Monday, July 8th 2013

Engineering Sample Of AMD Steamroller Based APU Spotted

Hardware news site WCCF Tech spotted an interesting entry listed in the Bionic research database. The ES (Engineering Sample) chip could be a part of AMD's next-generation APU series featuring the new and improved Steamroller core. While we don't expect performance to increase by leaps and bounds, but Steamroller builds on the Bulldozer architecture and has a target to offer as much as a 30% improvement in performance over the original core.
The ES code 2M186092H4467_23/18/12/05_1304 tells us even more. According to earlier observations (here and here), the four numbers in the middle part tell a bit about clock speeds. If the first one is not 00 (no turbo, see Kabini ES), it indicates a turbo clock of 2.3GHz. The "18? stands for 1.8GHz nominal frequency. I'm not so sure about the "12?. It could stand for 1.2Ghz North Bridge clock. Finally the "05? indicates a 500MHz GPU clock. The right part "1304? is the GPU code, which - thanks to earlier revelations - can be identified as AMD1304.1 = "KV SPECTRE MOBILE 35W (1304)".
A 2.3 GHz Turbo core is pretty low, which can be attributed to the early state of the Engineering Sample. Hopefully clock speeds hit further north of just 1.8 GHz CPU and 500 MHz graphics, especially for the 35W part. The next-generation chips will be manufactured on the new bulk 28nm manufacturing process at Global Foundries.
Source: WCCF Tech
Add your own comment

47 Comments on Engineering Sample Of AMD Steamroller Based APU Spotted

#26
TheinsanegamerN
FourstaffNo this is on GloFo's 28nm process (last sentence), which made me wonder. TSMC has a stable 28nm though, so it wouldn't be a surprise if they used TSMC instead.
whelp, i missed that part. they moved kabini and temash to TSMC, but not their main chips it appears...
Posted on Reply
#27
BbigTree
Over_LordHardware news site WCCF Tech spotted an interesting entry listed in the Bionic research database.
Wrong! Coz From WCCF Tech article every first paragraph:
AMD’s Steamroller APU has been spotted listed under the Bionic research database by Dresdenboy.
~ You even linked to that article, :rolleyes:
Posted on Reply
#28
TheoneandonlyMrK
TheinsanegamerNwhelp, i missed that part. they moved kabini and temash to TSMC, but not their main chips it appears...
Glo fo are probably busy enough on ms and sony lines and no ones making 20nm useable this quarter, it is interesting watching the main competitors to intel in both fab and chip ip terms and how much they arse around to get competitive chips out.
No one bar ibm has much chance of catching Intel on the node front but it is surprising how well they are all still doing with nodes nearly twice the size of intels best.

Ot the gpu has an interesting title 1036 (;) post bin maybe)sorry 1304 could that be an ass kicking shader count??.
Posted on Reply
#29
TheGuruStud
theoneandonlymrkGlo fo are probably busy enough on ms and sony lines and no ones making 20nm useable this quarter, it is interesting watching the main competitors to intel in both fab and chip ip terms and how much they arse around to get competitive chips out.
No one bar ibm has much chance of catching Intel on the node front but it is surprising how well they are all still doing with nodes nearly twice the size of intels best.

Ot the gpu has an interesting title 1036 (;) post bin maybe)sorry 1304 could that be an ass kicking shader count??.
Intel is not using actual 22nm. Their marketing team wins again!
Posted on Reply
#30
drdeathx
brandonwh64With the APU's I wished they would add the extra L3 cache of the higher end desktop CPU's to compete with intel more on the CPU side.
??? On the APU side AMD wins in CPU performance overall against Intel's i3's. That is the apple to apple comparison. Steam roller will improve and still be better than Haswell's i3's.
Posted on Reply
#31
de.das.dude
Pro Indian Modder
aceraceNo, you just love AMD too much. Are you a fanboy? Only you can answer that. :toast:
yes i am, but yeah, look at his posts.
Posted on Reply
#32
erocker
*
de.das.dudeyes i am, but yeah, look at his posts.
Where in the news article do you see any kind of bias? Keep on topic, instead of going into this childish fanboy finger pointing that holds no relevance to the article.

Thanks.
Posted on Reply
#33
TheoneandonlyMrK
TheGuruStudIntel is not using actual 22nm. Their marketing team wins again!
I said no one else is ,not that intel are afaik intel are at 14 nm but im not 100% sure which chips are finfet I assumed they all were but np I was sharing a thought :) mostly a hope to see a decent apu with mainstream gameing gpu and possibly later on an 8 core
Posted on Reply
#34
seronx
Intel 22-nm Ivy Bridge/Haswell, CPP 90-nm/M1 90-nm
TSMC 28-nm Jaguar & Steamroller, CPP Unknown-nm/M1 90-nm


Jaguar & Steamroller are on the same node to reduce costs.
Posted on Reply
#35
xvi
Over_LordA 2.3 GHz Turbo core is pretty low, which can be attributed to the early state of the Engineering Sample.
Engineering Samples are typically clocked fairly low, iirc, since they're worried about bugs due to the architecture/process and ramping up frequencies can easily be done later.
Posted on Reply
#36
cdawall
where the hell are my stars
xviEngineering Samples are typically clocked fairly low, iirc, since they're worried about bugs due to the architecture/process and ramping up frequencies can easily be done later.
It's only got a 35w TDP 2.3ghz might just be were that falls into the scaling.
Posted on Reply
#37
seronx
AMD Eng Sample ZM232263C4450_32/23/16_9900 [Family 21 Model 16 Stepping 0]
AMD Eng Sample 2M186092H4467_23/18/12/05_1304 [Family 21 Model 48 Stepping 0]

Trinity 35W Engineer Sample: (Bulldozer rules, Burst Mode/HFM/Northbridge)
2.3 GHz -> 3.2 GHz
Kaveri 35W Engineer Sample: (Jaguar rules, Burst Mode/HFM/LFM/GPU)
1.2 GHz -> 1.8 GHz / 2.3 GHz(Single Module Burst)

Trinity:
Identifier -> Z
Segment -> M
Model -> 2322
Roadmap -> 63
Package -> C
Cores -> 4
Cache Size -> 4
Part Definition -> 50

When tested it produced 4 logical threads/one thread per core.

Kaveri:
Identifier -> 2
Segment -> M
Model -> 1860
Roadmap -> 92
Package -> H
Cores -> 4
Cache Size -> 4
Part Definition -> 67

When tested it produced 4 logical threads/one thread per core.
Posted on Reply
#38
brandonwh64
Addicted to Bacon and StarCrunches!!!
drdeathx??? On the APU side AMD wins in CPU performance overall against Intel's i3's. That is the apple to apple comparison. Steam roller will improve and still be better than Haswell's i3's.
But wouldnt you like to see it compete with I5/I7/FX on a CPU level PLUS have the nice on board GPU :) Imagine a FX 8350 with the 7660D :)
Posted on Reply
#39
drdeathx
brandonwh64But wouldnt you like to see it compete with I5/I7/FX on a CPU level PLUS have the nice on board GPU :) Imagine a FX 8350 with the 7660D :)
I would but most persons who buy these chips use a dedicated GPU anyway. Makes no sense to me to take die space for a GPU on these chips
Posted on Reply
#41
Melvis
So steamroller will be 15% more performance then the current Piledriver, thats not to bad i guess, not as high as i was expecting, but should bring it up pretty close to intel?
Posted on Reply
#42
Mussels
Freshwater Moderator
brandonwh64But wouldnt you like to see it compete with I5/I7/FX on a CPU level PLUS have the nice on board GPU :) Imagine a FX 8350 with the 7660D :)
drdeathxI would but most persons who buy these chips use a dedicated GPU anyway. Makes no sense to me to take die space for a GPU on these chips
the IGP takes up space and adds heat. great for low heat products, not great for high end ones unless you want 200W coolers.
Posted on Reply
#43
EpicShweetness
TRWOVSteamroller should catch up with Bloomfield in terms of IPC if it maintains the 10-15% improvement from Bulldozer to Piledriver.

It's 28nm. We could see a 4Ghz 95w SKU *crosses fingers* Heck, make it a 10 core 125w SKU and I'll buy 3.
Piledriver does this already check it out.
cpuboss.com/cpus/Intel-Core-i7-965-vs-AMD-FX-8350

As for a 95w part no chance maybe the lower SKU, but AMD is gonna need that higher TDP to be keep competitive. All the 8-core's thus far have been 125w.
Posted on Reply
#44
TRWOV
EpicShweetnessPiledriver does this already check it out.
cpuboss.com/cpus/Intel-Core-i7-965-vs-AMD-FX-8350

As for a 95w part no chance maybe the lower SKU, but AMD is gonna need that higher TDP to be keep competitive. All the 8-core's thus far have been 125w.
The 965 turbo to 3.46Ghz in single thread. The 8350 turbo to 4.2Ghz. Clock for clock Bloomfield still has the advantage but AMD makes up with clockspeed (as in this case).

What I meant is that Steamroller should catch up in clock for clock performance with Bloomfield, although maybe Clarkdale is going to be a better match. You can do just so much maintaining the same socket for four generations.
Posted on Reply
#45
drdeathx
TRWOVThe 965 turbo to 3.46Ghz in single thread. The 8350 turbo to 4.2Ghz. Clock for clock Bloomfield still has the advantage but AMD makes up with clockspeed (as in this case).

What I meant is that Steamroller should catch up in clock for clock performance with Bloomfield, although maybe Clarkdale is going to be a better match. You can do just so much maintaining the same socket for four generations.
?? Ummm Piledriver keeps up with IvyBridge....(not far off) where do you guys get Bloomfield or Clarksdale out of this?
Posted on Reply
#46
Nordic
drdeathx?? Ummm Piledriver keeps up with IvyBridge....(not far off) where do you guys get Bloomfield or Clarksdale out of this?
With all its cores. He is talking strait ipc clock for clock.
Posted on Reply
#47
el etro
According to my calculations here the Trinty IPC is way behind Nehalem's.
In Phenom II era Deneb has ~10% less IPC compared to Wolfdale/Yorkfield. Compared to Nehalem, haves ~35% IPC. Only Llano tied Wolfdale/Yorkfield IPC. Anyway a 20% AVG boost(over Trinty) in STR IPC can make AMD cpus much more competitive on X86 market.
Posted on Reply
Add your own comment
May 6th, 2024 06:53 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts