AMD Ryzen Gen 2 Set For Q2 2018

[raghu78]

The Zen+ samples are not final performance according to James Prior of AMD. RR samples were clocked lower from final production chips.Listen to video at 8:20 - 8;53

https://www.youtube.com/watch?v=7MPxN6MEgbk

https://www.reddit.com/r/Amd/comments/7ooy69/ryzen_desktop_apu/?st=jea0slhl&sh=7964ba32

3.0 Ghz/ 3.5Ghz became 3.5 Ghz/3.7 Ghz on R3 2200G
3.3Ghz/3.5Ghz became 3.6 Ghz/3.9Ghz on R5 2400G

I stand by my earlier predictions that Pinnacle Ridge will be atleast >=10% higher for base clocks and >=15% for max turbo wrt Summit Ridge.

 

[The Stilt]

The Zen+ samples are not final performance according to James Prior of AMD. RR samples were clocked lower from final production chips.

https://www.reddit.com/r/Amd/comments/7ooy69/ryzen_desktop_apu/?st=jea0slhl&sh=7964ba32

3.0 Ghz/ 3.5Ghz became 3.5 Ghz/3.7 Ghz on R3 2200G
3.3Ghz/3.5Ghz became 3.6 Ghz/3.9Ghz on R5 2400G

I stand by my earlier predictions that Pinnacle Ridge will be atleast >=10% higher for base clocks and >=15% for max turbo wrt Summit Ridge.

ES / Prototype and PR / QS are two different things.

Weren't those Ravens GE PR parts (instead of G) anyway?

EDIT: They weren't, 65W TDP.

 

[anj2208]

How is that your conclusion?
I'd love a 4.0/4.5 offering, but the 2800X seems aimed at 3.9/4.3. OC headroom is a complete unknown at this point.

not a conclusion, it's a possibility. esp with new management, from what we can tell so for they are very conservative with target. Original SR ES was what? 3.2 and launching at 3.6 base.

and there's 10% process improvement on 12LP. I cant remember the source but improving clock is one of the highest priority in PR.

So yes, i am optimistic

 

[Peter Watts]

So what's going to be the lowest end one?

 

[exquisitechar]

The Zen+ samples are not final performance according to James Prior of AMD. RR samples were clocked lower from final production chips.

https://www.reddit.com/r/Amd/comments/7ooy69/ryzen_desktop_apu/?st=jea0slhl&sh=7964ba32

3.0 Ghz/ 3.5Ghz became 3.5 Ghz/3.7 Ghz on R3 2200G
3.3Ghz/3.5Ghz became 3.6 Ghz/3.9Ghz on R5 2400G

I stand by my earlier predictions that Pinnacle Ridge will be atleast >=10% higher for base clocks and >=15% for max turbo wrt Summit Ridge.

Interesting that it differs from the products that hit the shelves, given that the "ZD" in the codename means that that is a QS, as far as I know. It does indeed mean that there is a chance that these PR leaks we're seeing are not indicative of the final clocks.

 

[FIVR]

If the 2700X is 3.7/4.2 then what is the 2800X?

I'm hoping for 4.0/4.4

 

[Dayman1225]

If the 2700X is 3.7/4.2 then what is the 2800X?

I'm hoping for 4.0/4.4

I'm going to guess 3.9/4.3 possibly upto 4.4 with XFR

 

[moinmoin]

So what's going to be the lowest end one?

2200G

 

[Peter Watts]

2200G

2200g and 2400g are Zen+??? I was under the impression that they are not.

 

[HurleyBird]

2200g and 2400g are Zen+??? I was under the impression that they are not.

There's some evidence that Pinnacle and Raven Ridge are closely related (an interview with an AMD rep used language that suggested they are basically the same in terms of architectural advancements over Summit Ridge). I suppose we'll know for sure when they launch, but keep in mind that Raven Ridge has a neutered L3 so can't really be used for performance projections.

Raven Ridge also only has one CCX, and there's evidence that one of Pinnacle Ridge's improvements is better communication between CCXes given the significantly improved multithreaded scores for the 2600 sample.

 

[CatMerc]

The Zen+ samples are not final performance according to James Prior of AMD. RR samples were clocked lower from final production chips.Listen to video at 8:20 - 8;53

https://www.youtube.com/watch?v=7MPxN6MEgbk

https://www.reddit.com/r/Amd/comments/7ooy69/ryzen_desktop_apu/?st=jea0slhl&sh=7964ba32

3.0 Ghz/ 3.5Ghz became 3.5 Ghz/3.7 Ghz on R3 2200G
3.3Ghz/3.5Ghz became 3.6 Ghz/3.9Ghz on R5 2400G

I stand by my earlier predictions that Pinnacle Ridge will be atleast >=10% higher for base clocks and >=15% for max turbo wrt Summit Ridge.

The TDP of these isn't anything from known engineering sample naming. So it's not 95W, not 65W, and not 15W. Probably 35W or 45W, and not the 2200G/2400G we know.

There's some evidence that Pinnacle and Raven Ridge are closely related (an interview with an AMD rep used language that suggested they are basically the same in terms of architectural advancements over Summit Ridge). I suppose we'll know for sure when they launch, but keep in mind that Raven Ridge has a neutered L3 so can't really be used for performance projections.

Raven Ridge also only has one CCX, and there's evidence that one of Pinnacle Ridge's improvements is better communication between CCXes given the significantly improved multithreaded scores for the 2600 sample.

The memory controller on Pinnacle Ridge is completely different, and I believe the infinity fabric got updated too with tightened latencies. Branch prediction is also supposedly tweaked on Zen+.

 

[moinmoin]

2200g and 2400g are Zen+??? I was under the impression that they are not.

2200G is and very likely stays the lowest end Ryzen Gen 2 chip which is what you (implicitly since posting in this thread) asked about. It's not the lowest end 12LP/PR chip, that may be 2300(X) once one such gets a release.

 

[DisEnchantment]

https://www.3dmark.com/spy/3366892
3.7/4.125

https://www.3dmark.com/fs/15008158
3.7/4.2

https://videocardz.com/75159/amd-ryzen-7-2700x-spotted-with-clocks-up-to-4200-mhz

R7 2700X @3.7/4.1

So, R7 2800X @4.0/4.4?

The annoying thing with my 1700X are the damn segmentation faults when I do 16 threads compilation and the fan speed due to the Temperature offset which when operating in Linux causes a lot of Fan noise. I hear Kernel 4.15 fixes this. I cannot RMA my CPU as I am working almost 24/7 with my PC. Other than that I am happy with 1700X, first time I could have 2 really responsive VMs on two screens which makes developing for multiple target platforms so hassle free.
If Ryzen 2X00 fixes these things I will get one of these.

 

[EXCellR8]

opps, repost...

 

[meson2000]

I'm going to guess 3.9/4.3 possibly upto 4.4 with XFR

I am going to agree with the above. While 4ghz base would be nice, i think it will be 3.9/4.3 with 4.4 XFR. Then i have to decide if i want to upgrade my 1800X to a 2800X

 

[Asterox]

R7 2700X @3.7/4.1

So, R7 2800X @4.0/4.4?

The annoying thing with my 1700X are the damn segmentation faults when I do 16 threads compilation and the fan speed due to the Temperature offset which when operating in Linux causes a lot of Fan noise. I hear Kernel 4.15 fixes this. I cannot RMA my CPU as I am working almost 24/7 with my PC. Other than that I am happy with 1700X, first time I could have 2 really responsive VMs on two screens which makes developing for multiple target platforms so hassle free.
If Ryzen 2X00 fixes these things I will get one of these.

In this example, hm very likely R7 2700X for him Sandra says that 4.35ghz is max Turbocore.

http://www.hwbattle.com/bbs/board.php?bo_table=hottopic&wr_id=8803

 

[Gideon]

It doesn't say anywhere it's 2700X. But what makes me really suspicious is the AIDA mem-latency taken straight from Raven Ridge and 4.35 @ 1.2V

I would really like this is true, but am quite sceptical at least until we get a confirming leak.

 

[DisEnchantment]

In this example, hm very likely R7 2700X for him Sandra says that 4.35ghz is max Turbocore.

http://www.hwbattle.com/bbs/board.php?bo_table=hottopic&wr_id=8803

I would like this to be true, but I am more likely to believe it is an 2800X but more than happy to be wrong.

Out of curiosity, I upload this image to http://fotoforensics.com and guess what that CPU type seems to be Pinnacle Ridge. Give it a try.

I upload my result here

If it is really Pinnacle ridge, I am curious what would the memory latencies be at DDR 3200 MHZ / CL14, I speculate high 50s to low 60s .

Now hwbattle link is dead.

 

[eek2121]

2200g and 2400g are Zen+??? I was under the impression that they are not.

They aren't. The only thing they share in common is Precision Boost 2.0. The APUs are manufactured on 14LPP and the new PR chips are on 12LP.

 

[eek2121]

Threadripper, 8 cores, 4.3 GHz. Dual channel DDR 4 3200CL16 memory. The only thing I forgot to do is lower the memory speed. I will do that next.

EDIT: The new scores at DDR 2400CL14 were:

SC: 176
MC: 1856

This suggests to me the whole thing was faked.

 

[HurleyBird]

Why? I mean, there's a good chance anything like this is fake, but you matched the ST score while overshooting the MT score. What does that prove? And did they actually post a Cinebench screenshot or just put out the numbers in a bar graph? If the later, why would they even use real results from a TR system? In general, that's just a lot of work that a faker wouldn't need to do to generate a fake.

 

[xblax]

EDIT: The new scores at DDR 2400CL14 were:

SC: 176
MC: 1856

This suggests to me the whole thing was faked.

It wouldn't run 4.3Ghz allcore when 4.35Ghz is the max turbo. You're singlecore score however seems to be perfectly in line with the leak.

Don't expect any improvements in cinebench clock for clock. It's perfectly parallelized and doesn't profit from tightened cache and infinity fabric latencies. Ryzen 2000 will shine in workloads that access memory more randomly and require inter-thread synchronization (like gaming).

 

[HurleyBird]

Found another screenshot at SA forums. Big Dhrystone boost apparently, edging 8700K.

 

[CatMerc]

It's the 2700X because the base clock of 3.7GHz matches previous leaks. This is not the 2800X.

 

[eek2121]

It wouldn't run 4.3Ghz allcore when 4.35Ghz is the max turbo. You're singlecore score however seems to be perfectly in line with the leak.

Don't expect any improvements in cinebench clock for clock. It's perfectly parallelized and doesn't profit from tightened cache and infinity fabric latencies. Ryzen 2000 will shine in workloads that access memory more randomly and require inter-thread synchronization (like gaming).

Do you have a Ryzen chip? If so, set it to stock, open HWInfo, and note the clockspeed behavior while CB in single core and multicore. You'll then begin to understand why this is fake. PB2 will provide some improvement in this area, but my money is on:

1700X - 3.7/4.2
1800X - 4.0/4.4

AMD has stated publically that Ryzen will have memory latency and cache improvements. That's all they've said publicly about the chip thus far. Threadripper already has these improvments baked in, so similar clock speeds should obtain similar results. Now granted, thanks to the lack of control over precision boost, I can't reproduce the exact scenario, but after looking at the original page, it seemed pretty fake, or at the very least, it was an ES chip.

There is also a question of whether the new chips will have a higher single core turbo or not.

EDIT: Also don't forget that AMD will be saving the best binned dies for Threadripper.

EDIT #2: LPP processes tend to be different from LP processes. I suspect AMD is going to surprise us with the Threadripper 2900X. Don't be surprised if they push clock rates well beyond the 1800X thanks to the 180 watt TDP, depending on the voltage/clockspeed curve...which would potentially give enthusiasts a very good reason to go TR4. Just speculation of course.