AMD Bristol/Stoney Ridge Thread

Page 39 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
May 5, 2008
5
0
61
Hi!

What is the safe voltage for overclocking A8 9600 Bristol ridge APU 24/7 ?

On Air and the safe temp? tjmax 100 or 110?

Thanks!
 

NostaSeronx

Platinum Member
Sep 18, 2011
2,280
86
126
Hi!

What is the safe voltage for overclocking A8 9600 Bristol ridge APU 24/7 ?

On Air and the safe temp? tjmax 100 or 110?

Thanks!
4.1-4.4 GHz @ 1.4V-1.55V is the safest overclock for any Bristol Ridge.
You can do air for both as Bristol Ridge will hit the frequency-voltage wall before the power wall is hit. It is however highly dependent on the quality of the chip one gets.

Safe temperature is around 90 degrees, but the process used in Bristol Ridge "28HPA" is an automotive process. So, the actual Tjmax is above that. Do not exceed 90 degrees if you don't want the AVFS PVT(process-variation-temp) safeties to activate. Which was carried over from 28A Carrizo Pro, without upgraded process info. 28HPA should be 105-110 operating capable compared to 28A which is 85-90 operating capable.
 
Last edited:

ET

Senior member
Oct 12, 1999
508
21
91
If there's any truth to the specs quoted, it look like Bristol Ridge on FM2+. But even that is pretty cool to have.
 

ao_ika_red

Golden Member
Aug 11, 2016
1,234
65
106

amd6502

Senior member
Apr 21, 2017
422
48
61
Well, the first fm2+ BR 45W A8-7680 APUs are now showing up; available for export at ebay, shipped from Hong Kong and marked up considerably to $80 delivered.

While there I also saw a cheap $160 eMMC 6 Watt Dell A6 Stoney ( http://www.cpu-world.com/Compare_CPUs/AMD_A6-9220e,AMD_A9-9425/ ). Seems like a travel tablet substitute; 11.6 in screen seems nice and compact. However, only 4gb, and quite possibly neither ram upgrade nor storage sata based upgrade may exist. (It depends, some dell emmc laptops are expandable, some not).
 

NTMBK

Diamond Member
Nov 14, 2011
8,214
173
126
Well, the first fm2+ BR 45W A8-7680 APUs are now showing up; available for export at ebay, shipped from Hong Kong and marked up considerably to $80 delivered.

While there I also saw a cheap $160 eMMC 6 Watt Dell A6 Stoney ( http://www.cpu-world.com/Compare_CPUs/AMD_A6-9220e,AMD_A9-9425/ ). Seems like a travel tablet substitute; 11.6 in screen seems nice and compact. However, only 4gb, and quite possibly neither ram upgrade nor storage sata based upgrade may exist. (It depends, some dell emmc laptops are expandable, some not).
We used to call those "netbooks". ;)
 
Jan 29, 2014
273
10
91
Well, the first fm2+ BR 45W A8-7680 APUs are now showing up; available for export at ebay, shipped from Hong Kong and marked up considerably to $80 delivered.

While there I also saw a cheap $160 eMMC 6 Watt Dell A6 Stoney ( http://www.cpu-world.com/Compare_CPUs/AMD_A6-9220e,AMD_A9-9425/ ). Seems like a travel tablet substitute; 11.6 in screen seems nice and compact. However, only 4gb, and quite possibly neither ram upgrade nor storage sata based upgrade may exist. (It depends, some dell emmc laptops are expandable, some not).
I've had this laptop since March. It's decent for media consumption and simple office tasks. It's pretty slow for everything else.

The RAM is upgradable to 8gb (single stick from 4gb). There is a space in the chasis for a SATA SSD but the SATA connector isn't soldered. There is a rarer version with an a9-9420e with a 500gb HDD that could be swapped for an SSD but it's more expensive (approaching 2200u territory) so I passed on it.
 

Abwx

Diamond Member
Apr 2, 2011
8,792
122
126
Well, the first fm2+ BR 45W A8-7680 APUs are now showing up; available for export at ebay, shipped from Hong Kong and marked up considerably to $80 delivered.
53€ here, including 20% VAT.

https://www.pc21.fr/fiche/ad7680acabbox-a8-7680-3-8ghz-radeon-r7-series-i2456965.html

Same price as an Athlon 200GE, comparable perfs in MT and with a supposedly better GPU, if it wasnt officialy restricted to A68H MBs it would be a decent upgrade for PCs equiped with A6 Kaveris.
 

NostaSeronx

Platinum Member
Sep 18, 2011
2,280
86
126
So, next year is going to be interesting...

With three paths for AMD:
14LPP/12LP/7FF? with Raven2/Picasso2/Dali?... which might release or might not.
22FDX with Product Y... which might release or might not.
28ULP with Product X... which might release or might not.

14LPP/12LP/7FF is a known product with it is the ~100 mm squared Zen product or ~50 mm squared Zen2 product.
28ULP/22FDX is points towards a new core, but it might be a 2nd Excavator.

With the new core maybe being further Steamroller->Excavator; thinner libs, smaller L2, etc. With the optimization of the 7T-ULP in 28nm and the 7T-DULP in 22nm. ULP is Ultra Low Power, and DULP is Dense ULP, the reason for DULP is that 7.5T is ULP in 22FDX. Obviously, the new 7T lib is a more dense version of the 7.5T lib. Which also provides a more simple port from 28nm to 22nm.

However, 28nm ULP has a yet unnamed high performance counterpart. Which can be dropped in with a new mask at the FEOL side. Which would basically be a second Carrizo to Bristol Ridge event. There might need to be a deeper respin or new product to get the rest of the PHY enhancements: DDR4@3.2 GHz, LPDDR4@4.266 GHz, PCIeG4, 10GEthernet, plus more.

If Bristol/Stoney is Excavator mk2. Then, the 2017+ 28nm node version would be Excavator mk3. So far, I only expect Stoney Ridge to get the same treatment: Carrizo/Stoney -> Bristol/Painter___.
edit: Reason why 28-nm again => ULP option in SLP with Vdd of 0.8V results in additional 45% power reduction. Which is enough to be a shrink for Stoney. If the Core VID is the Vdd_nom then Stoney's e-suffix has 1.0 Vdd. Which with the ULP option it will become a 0.8 Vdd and get that 45% power reduction.
-> http://www.pasonisan.com/dell/inspiron11-2in1/3185-a6-9220e-a9-9420e.html
http://www.pasonisan.com/z-dell/inspiron11/a6-9220e-radeonr4.jpg
http://www.pasonisan.com/z-dell/inspiron11/a9-9420e-radeonr5.jpg
-> https://www.globalfoundries.com/technology-solutions/cmos/fdx/28nm-hkmg-technologies
Within options for SLP.


Mk2 history:
Beema -> Carrizo-L
Carrizo -> Bristol Ridge
Summit Ridge -> Pinnacle Ridge
Raven Ridge -> Picasso
Kaveri -> Godaveri
Trinity -> Richland
Zambezi -> Vishera
Temash -> Mullins
Stoney Ridge -> <<to be?>>

edit #2:
Scenario #1: 28ULP 9T
- Same design, power reduction; Carrizo -> Bristol Ridge similarity.

Scenario #2: 28ULP 7T 1st, 22FDX 7T 2nd
- New core, successor to 15h?
- Smaller branch predictor and L1i. New execution; unified integer and floating point. Still two core design, but shared FPU is dropped as HPC isn't the focus. Smaller L2 cache?
- There is a mention of 10T combined with the 7T library. Maybe, first standard mixed-cell height lib?
 
Last edited:

amd6502

Senior member
Apr 21, 2017
422
48
61
Well it sure could be a relatively cheap project. So this is what the low end lineup would look like (* means guess):

--28nm Stoney (very bottom end) 2c/2t + 3 CU: primarily for mobile 6w-15w, secondarily for bottom end bga OEM bare basic desktop 25w, thirdly for embedded.

-- 22fdx BR-L* (possible Stoney successor) 4c/4t or 2c+2c big-little*** + 3 CU: primarily for budget mobile ~10W and secondarily for low wattage budget desktop 15w-35w. (Since dual thread will cut it only so long, this is needed to remain competitive over the next few years with atom and acorn socs.)

Like nosta mentions, could be cut down core. But I think only minor change like halving BR's 2MB L2 cache to Stoney's 1MB L2.

-- 12nm RR-L* (Raven lite. RL is Bristol successor) 2c/4t + 5 CU (or maybe 4 to 6 CU) for both mainstream value mobile and budget desktop (Athlon 240ge and 220ge). dual core but upgraded to Zen+, and 12nm vega gpu shrink.

-- 14nm RR (good old RR) 4c/8t + 11 CU.

--12nm RR shrink*. (like Raven lite but 4c + 10CU ). mainstream everything: mobile 6w-25w. oem DT; consumer box APU 35w-65w.

alternatively less likely, rather than a 12nm RR shrink we could get something more exotic like a 12nm 6c APU.


*** comment on big little. Post-Piledriver CMT naturally does big+little like a champ via software. on linux:
cpufreq-set -c 0 -u 1.4GHz
cpufreq-set -c 1 -u 3.6GHz
cpufreq-set -c 2 -u 3.6GHz
cpufreq-set -c 3 -u 1.4GHz
Then have the OS taskset niced tasks to "little" cores 0 and 3. (Logical cores 0,1 are physical module 1; and cores 2,3 are module 2.)

So simply halving BR's 2MB L2 and writing pretty minor software updates would suit a big+little configuration. Given the node shrink it would be well smaller than Stoney and better in every way: threads, wattage, perf.
 
Last edited:

NostaSeronx

Platinum Member
Sep 18, 2011
2,280
86
126
Three keywords: Cost-effectiveness, Ultra low-power, and consumer-oriented Speed-demon or high frequency design.

Malta is returning to 28-nm, but only for the ULP 0.8 volt and UHP? 0.65? volt nodes. Dresden is shifting away from SLP/HPP to the newer 28-nm options.
by the way: 28SLP is 1.0 volt and 28HPP(+custom) is 0.85/0.73 volt for vdd_nom.

The 28-nm node at the moment is the cheapest high volume part at GlobalFoundries. With 22FDX->12FDX following as they get past 25% foundry utilization. That quarter and more is of Chengdu/Dresden capacity being 22FDX. 1/4th of Chengdu and 1/4th of Dresden, not the combined between both.
...
The cut down is for more transistors for higher frequency and wider scaling. The increase in transistors are hidden within a smaller average track height.
Excavator -> New module(FDX: 28BLK/22FDX/12FDX)
1x power + 1x perf => <0.5 power at one end + >1.25x perf at another end.

Example:
28BLK w/ 2017 transistors with 2019 enhancement: 3 watt - 2 GHz base to 10 watt - 3 GHz base
22FDX w/ 2019/2020 enhancement: 3 watt - 3 GHz base to 10 watt - 4 GHz base

Q1-Q2 2019 for 28BLK, then Q4 2019 to Q1 2020 for 22FDX?

Why the 28BLK SoC isn't much faster than Stoney Ridge is basically because of upgrades of GPU/NB/IO. -> GFXIP, VCN?, PCIe4, DDR4-3200, etc.
That or it is the Stoney Ridge rebrand.
 
Last edited:

hojnikb

Senior member
Sep 18, 2014
543
24
91
Three keywords: Cost-effectiveness, Ultra low-power, and consumer-oriented Speed-demon or high frequency design.

Malta is returning to 28-nm, but only for the ULP 0.8 volt and UHP? 0.65? volt nodes. Dresden is shifting away from SLP/HPP to the newer 28-nm options.
by the way: 28SLP is 1.0 volt and 28HPP(+custom) is 0.85/0.73 volt for vdd_nom.

The 28-nm node at the moment is the cheapest high volume part at GlobalFoundries. With 22FDX->12FDX following as they get past 25% foundry utilization. That quarter and more is of Chengdu/Dresden capacity being 22FDX. 1/4th of Chengdu and 1/4th of Dresden, not the combined between both.
...
The cut down is for more transistors for higher frequency and wider scaling. The increase in transistors are hidden within a smaller average track height.
Excavator -> New module(FDX: 28BLK/22FDX/12FDX)
1x power + 1x perf => <0.5 power at one end + >1.25x perf at another end.

Example:
28BLK w/ 2017 transistors with 2019 enhancement: 3 watt - 2 GHz base to 10 watt - 3 GHz base
22FDX w/ 2019/2020 enhancement: 3 watt - 3 GHz base to 10 watt - 4 GHz base

Q1-Q2 2019 for 28BLK, then Q4 2019 to Q1 2020 for 22FDX?

Why the 28BLK SoC isn't much faster than Stoney Ridge is basically because of upgrades of GPU/NB/IO. -> GFXIP, VCN?, PCIe4, DDR4-3200, etc.
That or it is the Stoney Ridge rebrand.
Any rough ideas how much does it cost to produce and package a product like Stoney Ridge ?
I always wonder if it would be possible to scale down x86 to 5-6$ SoC, which are used in likes like cheap tablets, SBCs etc...

It would be quite fun to see a SBC like Orange Pi PC (or PC2) which retails for 15/20$ but with a x86 based cpu.
 

amd6502

Senior member
Apr 21, 2017
422
48
61
Well, a sub 15 watt (~10w) pocket computer (approx NUC sized) with a sodimm slot or two would be great. Slightly smaller than NUC but approx fitting on the footprint of a 2.5in drive (3in x 4in) would be ideal.

I don't think trying to keep up with ARM prices is a good idea (they have the low segment), but having better features than Pi &co would be all that's needed to make them appeal, even at large price premiums. Old DDR3 sodimm support would be useful here imho since there's such an abundance of recycled RAM out there.

Q1-Q2 2019 for 28BLK, then Q4 2019 to Q1 2020 for 22FDX?

Why the 28BLK SoC isn't much faster than Stoney Ridge is basically because of upgrades of GPU/NB/IO. -> GFXIP, VCN?, PCIe4, DDR4-3200, etc.
That or it is the Stoney Ridge rebrand.
I would have hoped 22FDX was further along; that's still a bit of a wait!

Even so, imho no need for more 28nm APUs, Stoney should hold up well most of H1 2019, and anything higher powered will be from the zen/finfet (Raven Lite dual core die).

28BLK already had such a great (~half decade) run. Time to move on now though. I think 12LP (and successors, and alike) will be the new 28nm and should have just as good of a run, or longer. If GF could also license Samsung's 11LPP it'd be a good improvement of 14LPP and might bring the $/transistors down more and fit well for mobile <=10w APUs.

Faster DDR4 speeds aren't really important for the low end, since low end GPUs will be fine at single channel 2133 or 2400, and because higher frequency ram (given equal quality ram) also means more wattage consumed by the RAM.
 
Last edited:

dark zero

Platinum Member
Jun 2, 2015
2,500
5
61
22nm FDX is deader than dead. Only useful if a Raspberry Pi happens.
 

NTMBK

Diamond Member
Nov 14, 2011
8,214
173
126
22nm FDX is deader than dead. Only useful if a Raspberry Pi happens.
Nah, I'm sure it will still happen- lots of things it could be useful for. Just not high performance APUs.
 

amd6502

Senior member
Apr 21, 2017
422
48
61
could be good for low end APUs though, especially for sub 10w low end mobile.
 

ET

Senior member
Oct 12, 1999
508
21
91
--28nm Stoney (very bottom end) 2c/2t + 3 CU: primarily for mobile 6w-15w, secondarily for bottom end bga OEM bare basic desktop 25w, thirdly for embedded.

...
At this point I'm starting to feel like all these potential chips are rather pointless. Assuming that Raven Ridge 2 has the configuration of a Ryzen 3 2200U or Athlon 200GE, i.e., 2C/4T+3CU, and assuming that RR2 is the "Raven Ridge 2018" AMD talked about as part of its 25x20 initiative, so would likely fit into 6W-15W (35W for desktop), then what's the point of other chips on other processes?

Edit: After reading in AMD's 25x20 report card Mark Papermaster's comment that this is going to be the fastest APU AMD has ever made, I think it's obvious that the assumption that a cut down Raven Ridge is the promised more efficient CPU is wrong. Still, even if he's talking about Picasso, a cut down version of that would still achieve the desired characteristics, so it doesn't change the idea.

Sure, simpler chips on larger processes would be cheaper to produce, and perhaps even lower power, but what's the market for them? If Zen already covers most bases, why would AMD dilute its CPU message by adding an Atom? With Bristol/Stoney AMD had consolidated its offering: same architecture for both low performance, low power and higher performance, higher power. The rumours of RR2 as a cut down RR fit with this nicely.

With Ryzen, AMD has managed to gain back the image of producing high performance CPUs. It also managed to get what it wanted on a financial level: high margin CPUs. Sure, there's a market for cheaper, weaker CPUs, but why would AMD go there if its Zen solutions can fit everything from low end, low power to high end?
 
Last edited:

NostaSeronx

Platinum Member
Sep 18, 2011
2,280
86
126
Any rough ideas how much does it cost to produce and package a product like Stoney Ridge ?
More than Allwinner, but less than Athlon 200GE.
I would have hoped 22FDX was further along; that's still a bit of a wait!
22FDX is technically complete. While majority of research -> product for 22FDX won't be complete till 2021. Compared to 12FDX which also gets the weird completion thing and won't be done till 2023.
Logic1-enhancement(22FDX-base) is due to be finished by January 1st, 2019
Logic2-enhancement(22FDX-ULP/ULL/RFA/etc) is due to be finished by July 1st, 2019
^-- I might be off with this.

I do not know if this involves enhancement in 22FDX+(22FDX w/ 22FD+). If it relates to 14FDX which utilizes 14FD+ which is related to 22FD+. If 12FDX will be now using Gen 1 7FD or Gen 2 10FD, both of which use sSOI to get 50% power reduction over Leti's 14FDSOI. Which will allow a node shrink from 14FDX/22FDX+ and two node shrinks from 22FDX of performance gain/power reduction.
Sure, they'd be cheaper to produce, and perhaps even lower power, but what's the market for them? If Zen already covers most bases, why would AMD dilute its CPU message by adding an Atom?
The dilution is maybe because of GlobalFoundries.

The market is budget and emerging which is extremely low cost. Where Excavator(Stoney Ridge) is the dark horse beating out Atom, Core, Jaguar, Zen. However, Stoney Ridge was created because the then 28nm enhancement allowed it. So, an architecture that doesn't need those enhancements would be beneficial on 28-nm and later.

The one guiding the core specifics isn't AMD, but GlobalFoundries.
 
Last edited:

amd6502

Senior member
Apr 21, 2017
422
48
61
You both have really good points.

- why drag prices down on the low end down the cliff
- the production cost of Raven-L may not be significantly pricier than Bristol-L (I would think it is significant, though I think the point is partially valid and can be argued)
- the performance difference and price segmentation between RL and BL would be somewhat small so this additional BL project is almost superfluous.

on the other hand

+ the lower number of transistors and cheaper node adds up to significantly lower production cost (a controversial view that depends on estimate and opinion)
+ there is a potentially massive unit volume market for emerging market, and enough competition that Stoney won't hack it for too long (from atom and KX/via).
+GF reasons nosta mentions. I think finfet tech is licensed from samsung to GF. The cost is forwarded to AMD, and so the bottom line is that 22FDX pricing should have big advantages over 14LPP and related.

I wanted to add to Nosta's points

+ The addition of a BL project allows segmentation lineup to relax. This allows RL to be slightly larger in die area and GPU performance; instead of a minimal 3CU in Raven Lite , they add a tiny bit of die area and raise it to 5CU. That would give it the ability to keep up with BR's flagship A12 in GPU performance, and allow RL to be a serious contender for desktop low end gaming (including use in 2nd gen Atari VCS).

+ you don't want to loose this low end and the emerging markets to total acorn soc dominance, so you do want somewhat performance limited but perf-watt competitive APUs with very discounted pricing, and the existing excavator+22fdx combination is a very cost effective way (cheap and low risk ) to do it.
 
Last edited:

NostaSeronx

Platinum Member
Sep 18, 2011
2,280
86
126
So, straight off..

For 22FDX, 1.5x frequency gain does not require body biasing or increased voltage scaling for the e-suffix Stoney Ridge designs.

A9-9420e -> 2.7 GHz @ ~1.0v VID to Future design => 4.05 GHz @ ~1.0v VID to ~0.9v VID
A6-9220e -> 2.4 GHz @ ~1.0v VID to Future design => 3.6 GHz @ ~1.0v VID to ~0.9v VID
E2-9000e -> 2.0 GHz @ ~0.9 VID to Future design => 3.0 GHz @ ~0.9v VID to ~0.8v VID

In this case, ABB can be used to go even further for increased leakage(10W to 25W TDP) or to clamp down on power(sub-6W TDP).

With the CPU core however, AMD/GF would most likely target two post-Excavator style of designs:
- First is Migration, Optimization, Minimal Redesign => Ex: Piledriver -> Steamroller -> Excavator
- Second is a full Redesign => Ex: Bobcat -> Jaguar -> Zen
Either way we get a new core with a new name on a new process. Of the process which has three major nodes: 22FDX, 12FDX, 7FDX. 7FDX is in advanced pathfinding and isn't officially licensed, trademarked, restricted by GlobalFoundries yet.

With the first, the design remains dual-core and gets the above frequencies.
With the second, the design goes quad-core and gets the median of old and new frequencies:
FD2A => 3.375 GHz boost for quad.
FD2B => 3 GHz boost for quad.
FD2C => 2.5 GHz boost for quad.
With dual-core boosts probably peaking out as the above via ABB.

Either 22FDX design will definitely drop certain Stoney Ridge IP for Raven Ridge IP;
Infinity Fabric, Display Controller Next, Video Codec Next, the Sensor Hub, etc.

On the two major interconnects...DDR PHY and PCIE PHY. We should be looking at PCIe 5.0 and DDR5.
DDR5 mainly for the wide salvage zone: 3200 MHz to 6400 MHz is 3.2 GHz wide. Where as 2400 MHz to 3200 MHz is 0.8 GHz wide. So, DDR5 will quickly become cheaper than DDR3 and DDR4, simply because of frequency salvaging.

PCIe 5.0 support is mostly for switches and not for direct connections. For example: 4x PCIe 5.0 to 4x 2-lane PCIe 4.0 to 4x 4-lane PCIe 3.0 to 4x 8-lane PCIe 2.0 to 4x 16-lane PCIe 1.0, etc.

The SoC would be a wide-utility application processor. Thus, the possibilities are only limited by the designers who want to use it.

On the side: 22FDX is aggressively pointed at dual-core. 12FDX is aggressively pointed at quad-core.

Also: https://i.imgur.com/v1cwaGO.png
v0.2 PDK 12FDX in October 2018. Compared to 22FDX v0.2 in April 2016 and v1.0 in Q1 2017.
Given that info, then 12FDX will be ready in Q3 2019 for ramp.

// 22FDX => 14FDSOI 90CPP relaxed to 104CPP for Dense/Perf libs and 116CPP for Low-power libs
// 12FDX => 10FDSOI 64CPP relaxed to 78CPP? for Dense/Perf libs and 84CPP for Low-power libs
// 7FDX => 7FDSOI 46CPP/48CPP relaxed to 56CPP? for Dense/Perf libs and 64CPP? for Low-power libs
 
Last edited:

amd6502

Senior member
Apr 21, 2017
422
48
61
I like those frequencies, but I don't think dual core without SMT or some primitive multithreading would be a good idea for an upcoming project, esp'y one so far out.

22FDX is technically complete. While majority of research -> product for 22FDX won't be complete till 2021. Compared to 12FDX which also gets the weird completion thing and won't be done till 2023.
Logic1-enhancement(22FDX-base) is due to be finished by January 1st, 2019
Logic2-enhancement(22FDX-ULP/ULL/RFA/etc) is due to be finished by July 1st, 2019
^-- I might be off with this.
 

amd6502

Senior member
Apr 21, 2017
422
48
61
well, with intel joining the big+little crowd (not surprising) it'd have been nice if there were that XV+2 or 4 puma mini APU we'd all been hoping for. aSymmetric multithreading mode could be another way that Zen could achieve similar results, or it could be roughly emulated by XV quadcore too by asymmetric frequencies.
 

NostaSeronx

Platinum Member
Sep 18, 2011
2,280
86
126
I like those frequencies, but I don't think dual core without SMT or some primitive multithreading would be a good idea for an upcoming project, esp'y one so far out.
I think the delay was at first mostly GlobalFoundries fault. Which is mostly tied to capacity(SOITEC) and to the perf/yield tweaks(22FDX Gen2 through Gen4).

The second is AMD which shifted budget towards Zen and either cancelling or halting other projects. With Zen(Zen2) finished and cashing(pre-ordered) in, there is/was no stopping the FDX processor.

Which has lead to the awkward 22FDX APU. The major foundry to pay attention to is Fab11 from GlobalFoundries. Which has the completed Phase 1 moving from GloFo's mainstream nodes to the FDX nodes. 22FDX at Chengdu now peaks at 85k wpm with P1/P2, rather than just P2 at 65k wpm.

New cheap lane: Chengdu, China to Suzhou, China... for Diffused in China and Packaged in China.
Mature cheap lane: Dresden, Germany to Penang, Malaysia... for Diffused in Germany and Packaged in Malaysia.
---
it'd have been nice if there were that XV+2 or 4 puma mini APU we'd all been hoping for.
I do not expect to see v1 big.Little or v2(DynamIQ) big.Little in any of the FDSOI nodes. Zen will stay on FinFETs/NextGen(Nanosheets), and designs that use CMT will stay on planar(FDSOI).

Zen races for desktop to HPC. High TDP/High Frequency/Big IPC focus
// Maintain TDP or increase TDP to get higher IPC to Frequency gains.
This core races for fanless to laptops. Low TDP/High Frequency/Big EPI focus
// Extends from Steamroller 35W -> Excavator 15W -> This core 7.5W
 
Last edited:

Similar threads



ASK THE COMMUNITY

TRENDING THREADS