Will AMD Steamroller be their Sandy Bridge?

[bononos]

http://news.softpedia.com/news/AMD-s-Steamroller-To-Be-Faster-than-Intel-Haswell-289980.shtml

"Many are estimating conservative values ranging between a 20% and a 30% performance improvement over the current Bulldozer processors, but sources inside the engineering department at AMD are reportedly expecting 45% performance improvements."

"Steamroller is not Bulldozer Enhanced. F*** no. The layout might look the same but our LEGO blocks are completely different. When all is said and done we should get 45% improvement and this goes to show how the Bulldozer was f***** design. This is all what Bulldozer was supposed to be."

At least the source was right about BD being f*** design.

 

[SlowSpyder]

Do you not see where they took a step back from thuban?

The gain from thuban -> 8350 is half of the gain from 8150 -> 8350.

AMD was simply making up lost performance.

While that's true, it doesn't really apply to this scenario. AMD has the construction equipment cores going forward, so no matter where they left off before, what they're improving on started with Bulldozer. The next core (assuming there is one) will be based off of Bulldozer, not anything Phenom II.


BD to Vishera brought some IPC improvements and clockspeed improvements. Now with a new 28nm process and even more time, maybe we could see a really decent performer. But even if AMD can match an i7's raw performance, I doubt it could be done anywhere near the power levels the Intel chips use. But, even being the case that'd probably be enough for them to at least sell processors north of $200 for a change.

 

[grimpr]

FX 4300 vs i5 750
http://www.anandtech.com/bench/Product/700?vs=109

FX 6300 vs i7 860
http://www.anandtech.com/bench/Product/699?vs=108

FX 8350 vs i7 950
http://www.anandtech.com/bench/Product/697?vs=100

You can see here that with Vishera FX series they are roughly at the Nehalem level when comparing SKUs. A FX-4300 matches up pretty closely to an i5-750 and so on up the SKU chart. The problem is that Bulldozer was supposed to have at least that level of performance. AMD has been perpetually been behind schedule since the original Phenom.

It is actually slower in a few things in that link but really a lot of that is due to what being 20% of the market vs 80% of the market means for developer support than anything that can be designed for. After all AMD can't legally copy Intel's chips method of instruction support transistor for transistor, they can't even legally support every instruction set. AMD also can't control how things are coded and compiled, they don't have the market share and corresponding $ war chest.

If Steamroller is a 15-20% improvement and GlobalFoundries 28nm process performs better than their 32nm did at launch then wouldn't that make Steamroller FX SKUs similar to Sandy Bridge?

They can never touch Intels lead in single threaded ipc lead BUT with the dedicated decoders going in SR and other microarchitectual tweaks, they can leapthrough multithreaded performance above and beyond Sandy/Ivy Bridge, im pretty sure of that. The main bottleneck of the architecture, the shared decoder, is gone. Each integer core has its own independent decoder now, this alone should boost performance in single threaded ipc & total throughput.

 

[toyota]

the 8350 is already really close to Sandy and Ivy Bridge in some games so I hope Steamroller gets huge enough boost to exceed Haswell in some cases. Intel has not done hardly squat for the last 2 years when it comes to moving things forward at the same price bracket as the 2500k was at.

 

[Xpage]

I think the uops cache in SR will be a large benefit as well, i think it did wonders in Sandy and if AMD has something similar and it works at about he same performance then I expect at least a Nehalem -> Sandy IPC gain if not more, so I am hoping 15%, but I wouldn't be sad to see 20-25%. But most likely 15%

 

[sushiwarrior]

Do you not see where they took a step back from thuban?

The gain from thuban -> 8350 is half of the gain from 8150 -> 8350.

AMD was simply making up lost performance.

I'm sorry, was it said anywhere in here that we're comparing PileDriver and Thuban? Keep on topic. It's clearly shown than generational gains can easily give 10-20% performance. The half node shrink will help the power consumption part.

 

[guskline]

Most reviews on the 8150 to 8350 show 5 to 15% improvement. Be that as it may, where are any official announcements from AMD on the status of the Steamroller core? What "gossip" is there about engineering samples? We have more info on the 9590 and 9370 (Uber Piledrivers) than on SteamRoller.

 

[Enigmoid]

I'm sorry, was it said anywhere in here that we're comparing PileDriver and Thuban? Keep on topic. It's clearly shown than generational gains can easily give 10-20% performance. The half node shrink will help the power consumption part.

Sure I'm just saying that BD was WORSE that the last gen product despite being 32 nm vs 45 nm. If you take a step back then its easier to take a step forward.

8350 is marginally more efficient than thuban @ 45 nm. Thuban at 32 nm + minor tweaks would still be more efficient.

BD was simply so bad that it was easy to improve it. The better the design gets the harder it is to make improvements.

 

[NTMBK]

8350 is marginally more efficient than thuban @ 45 nm. Thuban at 32 nm + minor tweaks would still be more efficient.

Llano was 4 improved Stars cores at 32nm- and it was beaten by Trinity, which was 4 Piledriver cores. So how precisely would a 32nm Thuban be better than Vishera?

 

[BallaTheFeared]

Llano is also the Athlon II of Stars, and I don't think it was actually improved, just shrunk.

 

[NTMBK]

Llano is also the Athlon II of Stars, and I don't think it was actually improved, just shrunk.

It had some minor tweaks, actually- "claimed 6% IPC improvement". http://semiaccurate.com/2011/06/20/a-llook-at-the-llano-architecture/

 

[inf64]

Llano with no L3 has around ~6-7% higher IPC in legacy workloads Vs 45nm Stars core(with L3).

 

[BallaTheFeared]

Llano with no L3 has around ~6-7% higher IPC in legacy workloads Vs 45nm Stars core(with L3).

It's been awhile since I ran my Regor, but wasn't Athlon II a far bit slower per clock then Phenom II?

 

[littlebluesmurf]

It had some minor tweaks, actually- "claimed 6% IPC improvement". http://semiaccurate.com/2011/06/20/a-llook-at-the-llano-architecture/

As can be seen here

http://anandtech.com/bench/Product/403?vs=106

and here

http://anandtech.com/bench/Product/85?vs=403

The large L2 on Llano makes it quite hard to see if the improvement is mainly cache related or not.

 

[Enigmoid]

Llano was 4 improved Stars cores at 32nm- and it was beaten by Trinity, which was 4 Piledriver cores. So how precisely would a 32nm Thuban be better than Vishera?

Llano 3870k clocked at 3.0 ghz

Trinity a10-5800k clocked at 4.0 ghz.

33% higher clocks.

 

[NTMBK]

Llano 3870k clocked at 3.0 ghz

Trinity a10-5800k clocked at 4.0 ghz.

33% higher clocks.

So what? It's an architecture made to clock higher. What matters is the performance. And on the same process, at the same TDP, Trinity performed better than Llano (in both CPU and GPU).

 

[Enigmoid]

So what? It's an architecture made to clock higher. What matters is the performance. And on the same process, at the same TDP, Trinity performed better than Llano (in both CPU and GPU).

Didn't llano have manufacturing problems?

And yes performance did go up. However can AMD increase clockspeed another 33% on 4.0 ghz? Probably not. Most future performance gains must come from IPC and the IPC gain from llano -> trinity/richland was nonexistent. Gaining IPC is harder than gaining clockspeed.

 

[crashtech]

To the OP: I don't know if Steamroller will be AMD's SB, but honestly I see this whole Bulldozer era as AMD's Netburst. I sincerely hope they are able to move past it and come with a microarchitecture based on a different design, like Intel did when they were failing.

 

[sequoia464]

Be that as it may, where are any official announcements from AMD on the status of the Steamroller core?

I don't follow these developments as closely as most here, but I thought that AMD stated at one point they were not going to produce any straight CPU's after Piledriver - APU's only?

Did I misunderstand this??

 

[CHADBOGA]

Does anyone know when Steamroller is supposed to come out?

 

[Hitman928]

http://news.softpedia.com/news/AMD-s...l-289980.shtml

"Many are estimating conservative values ranging between a 20% and a 30% performance improvement over the current Bulldozer processors, but sources inside the engineering department at AMD are reportedly expecting 45% performance improvements."

"Steamroller is not Bulldozer Enhanced. F*** no. The layout might look the same but our LEGO blocks are completely different. When all is said and done we should get 45% improvement and this goes to show how the Bulldozer was f***** design. This is all what Bulldozer was supposed to be."

At least the source was right about BD being f*** design.

Follow the links and you get VR-Zone as a source with date and authorship of:

By Theo Valich on August 31, 2012 10:14 am

Personally, I'm not setting my hopes so high. It'll be a little while before we know for sure either way.

 

[Hitman928]

Most reviews on the 8150 to 8350 show 5 to 15% improvement. Be that as it may, where are any official announcements from AMD on the status of the Steamroller core? What "gossip" is there about engineering samples? We have more info on the 9590 and 9370 (Uber Piledrivers) than on SteamRoller.

All we know is that Kaveri (steamroller based) APU's are just now being shipped to partners to play with and test and they expect availability by the end of the year.

We know more about the new PED's (see what I did there :sneaky because they're launching this summer.

 

[LoveMachine]

I don't follow these developments as closely as most here, but I thought that AMD stated at one point they were not going to produce any straight CPU's after Piledriver - APU's only?

Did I misunderstand this??

I think their goals going forward are to simply not try to out-compete Intel in the desktop CPU space. They will still design and sell CPU-only chips, but their main focus and R and D will go towards APUs. The reality is that this makes sense for the general market. Enthusiasts that make up many of the forum members here are a shrinking lot, and the CPU speed wars really don't matter much anymore.