AMD Bristol/Stoney Ridge Thread

[KTE]

This is a very poor i7-6500U score, it scores over 300 usually.

http://www.notebookcheck.net/Intel-Core-i7-6500U-Notebook-Processor.149431.0.html

My old i7-3667U scores 113/268 random non-op run.

//cloud.tapatalk.com/s/57d85f7ee9202/x1c-stock.cb15.jpg?

So 258 is a very low score, indeed.

Edit: and 2922 with CPUZ MT (2311 for BR).

Sent from HTC 10
(Opinions are own)

 

[DrMrLordX]

Could someone explain me the math behind this (no endnotes available)?

No. Not without knowing more about Bristol Ridge's memory controller. If it's the same IMC as in Carrizo then . . . it makes no sense.

 

[The Stilt]

No. Not without knowing more about Bristol Ridge's memory controller. If it's the same IMC as in Carrizo then . . . it makes no sense.

There is no difference.
The only part of the software stack (AGESA, µcode and co-processor firmwares) which isn't shared between Carrizo & Bristol Ridge is the SMU configuration block. The µcode, VBIOS, AGESA, DSP firmwares and most of the SMU firmware(s) are identical between the two. It is pretty funny, since even a minor die steppings (e.g CZ-A0 & CZ-A1) always require their own µcode but Trinity & Richland, Kaveri & Godavari, Carrizo & Bristol Ridge share the same codes, despite they are supposed to be a different design. Both Carrizo and Bristol Ridge use firmware version 9221 for the PMU (memory controller). There are differences in the SMU configuration settings, however SMU cannot directly control any clocks. It can toggle the states of all domains (DPMs), but it cannot freely adjust any clocks outside them.

 

[The Stilt]

Found the end-notes for the claim

Memory Bandwidth: Testing by AMD Performance labs. PC manufacturers may vary configurations yielding different results. AMD Internal lab testing using an AMD A12-9800 processor scored 13216.96MB/s in the WinSAT Memory Bandwidth benchmark vs. an AMD A10-8850B processor which scored 10841.72MB/s in the benchmark, for a benchmark score difference of 13,216/10,841 = 1.22X or 22% more.

It is sad to see that most of the claims AMD nowdays make, make no sense what so ever unless you manage to fish out the end-notes / footnotes somewhere

 

[Zstream]

Found the end-notes for the claim



It is sad to see that most of the claims AMD nowdays make, make no sense what so ever unless you manage to fish out the end-notes / footnotes somewhere

Because people can't read footnotes. Give me a break...

 

[The Stilt]

Because people can't read footnotes. Give me a break...

You mean those footnotes, which are placed to the very last pages of the presentations and which generally for that reason are unavailable to the end-users?
It is hard to read something you don't have access to. Take a look, how many sites who reported about the 7th gen. APUs (including the 22% claim) actually included the footnotes from the end of the slide, from the last pages which contained nothing but footnotes?
The only reason why I found the explanation is that I managed to find the original PDF.

It wouldn't be too hard for AMD to add either *difference of 13,216/10,841 (MB/s) = 1.22X or 22% more. or even *measured using WinSAT notes to the very same page they make these claims in the first place.

They could claim as well that they have 347 billion dollars of assets² and then provide footnotes 20 pages later in small print stating: Zimbabwean dollars, 1 Z$ = 0.00276 US$ as of 09/14/2016.

 

[NostaSeronx]

No. Not without knowing more about Bristol Ridge's memory controller. If it's the same IMC as in Carrizo then . . . it makes no sense.

There is no difference.

It should be using the same IMC in Stoney Ridge.

Carrizo = 2 DDR4 Channels => 2 DDR4 DIMMs
Stoney Ridge = 1 DDR4 Channels => 2 DDR4 DIMMs [DIMM_A0/DIMM_B0]
Bristol Ridge = 2 DDR4 Channels => 4 DDR4 DIMMs [DIMM_A0/1/DIMMB0/1] (Technically, it could also work in quad-channel systems A0/B0/C0/D0, much like stoney can work in dual-channel systems)

The above is just one of the changes.. Another is that it should support 3200 Mbps/3.2 Gbps; http://www.design-reuse.com/sip/3-2gbps-tsmc-28nm-ddr4-3-lpddr3-2-combo-ddr-phy-ip-39431/
(28HPC/HPC+ => GF28A/GF28HPA)

From Cadence Denali DDR4/DDR3 2400/2133 Mbps support to Rambus R+ DDR4/DDR3 3200/2133 Mbps.

 

[The Stilt]

Stoney has completely different DRAM PHY IP.
It is 1x 64-bit DDR4, while Carrizo and Bristol Ridge are using 2x 64-bit DDR3/DDR4 IP.

 

[NostaSeronx]

Stoney has completely different DRAM PHY IP.
It is 1x 64-bit DDR4, while Carrizo and Bristol Ridge are using 2x 64-bit DDR3/DDR4 IP.

Both Stoney & Bristol are using the same DRAM PHY IP. Rambus's R+ DDR4/DDR3 3200 Mbps/2133 Mbps PHY. Both are using Polaris GFX_IP on both dies, which means both support HDMI 2.0b w/ HDR Rec2100, Adaptive Sync HDMI. Both have the Quad-MBFF & the AVFS system from Polaris. Hence, the higher clock rate at lower TDPs in comparison to Carrizo/Carrizo-L.

@KTE for

Edit: and 2922 with CPUZ MT (2311 for BR).

Single-threaded @ 2.9 GHz and Multi-threaded @ 2.9 GHz. The score is limited by Northbridge & DRAM speed.
http://valid.x86.fr/yajvy9

BIOS v1.09 is out for Stoney Ridge, E5-523G. Waiting for it, on E5-553G. Comes with more 65h-7Fh specific BIOS changes.

http://www.jagatreview.com/2016/06/hands-on-preview-amd-fx-9800p-bristol-ridge-apu-7th-gen/
BIOS v1.05 => 64-bit Northbridge/DRAM
BIOS v1.08 => 128-bit Northbridge/DRAM
BIOS v1.09/v1.xx => 256-bit Northbridge/128-bit DRAM, so ideally that Multithreaded score should hit much closer to Kaveri Reference.

 

[lolfail9001]

Found the end-notes for the claim



It is sad to see that most of the claims AMD nowdays make, make no sense what so ever unless you manage to fish out the end-notes / footnotes somewhere

Wait, they did not even specify what actual, uh, memory sticks were used?

 

[The Stilt]

Wait, they did not even specify what actual, uh, memory sticks were used?

2x 4GB 2133 for Godavari, 2x 4GB 2400 for Bristol Ridge.

 

[Zstream]

You mean those footnotes, which are placed to the very last pages of the presentations and which generally for that reason are unavailable to the end-users?
It is hard to read something you don't have access to. Take a look, how many sites who reported about the 7th gen. APUs (including the 22% claim) actually included the footnotes from the end of the slide, from the last pages which contained nothing but footnotes?
The only reason why I found the explanation is that I managed to find the original PDF.

It wouldn't be too hard for AMD to add either *difference of 13,216/10,841 (MB/s) = 1.22X or 22% more. or even *measured using WinSAT notes to the very same page they make these claims in the first place.

They could claim as well that they have 347 billion dollars of assets² and then provide footnotes 20 pages later in small print stating: Zimbabwean dollars, 1 Z$ = 0.00276 US$ as of 09/14/2016.

Ok, whatever mate. Too lazy to look at footnotes and also as you stated, the "original" documents show it in very clear footnotes.

 

[PhonakV30]

Found the end-notes for the claim

It is sad to see that most of the claims AMD nowdays make, make no sense what so ever unless you manage to fish out the end-notes / footnotes somewhere

I don't get it.you say they should place on Memory bandwidth page because of your post (342)?

on the slide you can see "See endnotes for detail"

Edit : Winsat Bench Is Valid for Memory bandwidth ?

 

[The Stilt]

The reason why I am complaining is because all of the necessary information, required to evaluate the claim properly is "hidden" away in the back pages (footnotes). AMD generally doesn't provide this kind of marketing material to consumers, so when the media show these slides they usually always are lacking the "footnotes". For clarity they should include a short explanation (other than see "footnote", which is not available to the end-user), or alternatively somehow force the media the make them available in their news article. The media excluding the footnotes is not AMDs fault, their fault is the excess use of footnotes in the first place.

WinSAT is not something commonly used to evaluate the memory bandwidth and I personally certainly wouldn't use it for the purpose. Most likely the reported results are close enough correct thou. It seems the be the write bandwidth they decided to report.

 

[Justinbaileyman]

So is Bristol Ridge coming before ZEN? If so is it going to be only for laptops or desktop as well?
Also will it be BGA or Standard Socket Type? How soon are we to seeing these released?

 

[KTE]

The same problem existed with Bobcat/Zacate. It's an AMD Partner/OEM problem.

The boards were piss poor, and the single channel memory configs atrocious. This could hamstring the platform heavily on its own.


Sent from HTC 10
(Opinions are own)

 

[DrMrLordX]

So is Bristol Ridge coming before ZEN? If so is it going to be only for laptops or desktop as well?
Also will it be BGA or Standard Socket Type? How soon are we to seeing these released?

Bristol Ridge is already out for OEMs on FP4. It is coming next month to the DIY market for socket AM4. The top-end AM4 chipset for the desktop set is B350 which also comes out in October. In Feb 2017, Summit Ridge (aka Zen) comes out with AM4 chipset X370 which is the "enthusiast" AM4 chipset. Unknown as to whether or not B350 will handle Summit Ridge properly.

FP4 = BGA
AM4 = PGA

 

[Justinbaileyman]

Awesome News thank you so much!! I am going to snag one of these next month with a B350 AM4 motherboard for an upgrade to my HTPC. Man I am excited!!

 

[DrMrLordX]

Be advised that it appears that all the desktop AM4 chips have locked multipliers. The base clockspeeds on the A12-9800 look pretty good regardless (3.8 GHz base/4.2 GHz turbo), but overclocking will probably require bclk adjustment. That could get a little funky.

 

[The Stilt]

Both Stoney & Bristol are using the same DRAM PHY IP. Rambus's R+ DDR4/DDR3 3200 Mbps/2133 Mbps PHY. Both are using Polaris GFX_IP on both dies, which means both support HDMI 2.0b w/ HDR Rec2100, Adaptive Sync HDMI.

Carrizo, Bristol Ridge and Stoney Ridge are all using the same GFX 8.1 IP and they all support HDMI 2.0. Stoney Ridge has newer UVD block than the other two.

So you are saying AMD has included an additional 64-bit controller, without any intention of using it, to the brand new die on which they took huge efforts in reducing the die size to the minimum?
The penalty from single channel even on the tiny GPU of Stoney is huge, so it doesn't make any sense even in theory. And judging from the die size, there definitely is no second controller either.

Additionally, the DRAM IPs used in all Excavator based APUs are not from Rambus. These controllers support up DDR3/DDR4 up to 2400MHz or DDR4 up to 2400MHz, depending on the controller.

 

[Shehriazad]

I really hope I don't have to wait long for a BR Athlon 960(K).

Even without the K it would be at 3.8 Ghz/4.2Ghz + new Chipset + DDR4 ram.

I'm pretty sure it would be a "low cost" upgrade to a DDR4 platform that actually beats the stock 860K and 870K. It may not be a perfect upgrade since Intel is a thing...but this absolutely needs to happen.

I actually feel like it would be a mistake to not instantly have the CPU only equivalent of the highest-clocking APU out at the same time as the APU

 

[Justinbaileyman]

Be advised that it appears that all the desktop AM4 chips have locked multipliers. The base clockspeeds on the A12-9800 look pretty good regardless (3.8 GHz base/4.2 GHz turbo), but overclocking will probably require bclk adjustment. That could get a little funky.

Thats totally bogus!! whats the point of even releasing something like that if it barely keeps up with intels offerings and wont even be unlocked?? I am wanting like a 860K or 880K version that you can overclock the snot out of. Come on AMD if its not one thing its another.

 

[Shehriazad]

Thats totally bogus!! whats the point of even releasing something like that if it barely keeps up with intels offerings and wont even be unlocked?? I am wanting like a 860K or 880K version that you can overclock the snot out of. Come on AMD if its not one thing its another.

I feel like that these comments are based on the fact that the OEM versions are locked. But OEM =/= Self-builder Hardware. Unless the CPUs are already being pushed to their absolute maximum I dont see any reason to not have it be unlocked.
Zen is gonna beat up BR no matter what...so its not like they'd eat up their own shares by releasing K....the opposite is the case imho. Some cheap CPUs that let you play around until you get the real thing? Seems more attractive.

 

[The Stilt]

I guess the reason for the lack of "K" models is simply because at 4.2GHz (9800 and some other SKUs) there is no headroom left. The 9800 AM4 SKUs have default turbo voltage of up to 1.525V, which is absolutely crazy for a 28nm part, which essentially is a mobile design. It's higher than the maximum default voltage for FX-9590 for example. On AM4 Bristol Ridge is merely a seat warmer for Zeppelin & Raven, so I personally find Bristol Ridge itself let alone overclocking it completely irrelevant. If you are going for AM4 anyway, why wouldn't you wait couple more months and purchase Zeppelin when it is available?

 

[Abwx]

The 9800 AM4 SKUs have default turbo voltage of up to 1.525V, which is absolutely crazy for a 28nm part, which essentially is a mobile design. It's higher than the maximum default voltage for FX-9590 for example.

Voltage being higher than the FX mean nothing, comparatively it s like talking of a given volume of matter without specifying the density..