especially if you have a FreeSync monitor. Last I checked the Vega 64 is trading blows with the 1080, and they're about the same price. The failure part was allowing these cards to be so scare that sellers could charge $100-$200 over MSRP. That's the biggest failure with Vega, the pricing bullshit. The actual product, is pretty good (IMO).
    Posted on Reply
    #19
    GoldenX
    efikkan said:
    Why are you sugar-coating it?
    Vega is the largest failure in many years for AMD, and there is no reason to buy it for gaming. So when a product is inferior, the fans keep focusing on theoretical specs over actual performance…
    You can say the same about the Titan cards, that gaming performance at that price is totally stupid.
    Posted on Reply
    Add your own comment