Amd Blog Update: Steamroller

[Nordic]

AMD Steamroller FX 9650 - 4.5 Ghz / 4.8 Ghz Turbo Benchmarks!

AMD Back in the High End
1090FX Entusiast


30% more Ops / Cycle = 20% IPC Benefit over Piledriver


Assuming 45% increase in performance vs Bulldozer (First Gen) and a 32% increase in performance vs Piledriver (Second Gen). This is taken directly from AMD!


source

IPC will lie somewhere between Nahelem and Sandy Bridge , falling behind Haswell by only 25-30%.


POV Ray 3.7 - Multithreaded Test Score

Intel

4C/8T - Intel i7 3770k : 1363

6C/12T - Intel i7 3970x : 1888

AMD

4M/8T - AMD FX 8350 : 1505

4M/8T - AMD FX 9450 : 1636

4M/8T - AMD FX 9650 : 1985

(determined with 32% increase in performance over the FX 8350)

source

Cinebench 11.5 Score - Single Core Test

Intel

4C/8T - Intel i7 3770k : 1.66

6C/12T - Intel i7 3970x : 1.62

AMD

4M/8T - AMD FX 8350 : 1.11

4M/8T - AMD FX 9450 : 1.21

4M/8T - AMD FX 9650 : 1.47


Cinebench 11.5 Score - Multi-core Test


Intel

4C/8T - Intel i7 3770k : 7.52

6C/12T - Intel i7 3970x : 10.84

AMD

4M/8T - AMD FX 8350 : 6.93

4M/8T - AMD FX 9450 : 7.55

4M/8T - AMD FX 9650 : 9.15

source

---

Due to the 20% difference in single core performance when compared to Haswell, the FX 9650 will perform on par with a 2nd gen intel i7 2600k in gaming, still falling behind the 4770k by a hair when paired with two high end graphics card solutions. For games, like Crysis 3, that take advantage of more than 4 cores, the FX 9650 will perform on par with its intel counter parts. The 32 % performance increase comes with a 20% increase in IPC (vs Piledriver) and around a 10% Clock boost.


The only downside with these chips is the overclockability. AMD generally pushes their stock clocks high enough to not leave any room for overclocking. Although AMD has put alot of effort into making these chips less power hungry, they only succeeded in allowing for higher clocks and typically higher overclock under the given TDP. The total power consumption is only 5-8% less than its FX 8350 counterpart. Unlike with the 4770k, which can handle a hefty +43% overclock from 3.5 Ghz to 5.0 Ghz , while the AMD FX 9650 will only handle a 20% overclock from 4.5 Ghz to 5.4 Ghz.

Note: It should deffinitly be noted that the intel processor overclocked to 5 Ghz will still use a whopping 40% less power than AMD's Steamroller Flagship.

The FX 9650 will lie somewhere between the 4770k and the 4960x in performance, and the price will reflect that. To directly compete with the 4770k, AMD will release a cheaper variant clocked at 3.8 Ghz with a 4.2 Ghz Turbo.

My Price estimate for AMD's flagship FX chip 4.5 Ghz with 4.8 Ghz Turbo is $449, with the lower model sporting a 3.8 Ghz / 4.2 Ghz Turbo Clock for $249. The FX 9650 will compete directly with the i7 4930k, while the FX 9450 will compete with the i7 4770k.

You Choose Who will be King!

Update: *Given AMD's flagship Kaveri APU will sport 3M/6T , (source) it wll be interesting to see if AMD releases a 6M/12T Centurian FX variant flagship in the 900-1000 USD range. (If not 4M/8T) Although this chip will be plagued with terrible overclockability, it will nevertheless give the 4960x a run for its money.

http://amdfx.blogspot.ca/2013/05/amd-steamroller-fx-9650-45-ghz-48-ghz.html

I found this interesting to read. Take what you read with a grain of salt. Figured I would post this here for discussion.

 

[Mathragh]

This seems to be just a random person taking an article from 2012 (The "source") and making up some numbers based on very rough estimates.

Sure, it sounds awesome, and I wish it were true, but there is nothing that suggests this is anything more than just someones fantasy at this moment.

 

[erocker]

This seems to be just a random person taking an article from 2012 (The "source") and making up some numbers based on very rough estimates.

Seems to be exactly what it is.

 

[RejZoR]

I'm a bit concerned about their very high clocks. When Intel was going that way, they were using NetBurst which was rubbish. With not much room around or over 4GHz mark, they'll have to do something, because they won't be able to keep on doing this forever...

 

[Mathragh]

I'm a bit concerned about their very high clocks. When Intel was going that way, they were using NetBurst which was rubbish. With not much room around or over 4GHz mark, they'll have to do something, because they won't be able to keep on doing this forever...

Do note though, that those high clocks were AMD's goal all along. Also, compared to netburst, AMD has a lot of advantages, such as a better process, resonant clock mesh, and arguably just a better design(compare the IPC of netburst with bulldozer).

Furthermore, high frequency designs aren't bad by default. There have been a lot of "speed demons" which have been very succesfull, but just not in the x86 world(where netburst actually was the only attempt).

EDIT: the link at the bottom of the linked article does sound interesting though, and is quite an interesting read

Some interesting points: most of them aimed the current bottlenecks(orly, but still), and lower latencies(not the cache latency everyone keeps going on about, but stuff that matters for instruction throughput)
-A shortening of the FP pipeline(from 4 to 3 stages)(should matter a lot for the currently weakest point of BD)
-A lot of caches and queues get increased in size
-Doubling of decoding stage(that was fairly know iirc)

 

[Nordic]

Funny little thing to note. I voted for the intel haswell on the little pole at the top. Pole says 4 votes all for steamroller. I vote again. 4 votes for steam roller.

 

[Deleted member 24505]

Funny little thing to note. I voted for the intel haswell on the little pole at the top. Pole says 4 votes all for steamroller. I vote again. 4 votes for steam roller.

I just voted too, still says 0 votes for Haswell

 

[Mathragh]

Its the same the other way around, voted for steamroller, no increase
Seems like the poll isn't updating/registering

 

[Nordic]

Might as well post this here as it is along the same lines.

 

[silkstone]

I always used to buy AMD, then went a few years without a computer before coming back. I decided to go with a C2D, then SB.

I really do hope that they start competing more at the enthusiast level. Having only 2 chip manufacturers is bad enough, having one will not be good for consumers.
(I think i had a Cyrix back in the days of 486's)

 

[Novulux]

Who says this is from AMD? It's just a blogspot page that refers to AMD as a separate entity in other posts...could've been made by anyone.

 

[Nordic]

Who says this is from AMD? It's just a blogspot page that refers to AMD as a separate entity in other posts...could've been made by anyone.

I said out of ignorance. Still though, should be taken with a grain of salt.

 

[Frick]

I really do hope that they start competing more at the enthusiast level. Having only 2 chip manufacturers is bad enough, having one will not be good for consumers.
(I think i had a Cyrix back in the days of 486's)

I'be been thinking about this. How big is the enthusiast market anyway? People don't need enthusiast parts, and it feels like less and less people are enthusiast these days anyway.. I think competing there will become less and less important. If that is the case, AMD is ahead.

 

[Nordic]

If that is the case, AMD is ahead.

With the exception of apu's, how are they? Are you thinking price/performance?

 

[jihadjoe]

So this guy extrapolates numbers, arbitrarily assuming a fixed 45% performance increase over Bulldozer, and claims they are "benchmark results."

lol

 

[Frick]

With the exception of apu's, how are they? Are you thinking price/performance?

Considering most of their lineup are APU's... I'd say they are ahead. CPU power isn't that important nowadays (for avarage users) and while they could do with lower power consumtion they are doing good there.

 

[eidairaman1]

With the exception of apu's, how are they? Are you thinking price/performance?

ya they ahead in the APU dept. Pretty good product actually, have a llano based unit here for the GP. And the way TDP is measured is different between Intel and AMD, so its a really moot point

 

[Nordic]

I thought intel across the board had better power efficiency compared to amd because of better sleep states and a smaller process node (I believe the term is called).

 

[Frick]

And the way TDP is measured is different between Intel and AMD, so its a really moot point

Just look at what they actually use. The low ends are nice (Kabini is doing very nice), but the higher end is a bit high. That is obviously because they have decent GPU's in them, but still. It can only go forward.

 

[eidairaman1]

Just look at what they actually use. The low ends are nice (Kabini is doing very nice), but the higher end is a bit high. That is obviously because they have decent GPU's in them, but still. It can only go forward.

so Kabini is the BGA/ System on a board, because I recall Richland being soon to launch

 

[Ravenas]

What socket is steamroller?

 

[de.das.dude]

i hope it wont be a new socket. just got a new motherboard.....

 

[erocker]

Yep.

 

[Frick]

What socket is steamroller?

I think I've seen at least rumours that it's AM3+. Don't remember at all where I have seen it, but I think that is a possibility at least.

 

[Mathragh]

What socket is steamroller?

99% certain its AM3+. Dont remember the source, but i remember reading about amd saying SR being AM3+. It'll prolly be the last cpu for that socket though