AMD Raven Ridge 'Zen APU' Thread

Page 25 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

DisEnchantment

Golden Member
Mar 3, 2017
1,687
6,243
136
I thought AMD had said as much but it's darn near fact that any Zen based CPU will be using Vega. It's one of the biggest points of them developing IF.

2M3001C3T4MF2_33/30_N + AMD 15DD
2M2000C4T4MF2_33/20_N + AMD 15DD
2M1901C4T4MF2_30/19_N + AMD 15DD

All the 15DD are Raven IDs. Checking the sources it is pretty clear Vega and Raven use GFX9.
I checked out out the agd5f repo, now browsing the branch drm-next-4.13-wip

Some Info from the sources as of this commit 9aafbfcdead1f9f7d6eaade6f87c86b7069815a2
- Max VRAM width is 2048, read from register ranges 128 bit to 2048 bit. From the code it seems some chips will have HBM widths from 128, 256, 512, 1024, 2048. Some chips apparently have no HBM?
- There are bin files/microcode which are loaded into the asic everytime when driver is loaded?
- Vega and Raven belongs to new chip family unlike Polaris which belong to a big family of chips (Tonga, Fiji, Polaris10/11)
- Vega has many new ip blocks,
- Raven has integrated audio it seems.
- Raven is having many similar ip blocks, including the memory controller for gfx9, but it lacks uvd and vce has vcn instead
- It seems that vcn has both encode and decode blocks, I am not sure but seems so.
amdgpu_ip_block_add(adev, &uvd_v7_0_ip_block);
amdgpu_ip_block_add(adev, &vce_v4_0_ip_block);

Raven has vcn instead, which according to phoronix is doing similar job.
amdgpu_ip_block_add(adev, &vcn_v1_0_ip_block);
 
Last edited:

DisEnchantment

Golden Member
Mar 3, 2017
1,687
6,243
136
This section of code intrigues me for Raven
https://cgit.freedesktop.org/~agd5f...p&id=790bd5b4935cef137ac62b73599688bd390d5a58

Code:
/* temporary hardcoded clock voltage breakdown tables */
DpmClock_t VddDcfClk[]= {
    { 300, 2600},
    { 600, 3200},
    { 600, 3600},
};

DpmClock_t VddSocClk[]= {
    { 478, 2600},
    { 722, 3200},
    { 722, 3600},
};

DpmClock_t VddFClk[]= {
    { 400, 2600},
    {1200, 3200},
    {1200, 3600},
};
...

Which one is core clock? It is used somewhere later down in the source. Could this hint something?
 

raghu78

Diamond Member
Aug 23, 2012
4,093
1,475
136
I think we will see two mobile Ryzen APUs as we saw in the embedded slides from a few days back. A 2C/4T APU with 192 Vega sp (3 compute units). This chip is targetted at 4-15w TDP and will definitely find a place in premium ultrathins and tablets. The larger 4C/8T APU with 704 Vega sp (11 compute units) is targetted at 15-45w TDP. This should sell into a range of devices from ultrathins to full sized 15.6 inch high performance notebooks. We have already seen that a 4C/8T R5 1500x easily beats a 4M/8T Piledriver based FX 9590 CPU even in multithread performance.

http://www.hardwarecanucks.com/foru...ryzen-5-1600x-1500x-performance-review-8.html

In single thread performance Zen kills the Bulldozer based architectures.

http://www.hardwarecanucks.com/foru...ryzen-5-1600x-1500x-performance-review-6.html

So its very obvious that 2C/4T Zen APUs will be faster than 2M/4T Carrizo APUs for CPU performance but at less than half the power. I am more interested in seeing how the 192 sp Vega GPU with GFX 9 arch and with significantly higher clocks compares against the full 512 sp Carrizo GPU based on GFX 8 architecture used in Tonga running at much lower clocks.My guess is we might see the GPU performance very close. So its likely that 2C/4T Zen APUs at 15w TDP will have roughly same gaming performance as the high end Carrizo notebook APUs at 35W . the 4C/8T Zen APUs will bring a new level of graphics performance which is going to get very close to discrete GPUs.
 

Bouowmx

Golden Member
Nov 13, 2016
1,142
550
146
Clearly 2×Summit Ridge (16 Zen cores) has its use, but thinking about Topweasel's proposal of 2×Raven Ridge (2×4 Zen cores + 2×704 Vega shaders), I wonder about its suitable use cases.
 

Topweasel

Diamond Member
Oct 19, 2000
5,436
1,655
136
Clearly 2×Summit Ridge (16 Zen cores) has its use, but thinking about Topweasel's proposal of 2×Raven Ridge (2×4 Zen cores + 2×704 Vega shaders), I wonder about its suitable use cases.
8 Core laptop with near Desktop level graphics onboard. It would be better power wise than most 4c discreet Video setups. Great for mobile video editing and rendering.

I am just wondering because if it's just a really low power Pinnacle Ridge than why wait till after Raven Ridge to release it?

Sent from my Pixel XL using Tapatalk
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
AMD "THREADRIPPER" based on 2×Raven Ridge: interesting, but the appeal?

How could it be Raven based, when Raven has a single CCX (4C/8T) per die and Threadripper has 16C/32T with two dies in MCM config?
 

Tuna-Fish

Golden Member
Mar 4, 2011
1,422
1,759
136
Which one is core clock? It is used somewhere later down in the source. Could this hint something?

No. Some information was once leaked by the drivers, so nowadays they use temporary values that they replace with final ones after the full specifications are unveiled.
 

otinane

Member
Oct 13, 2016
68
13
36
So according to yesterday's event, there won't be a Pinnacle Ridge RR version. Just what we know till now, and at 2019 the 7nm Gray Hawk.

Pinnacle Ridge seems to be only for Ryzen CPUs.
 

T1beriu

Member
Mar 3, 2017
165
150
81
What's the chance of mobile Ryzen Pro being like Ryzen Pro and having two Raven Ridge dues in MCM? 8 cores 22 CU's.

Ryzen Pro is not MCM. Ryzen Pro is Ryzen for the commercial OEMs with increase in validation and support with an extra price tag.

And I see no chance of RR MCM because I can't see the value scenario/complexity/manufacturing price for a laptop.
 

T1beriu

Member
Mar 3, 2017
165
150
81
So according to yesterday's event, there won't be a Pinnacle Ridge RR version.

My hunch seems to be on point - Pinnacle Ridge is a Summit Ridge rebrand/refresh, just like Godavari. I think we'll see an increase of 100MHz.

People expected it to be Zen2/Zen+ in 2018, but I always thought that AMD can't have the resources to work on this many designs at the same time in such a short time.
 

Topweasel

Diamond Member
Oct 19, 2000
5,436
1,655
136
Ryzen Pro is not MCM. Ryzen Pro is Ryzen for the commercial OEMs with increase in validation and support with an extra price tag.

And I see no chance of RR MCM because I can't see the value scenario/complexity/manufacturing price for a laptop.

I only brought up MCM because we know that Ryzen Pro will be the 12c and 16c (ignoring the other configurations guessed). I really don't see why they would need to go to the trouble of qualifying the same chip twice since you can have retail CPU in workstation. AMD's goal during all of this is to massively shake up all markets. A 4C8T iGPU APU isn't a market shake up chip specially in the pro field. Now on the other hand it could be them going with a mobile pinnacle ridge and offering a 8c16t laptop chip. Which is probably Occam's Razor here. I was just throwing out the MCM option because a 4 channel memory controller, 22 GC CU, 8c16t, 60w-80w mobile CPU would be insanely powerful and the 22 CU portion would be competitive with discrete laptop solutions. It could also be a HBM2 solution that some people have been looking for. I just doubt its as simple as solo Raven Ridge rebadged.
 

T1beriu

Member
Mar 3, 2017
165
150
81
I only brought up MCM because we know that Ryzen Pro will be the 12c and 16c

Again. Ryzen Pro is Ryzen for OEMs. Ryzen Pro is not ThreadRipper and ThreadRipper is not Ryzen Pro. I don't know where got this from.

 
Last edited:

Topweasel

Diamond Member
Oct 19, 2000
5,436
1,655
136
Again. Ryzen Pro is Ryzen for OEMs. Ryzen Pro is not ThreadRipper and ThreadRipper is not Ryzen Pro. I don't know where got this from.
Well guess this is a we'll see moment. I don't know why people would assume that with the amount of smoke around a HDET platform with Threadripper, that it doesn't fall right into the wheel house of a Workstation processor. Seems silly to assume Ryzen Pro is just Pinnacle Ridge qualified for registered ECC memory.

Though I did look up to see if there was any specifics in the meeting yesterday and I did see a slide that referred to a Ryzen 5 Pro 1500 beating a 7500 with mediocre numbers. So you could be right. But then it just comes off as the most pointless branding ever. Workstations could really use the extra PCIe lanes of Threadripper and there really isn't a need to rebrand, they could just offer a workstation chip-set and call it good.

So you could be right. But it would be insanely worthless and boring if that is the case.
 

CatMerc

Golden Member
Jul 16, 2016
1,114
1,153
136
My hunch seems to be on point - Pinnacle Ridge is a Summit Ridge rebrand/refresh, just like Godavari. I think we'll see an increase of 100MHz.

People expected it to be Zen2/Zen+ in 2018, but I always thought that AMD can't have the resources to work on this many designs at the same time in such a short time.
100MHz sounds rather pessimistic, assuming they use an improved node like Samsung's 14nm LPU.
 

Abwx

Lifer
Apr 2, 2011
11,167
3,862
136
GF wouldnt bother tweaking their process just for 2.5% higher max frequency, they ll likely target 10% higher Fmax as well as reshaping the voltage/frequency characteristic between 3.3 and 4GHz, where it s no more in the same rate as below 3.3GHz, and hence to increase substancially the perf/watt in this range.

 

Topweasel

Diamond Member
Oct 19, 2000
5,436
1,655
136
Even the mobile version makes sense. Intel's had some success selling Xeon branded mobile processors.
A little bit limited and it's not like they are doing a full lineup, it's high watt, either iris versions of the i7 or clocked well above the i7. On the desktop all Xeon's aimed at workstations are using the same platform as the HEDT market.

That's what is bothering me about the R5 Pro shot. If Ryzen Pro is just a rebrand of AM4 desktop parts with slightly higher clocks all it will do is further confuse the market and be utterly pointless because OEM's don't need Workstation specific CPU's outside the qualifications AMD already has to do on the CPU's for selling them it's the OEM's that have to do the brunt of testing. So do they really need 4/6/8 core CPU's branded specifically for them, if in the same situation they could still sell a normal workstation with an i7? It seems like a waste of Branding and probably a non-starter if you ask me.

So now how are they going to separate the X399 stuff? Create a second consumer and second Pro workstation platform as a Ryzen 9 and Ryzen 9 Pro? That's not the stability what businesses want on their "Professional" systems.
 

jpiniero

Lifer
Oct 1, 2010
14,841
5,456
136
A little bit limited and it's not like they are doing a full lineup, it's high watt, either iris versions of the i7 or clocked well above the i7. On the desktop all Xeon's aimed at workstations are using the same platform as the HEDT market.

But it has ECC enabled and validated. That's the whole point.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |