"Steamroller" could crush "Haswell"?!

[Centauri]

I'm as big of an AMD fanboy as anybody, but the entire premise behind this thread is stupid.

 

[2is]

chernobog, do you know galego?

I was going to say the same thing.

Ok, I wasn't alone on this one lol

Btw, I calculated Haswell is going to be faster than Steamroller. My calculation is based on seeing the same trend now that I've seen every time AMD is about to release a new CPU for the last decade.

AMD releases details
Enthusiasts get exited, think its the next "Athlon 64"
Enthusiasts make calculations
AMD releases processor that's marginally better and sometimes worse than its predecessor.

 

[NostaSeronx]

Btw, I calculated Haswell is going to be faster than Steamroller. My calculation is based on seeing the same trend now that I've seen every time AMD is about to release a new CPU for the last decade.

The first rule of showing calculations is showing your work.

I showed mine.

If you up the clock on Berlin, to match the i7 4670K clock rate, it would score somewhere around 6.8 pts on Cinebench R11.5. If you upped it to 4 GHz like the FX-8350, it would be around 8 pts.
http://www.bit-tech.net/hardware/2013/06/12/intel-core-i5-4670k-haswell-cpu-review/6

 

[2is]

The first rule of showing calculations is showing your work.

Poor choice of words on my part. Mines really more of a prediction based on historical evidence/events. Namely people like the OP who have attempted similar calculations that never came to fruition for the last 10 years.

And every time AMD fails to deliver, there's an excuse.

Apps aren't optimized for AMD (yet)/(still)
Intel sabotaged performance on AMDs processor
Windows scheduler doesn't know how to allocate resources for AMD processors.
Old staff was the problem

I'm sure I'm missing a few.

 

[rgallant]

amd should have walked away from gf , when they sold out.
at least they would have a shot at 20nm Q1 2014 with tsmc , instead of Q3-4 2014 32nm server chip release.

 

[crashtech]

Now that I think about it, there is just one thing missing from this thread. The letter "A."

"A Steamroller could crush Haswell!"

Although it might be hard to find one in operating condition, since virtually all rollers are diesel powered these days.

 

[Greenman]

"Could"... "May".....

All pointless time wasting speculation. I am reminded of something Red Dawn used to say over a decade ago on these very forums. He probably still says it wherever he is. Forgive me if I don't get it exactly correct.

"If my aunt had balls she would be my uncle."

I know where he is, and I've never seen him write that. But it certainly sounds like something would say.

 

[monstercameron]

If a certain member starting with G - stops activity posting wise.. well just saying.


@Chernobog:

Dear dis-illusioned AMD fan (because you are one, in my humble oppinion).

Remember Bulldozer?

Remember how estimating from internal slides went?

Remember how listening to a direct employee from within the Marketing Arm of AMD went?

What makes you think on a worse process too - that AMD can somehow increase ST\MT 20% on constrained budgets - after laying off a SIZEABLE chunk of they're talent.

Can somehow turn it around - on a uARCH that's been described as well nearly as bad as Netburst - within 2 small years?


Do you think this is "logicly fallable" within your own personal judgement?

Do you think it's probable given the scenario AMD is in?



If yes - please stop posting any pro AMD thread again - and wait til SR is released.
(especially the FX as you claim) - and then watch and enjoy rubbing it in all the "pro" intel people here.

Why would you (after twice in a row failing with uArch's upon released performance results vs pre-marketing slides) - not atleast say "Hey... it looks pretty good - but lets wait til we get actual results!" ?

Why?

jaguar was right on point...

 

[moonbogg]

What I am just dying to know is will Steamroller have solder or thermal paste!?

 

[sourn]

I've got nothing against AMD, but with the way they're going all I can say is I'll believe it when I see it.

 

[monstercameron]

I've got nothing against AMD, but with the way they're going all I can say is I'll believe it when I see it.

again they were relatively accurate with jaguar...

 

[2is]

again they were relatively accurate with jaguar...

And inaccurate with everything else...

 

[monstercameron]

And inaccurate with everything else...

no, pretty spot on with vishera and richland too...

 

[2is]

And still nothing to challenge Intel. That's the rub. AMD was able to do this once, sice then the enthusiasts have had a bad case of wishful thinking for 10 years. Erecting threads like this. It fizzles out as soon as the product releases, rinse and repeat a year later.

Vishera was an improvement over Zambezi. Zambezi should have been performing at Vishera levels to begin with and Vishera should have been an improvemt over that. Pretty weak argument when it only works when you factor in just how bad its predecessor was.

 

[Vesku]

It should be noted all AMD has shown of Kaveri is the chip itself. They have yet to put out any estimates along the lines of what they did for Vishera and Jaguar.

 

[monstercameron]

And still nothing to challenge Intel. That's the rub. AMD was able to do this once, sice then the enthusiasts have had a bad case of wishful thinking for 10 years. Erecting threads like this. It fizzles out as soon as the product releases, rinse and repeat a year later.

Vishera was an improvement over Zambezi. Zambezi should have been performing at Vishera levels to begin with and Vishera should have been an improvemt over that. Pretty weak argument when it only works when you factor in just how bad its predecessor was.

not the point though, their marketing was accurate for the past 3 releases

 

[2is]

not the point though, their marketing was accurate for the past 3 releases

It's exactly the point if we're sticking to the topic of Steamroller crushing Haswell.

Besides that, Vishera is still a bad example. "accurate" marketing by virtue of inaccurate marketing on it's predecessor isn't something to brag about. That's an important distinction. If their claims were accurate for Zambezi and accurate again for Vishera, an 8 core Vishera would easily compete with i7's in just about anything.

 

[monstercameron]

It's exactly the point if we're sticking to the topic of Steamroller crushing Haswell.

Besides that, Vishera is still a bad example. "accurate" marketing by virtue of inaccurate marketing on it's predecessor isn't something to brag about. That's an important distinction. If their claims were accurate for Zambezi and accurate again for Vishera, an 8 core Vishera would easily compete with i7's in just about anything.

let me clarify, I am not talking about steamroller, just wanted to debunk the notion that amds marketing is lying or very inaccurate.

 

[MisterMac]

jaguar was right on point...

Jaguar is not Big Core.


I expect they'll become even more of a competitor in those markets - if they keep up their work cadence first set with brazos



...unfourtunately for reasons perhaps questionably ethical - it doesn't seem like Temash\Kabini are winning as much as they "should".


I doubt we'll see FX 4M SR - if so it'll be likthe 9xxx chips.
Marginal sales - stupid crazy prices.
They can't afford to sell 4M cores @ 200 USD.

And if SR is godsend - then they'd tell us.
If Intel had a SB or Conroe up their sleaves with Haswell they'd tell us.

They had a perf\watt mobile chip on their sleave - not a Desktop performance one.

 

[NTMBK]

I should really bookmark this thread just for future comic effect. There's no way in hell your predictions will be as accurate as you think they will, chernobog.

 

[ShintaiDK]

And if SR is godsend - then they'd tell us.
If Intel had a SB or Conroe up their sleaves with Haswell they'd tell us.

They had a perf\watt mobile chip on their sleave - not a Desktop performance one.

Exactly.

K8 for desktop we heard about half to a full year in advance or so with full fanfare.
Conroe for desktop.
Nehalem for servers.
SB for desktop.
IB for laptops.
HW for laptops.

AMD tried the lies for PR with Phenom and Bulldozer uarchs. But it was so obvious it wouldnt happen.

 

[Ancalagon44]

Exactly.

K8 for desktop we heard about half to a full year in advance or so with full fanfare.
Conroe for desktop.
Nehalem for servers.
SB for desktop.
IB for laptops.
HW for laptops.

AMD tried the lies for PR with Phenom and Bulldozer uarchs. But it was so obvious it wouldnt happen.

I dont think AMD does as much as it used to, probably after it got rightly burned after the BD release. I mean, look at Jaguar - it did everything AMD said it would, but they didnt shout about it from the rooftops.

 

[ShintaiDK]

I dont think AMD does as much as it used to, probably after it got rightly burned after the BD release. I mean, look at Jaguar - it did everything AMD said it would, but they didnt shout about it from the rooftops.

Jaguar is the product they needed a year ago. So thats why.