New Zen microarchitecture details

[dacostafilipe]

There is an official OPN decoder available. Saw it last week but can't find it right now.

Please PM me if you find it, thx!

 

[Dresdenboy]

Thx.

I'm still missing some information to decode final SKUs (TDP, Stepping, ...) : http://dzencoder.herokuapp.com/?code=YD180XBCM88AE

Some quick notes:
I think, "Y" is new (for "RYzen"). The 180 is for 1800. X is for "XFR" -> 1800X.
"AE" could still map to the stepping (likely B1 here).


For ES you can use this table: http://www.cpu-world.com/info/id/AMD-ES-identification.html and the pic made by a Redditor. Note the difference in first 3 digits, which denote base clock frequency/MACB, not just the first 2 digits.

 

[piesquared]

Some quick notes:
I think, "Y" is new (for "RYzen"). The 180 is for 1800. X is for "XFR" -> 1800X.
"AE" could still map to the stepping (likely B1 here).


For ES you can use this table: http://www.cpu-world.com/info/id/AMD-ES-identification.html and the pic made by a Redditor. Note the difference in first 3 digits, which denote base clock frequency/MACB, not just the first 2 digits.

[edit]
Nope forget all that my memory completely failed me. I found the decoder on Videocardz comments.

Although the one i saw did define YD as retail. (and probably isn't official either)

 

[Mopetar]

16 CU APU doesn't make sense unless you have HBM or it will be so memory starved it's sad. 16 CU also eats up more die space which makes the chip bigger, which adds to the cost and lowers yields. Never mind that 16 CU stays into 460 territory and and makes that GPU largely pointless so unless AMD is going to drop that part of the market completely, or HBM2 is ready for mass production and they can put it on an APU, I don't see the point in a 16 CU APU unless they use Vega and it has the worlds most magical memory compression technology.

 

[Glo.]

16 CU APU doesn't make sense unless you have HBM or it will be so memory starved it's sad. 16 CU also eats up more die space which makes the chip bigger, which adds to the cost and lowers yields. Never mind that 16 CU stays into 460 territory and and makes that GPU largely pointless so unless AMD is going to drop that part of the market completely, or HBM2 is ready for mass production and they can put it on an APU, I don't see the point in a 16 CU APU unless they use Vega and it has the worlds most magical memory compression technology.

Raven Ridge NPU's are combining Ryzen CPUs with Vega GPUs. Vega is designed to work with HBM2. 16CU chip with 4 32 bit memory controllers is 123 mm2 Ryzen CCX has around 46mm2. There is a lot of room for the rest of the chip, if whole die is supposed to be around 210 mm2.

P.S. 16CU Vega chip will be faster than RX 460 .

 

[The Stilt]

I'm looking for some known Zen OPN codes to test my decode script.

For now I have those here:

ZD3601BAM88F4_40/36_Y
ZD3201BBM4KF4_34/32_Y
ZD3301BBM6IF4_37/33_Y
ZD3406BAM88F4_38/34_Y

If you have others, please post them here or PM me.
Thanks!

btw. The fourth number in the ES SKUs has nothing to do with the die configuration (as I speculated) or the revision (like PCGH claimed)

 

[iBoMbY]

Maybe 6 means XFR, and 1 means normal Turbo?

 

[dacostafilipe]

Maybe 6 means XFR, and 1 means normal Turbo?

Yeah possible.

That position seems to indicate XFR on the final SKUs.

 

[CatMerc]

Maybe 6 means XFR, and 1 means normal Turbo?

The 3.6/4.0 SKU doesn't have the 6, so probably not.

btw. The fourth number in the ES SKUs has nothing to do with the die configuration (as I speculated) or the revision (like PCGH claimed)

Gonna take a wild guess and say that the 6 indicates a PRO SKU.
What that would mean exactly I've no clue.

 

[Glo.]

Which also would suggest that my theory, that leaked 6 core CPU scores is NOT for 95W+ TDP CPU, and there will be one higher end CPU.

 

[Glo.]

The 3.6/4.0 SKU doesn't have the 6, so probably not.


Gonna take a wild guess and say that the 6 indicates a PRO SKU.

BB - 65W, BA - 95W.
The 4C/8T model with XFR is Ryzen 5 1300. Which one do you think, the 3.6/4.0 GHz 8C/16T, with BA TDP target would end up? Yup. 1800 Pro. So it is entirely possible.

Also, bare in mind that WTFTech incorrectly suggests that 6C/12T is 95W TDP. Its not, as we can see from the indication of TDP target - 65W. So it is either Ryzen 5 1500 or 1500 Pro.

 

[Dresdenboy]

[pic removed]
BB - 65W, BA - 95W.
The 4C/8T model with XFR is Ryzen 5 1300. Which one do you think, the 3.6/4.0 GHz 8C/16T, with BA TDP target would end up? Yup. 1800 Pro. So it is entirely possible.

Also, bare in mind that WTFTech incorrectly suggests that 6C/12T is 95W TDP. Its not, as we can see from the indication of TDP target - 65W. So it is either Ryzen 5 1500 or 1500 Pro.

My try:
Y = Ryzen (it's "F" for FX for example)
D = Desktop
180 = 1800, etc. (8350 with 4th digit for FX-8350)
0/B/X = std/Pro/XFR
BA/BB/BC = 95W HS65 / 65W HS55 / 95W + SR3+ HS81
Those differ depending on tray/memory package (MPK), this is for tray:
M = AM4 package (as with ES)?
4/6/8 = # cores
8/I/G/K = caches? threads? -> different for 4C/4T and 4C/8T
AE = stepping?

Tray vs MPK:
YD1700BBM88AE YD1700BBAEMPK
YD170XBCM88AE YD170XBCAEMPK
YD180XBCM88AE YD180XBCAEMPK

Edit: included The Stilt's remark and some info from this posting.

 

[The Stilt]

B & A/B/C belong together.

 

[bjt2]

http://imgur.com/a/yM9l4

Chinese leak.
R5 1400X (4c/8t 65W) rated 4GHz by windows. Included a screen with lightly loaded CPU and mean clock of 4.52GHz (XFR in action i presume)

 

[Glo.]

http://imgur.com/a/yM9l4

Chinese leak.
R5 1400X (4c/8t 65W) rated 4GHz by windows. Included a screen with lightly loaded CPU and mean clock of 4.52GHz (XFR in action i presume)

So the leaks from WTFTech appear to be completely incorrect, in terms of TDP ratings and core clocks.

This CPU is 199$. Holy sh**.

 

[Phynaz]

So the leaks from WTFTech appear to be completely incorrect, in terms of TDP ratings and core clocks.

This CPU is 199$. Holy sh**.

I'm lost, where did you get $199 from?

 

[lolfail9001]

So the leaks from WTFTech appear to be completely incorrect, in terms of TDP ratings and core clocks.

Mind you: it could be XFR acting up. Also, is not Windows task manager notorious for reporting wrong frequency?

 

[CatMerc]

I'd say this is fake. It has the actual name rather than an OPN like the rest of the leaks.
It would imply review samples are out, which is not the case according to multiple sources.

Edit:
And it says 1MB of L2. A quad core Ryzen should have 2MB.
This is a renamed Intel CPU, probably a 6700K judging by the base clock reported and the cache.

 

[Atari2600]

http://imgur.com/a/yM9l4

Chinese leak.
R5 1400X (4c/8t 65W) rated 4GHz by windows. Included a screen with lightly loaded CPU and mean clock of 4.52GHz (XFR in action i presume)

Ryzen TM in lowercase?

Nah. Fake.

 

[Glo.]

I'd say this is fake. It has the actual name rather than an OPN like the rest of the leaks.
It would imply review samples are out, which is not the case according to multiple sources.

Edit:
And it says 1MB of L2. A quad core Ryzen should have 2MB.
This is a renamed Intel CPU, probably a 6700K judging by the base clock reported and the cache.

Skylake has 128 KB L1 cache.

 

[CatMerc]

Skylake has 128 KB L1 cache.

 

[Glo.]

Yup, its Skylake Faked as Ryzen 5.

 

[bjt2]

I found the chinese thread: http://tieba.baidu.com/p/4985247641
With gtranslate it seems a shopped 6700K...
My bad.

 

[CatMerc]

Anyone else finds odd that in the CPU-Z leak, the benchmark is in an English OS, while the CPU-Z pic is in Chinese?

Taken with a different camera too, judging by the pic size.

 

[lolfail9001]

Anyone else finds odd that in the CPU-Z leak, the benchmark is in an English OS, while the CPU-Z pic is in Chinese?

Wait a minute... Indeed.

Well, we have already had 1 fake from Chinese, time to increase salt levels.