At the Hot Chips conference, Advanced Micro Devices has revealed the first details about its next-generation x86 core code-named Steamroller, which is presumably due in 2013. Steamroller contains a number of improvements specifically made to increase performance in traditional x86 apps, but in general Steamroller is a highly-improved Bulldozer, not something brand new.
Just like in case of the Bulldozer architecture, Steamroller x86 cores - which will power AMD's future high-performance Opteron and FX chips - will be located inside dual-core modules and therefore processors on its base should be similar by design with Orochi and Viperfish, with some minor exceptions that will not be truly important (new memory controller, different internal buses additional tweaks, etc) foe x86 performance. The main improvements will be independent instruction decoders for each core within a module, better schedulers, larger and smarter caches, more register resources and some other enhancements.

One of the reasons why dual-core Bulldozer modules [the same may be said about Piledriver] are not completely efficient is because they have only one instruction decoder for two ALUs and one FPU. With steamroller, AMD not only incorporated two decoders per module, but also increased instruction cache size (to lower i-cache misses by 30%), enhanced instruction pre-fetch (the number of mis-predicted branches is down by 20% compared to Bulldozer ) as well as improved max-width dispatches per thread by 25%. AMD believes that Steamroller will provide 30% improvement in ops per cycle.

AMD also advanced single-core execution by implementing 5%-10% more efficient scheduling, incorporated higher-capacity register files and performed some other tweaks. It should be noted that while integer pipes of Steamroller will not be too different from existing ones, the floating point pipe will be a bit redesigned. In general, AMD promises that both integer and floating point per-core performance of Steamroller will be higher than they are today with Bulldozer micro-architecture.

One of the interesting features of AMD Steamroller will be its ability to disable unused parts of L2 cache. Since not all apps are cache-bound, this may result in decreased power consumption and/or AMD's ability to boost clock-speeds of its microprocessors dynamically.

It is noteworthy that AMD decided to talk about its Steamroller micro-architecture that will be utilized inside microprocessors made using 28nm process technology approximately a year or more ahead of their roll-out. Potentially, it means that the company is pins a lot of hopes on the new micro-architecture, even more that it pins on the nearly available Piledriver.
Tags: AMD, Steamroller, Piledriver, Bulldozer, Opteron, FX, 28nm, Globalfoundries, 32nm
Comments currently:
53
Discussion started: 08/29/12 07:51:31 PM
Latest comment: 09/22/12 04:34:12 PM
Expand all threads |
Collapse all threads
[1-11]
1.
AMD appears to have some good stuff coming to fruition. If they keep hitting their 15% performance bumps per iteration they will be just fine. I expect that Steamroller will be pulled ahead slightly to allow for newer products to be rolled out sooner than the current AMD roadmap shows.
23
11
[
Posted by: beenthere

|
Date: 08/29/12 07:51:31 PM]
2.
Shame that they didn't beef up their FPU to something from 128-bit pipes to 254-bit pipes. Well, I also hope that they are also increasing base clock speeds deep into 4GHz territory and maybe even having turbo boos speeds get into 5GHz. Would be great to have Steamrollers running 5GHz on air on enthusiast systems.
Anyway, here's hoping that it can go head-to-head with Intel's Haswell.
show the post
2
5
[
Posted by: RtFusion

|
Date: 08/29/12 09:31:47 PM]
+ expand thread (1 answer)
- collapse thread
Dual 128-bit FPU has more flexibility versus a single 256-bit. And an insider already stated Steamroller Clock for Clock will be about 45% faster than today's Bulldozer with lower power consumption. Excavator coming in 2014 should be another 20% to 30% faster than Steamroller. The insider speculation stated that Excavator will be a highly enhanced Bulldozer, such as a complete overhaul of some of its key components, sticking to the Bulldozer principles, but this time performing the way it was meant to.
Steamroller is the start of this modification.
3
0
[
Posted by: nt300

|
Date: 09/22/12 02:09:42 PM]
3.
We see, SR is just a shrink of BD with some tuning.
BTW, AMD did a small step away from stupid BD module design with allocation more dedicated resources per each core.
BTW, AMD mentioned about more perf/watt and forgot to mention about more performance. Do you need another proof that SR performance will be lower than one of BD due to much lower frequency?
show the post
3
21
[
Posted by: Azazel

|
Date: 08/29/12 10:34:41 PM]
+ expand thread (1 answer)
- collapse thread
We don't actually need another proof. We need a proof.
What you mentioned isn't one.
4
1
[
Posted by: sirroman

|
Date: 08/30/12 02:38:04 PM]
4.
I can't get in what AMD engineers will do if the new software Library makes their job much better?
show the post
5
17
[
Posted by: Azazel

|
Date: 08/29/12 10:36:28 PM]
5.
3rd time is a charm??
show the post
4
18
[
Posted by: TAViX

|
Date: 08/30/12 12:02:58 AM]
6.
Still 28 nm? Yawn...that whore is all stretched out. Intel will already be running engineering wafers of 14nm Broadwell by the time Steamroller is out.
4
2
[
Posted by: AnonymousGuy

|
Date: 08/30/12 12:58:01 AM]
+ expand thread (4 answers)
- collapse thread
SB on 32 nm is better than IB 3d in smaller process ,too hot.Dont expect so much from 14 nm.
show the post
3
6
[
Posted by: Blackcode

|
Date: 08/30/12 04:36:29 AM]
Errr no. You only get thermal runaway on IVB at 4.8 Ghz, and the IPC gains over Sandy Bridge make it so 4.8 Ivy = 5.0 Sandy anyways.
2
3
[
Posted by: AnonymousGuy

|
Date: 08/30/12 11:39:47 PM]
IB is still superior to SB on both CPU and GPU side.
What more important, IB has insanely small die size. Quad core + HD4000 is only 160mm^2
Thats almost half die size of bulldozer, despite the fact that IB has GPU and a superior cpu
With 14nm, AMD is going to be at worse situation if they stick with 28nm. AMD won't make lot of profit if they sell a big die size at cheaper or same price as much smaller die sizer
3
4
[
Posted by: maroon1

|
Date: 08/31/12 03:32:22 AM]
Intel's 22nm 3D transistor is amazing. If you delid 3570/3770K processors, you get a 20*C drop in temperatures. The problem is the solder between the CPU die and the heat-spreader.
17
6
[
Posted by: BestJinjo

|
Date: 09/01/12 10:57:31 PM]
7.
(at least) 10% clock increase * 30% increased IPC = 43% increase in net performance per core over Bulldozer. Sounds interesting!
3
5
[
Posted by: jasmith007

|
Date: 08/30/12 01:49:46 AM]
+ expand thread (1 answer)
- collapse thread
Sound interesting if it is true
But giving the fact that those are AMD claims, it is highly unlikely
6
0
[
Posted by: maroon1

|
Date: 08/31/12 03:40:05 AM]
8.
AMD needs to correct IPC.
4
6
[
Posted by: Blackcode

|
Date: 08/30/12 04:37:53 AM]
9.
Most consumers will be perfectly happy with Piledriver and Steamroller and that's all that really matters.
15
14
[
Posted by: beenthere

|
Date: 08/30/12 06:07:09 AM]
10.
Will the Steamroller run on AM3+ motherboard ? Or they will do an AM4 for this CPU ?
8
2
[
Posted by: ojohn

|
Date: 08/30/12 10:35:30 AM]
+ expand thread (1 answer)
- collapse thread
AvonX & jmlxg are the same person , and he is a Troll so ignore him !!!
13
3
[
Posted by: ojohn

|
Date: 08/31/12 07:25:01 AM]
11.
AMD is very good at talking these days.
AMD was telling us how good bulldozer was, and they were cherrypicking some benchmarks to show how good bulldozer is.
You can't say that Steamroller is going to be great based on AMD words. Let us wait for unbiased 3rd party reviews.
7
3
[
Posted by: maroon1

|
Date: 08/31/12 03:20:07 AM]
[1-11]