I believe that those roadmaps were done with the belief that BD was going to be faster than it is. Maybe there's nothing they'll be able to do about it? Maybe they can get some of the performance though that they were originally shooting for and then have further 10%-15$ improvements on top of that. Bulldozer, as it is, just isn't at all competitive. If they can't get more substantial improvements (across the board, power, clocks/IPC) then they need to just move on from it. A six core Thuban at the same clocks beats an 8 core Bulldozer in both single and multi-threaded work loads. That's really Bulldozers failing. I never expected it to be faster per core/per clock than Intel's current offerings. I did expect it to beat PhII though.
Its there to beat BD. Its AMD's fault the 2700k is not much of an improvement, if intel needed to up clocks 500Mhz to beat BD they would have. But since AMD screwed up BD and intel only needed 100Mhz to beat it thats what we ended up with. Obviously its not ment as an upgrade for current SB users.
Agreed.
But that extra 15$ above the 2600K price is just adding insult to injury, isn't it? I mean, what the heck for? A measly 100MHz more; what's more, compared to another multiplier unlocked part. And 9 months later?!
Maybe we'll find out that Intel have been binning the very best SBs for a while now and that this SKU will routinely hit 5.5GHz on air, with higher multipliers for those people who don't mind pushing 1.55+ volts through Sandy Bridges.
Though I'm not holding my breath.
Agreed.
But that extra 15$ above the 2600K price is just adding insult to injury, isn't it? I mean, what the heck for? A measly 100MHz more; what's more, compared to another multiplier unlocked part. And 9 months later?!
Maybe we'll find out that Intel have been binning the very best SBs for a while now and that this SKU will routinely hit 5.5GHz on air, with higher multipliers for those people who don't mind pushing 1.55+ volts through Sandy Bridges.
Though I'm not holding my breath.
Yeah but the problem is how do you take advantage of a gpu in your software? Lets say I have an array of 16 bit unsigned integer variables. And lets say I want to multiply each one by 2. Sounds like a great job for a big wide SIMD in a gpu! But how can it possibly be faster to send that entire array out of my L1 or L2 cache all the way to the L2 cache onboard the gpu to be processed, and then returned back to the cpu? How can that be any faster than just running an SSE inside the core? Obviously the only way it can be faster is if it is a pretty big array. Like say an array of 65536 16 bit values. But even then I wonder how much of a gain would I really see? I have no way to quantify it, but my gut tells me it wouldnt buy me much performance. AMD may have a gpu on the same silicon. But until they place their radeon cores right inside the cpu fpu, it doesnt mean much.
What I want to see is a 4 core bulldozer module with a shared GCN fpu consisting of at least four 16-wide vector SIMDs. Yes, an entire GCN compute unit inside each bulldozer module. Then we'll have something.
Hmm, try SIMD vector(SSE) on 4 Module 8 Thread Bulldozer, I believe it will be much faster than 4 Core 8 Thread SB
Agreed.
But that extra 15$ above the 2600K price is just adding insult to injury, isn't it? I mean, what the heck for? A measly 100MHz more; what's more, compared to another multiplier unlocked part. And 9 months later?!
Maybe we'll find out that Intel have been binning the very best SBs for a while now and that this SKU will routinely hit 5.5GHz on air, with higher multipliers for those people who don't mind pushing 1.55+ volts through Sandy Bridges.
Though I'm not holding my breath.
Nov 15th can not get here fast enough.
Have you looked at CPU prices before? Eeveryone does this.
Is AMD 'price-gouging' X4-980 by adding $25-30 for 200 mhz over the 970?
These type of statements are getting rediculous. 4-5% more $$$ for about 3% more performance is business.
With a complete lack of competition at the high end, innovation is going to be stifled for at least 2-3 years.
I hope you all like your current CPUs.
With a complete lack of competition at the high end, innovation is going to be stifled for at least 2-3 years.
The best AMD seems able to do is a 4.4ghz Phenom II.
It will be interesting to see what AMD does with an absolutely puny R&D budget going forward.
Intel can just bide its time at this point. They're really only competing with themselves at the high end.
You know what, though, ARM might catch up and release something at 3ghz relatively quickly.
It won't suspend forward progress but things might be different if AMD was more competitive. But I don't think we should draw parallels between decreasing improvements each generation and lack of competition.
Take video cards for example. Arguably, nV and AMD are equal in regards to performance, yet the 6900 series didn't improve on things as much as say, the HD5800 series over the HD4800, the X800XT over the 9800PRO, or even the Radeon 9700 over the Radeon 8000 series.
Lithography advancements are slowing down. I see a trend of improvements moving to a more linear model as opposed to the exponential Moore's law.
Not quite what I was saying, though, is it? I didn't mention "price-gouging", so what purpose do those "quotes" serve?
Using your own example of AMD top end Phenom II quads:
04 / 2009 - Introduction of AMD's Phenom II X4 955 Black Edition, 3.2GHz - 245$
08 / 2009 - Phenom II X4 965 Black Edition, 3.4GHz - Which assumes the 245$ price point
09 / 2010 - Phenom II X4 970 BE, 3.5GHz at $185
01 / 2011 - Phenom II X4 975 BE, 3.6GHz @ 195$
05 / 2011 - Phenom II X4 980 BE, 3.7GHz keeps the $195 price, but increases the clock
The only time AMD charged more for one of their incremental upgrades in two years was with the 975. Otherwise, they would usually replace their top SKU with something slightly faster, every 4-5 months or so, and keep the price the same. (the exception being the 970 which came a full year after the 965, but did drop the price 60$)
So Intel introducing a new part with only the very slightest of speed bumps after nine months, while raising the price point of their top SKU, feels like stagnation. If not downright regression.
All in keeping with the "Welcome to the Dark Ages" theme of the thread.
Particularly if that application you use shows improvements of 10% to 20% with SSE4.1/4.2 where you will see 15% to 30% with XOP
XOP just owns SSE4.1 and 4.2 which it was meant to replace
You either had to put up with hot and slow Intel CPUs, or very expensive Athlon X2 CPUs. $220 2500k that will be fast enough for just about anything until 2013-2014, and have very high overclocking headroom, is not dark ages.
Not quite what I was saying, though, is it? I didn't mention "price-gouging", so what purpose do those "quotes" serve?
Using your own example of AMD top end Phenom II quads:
04 / 2009 - Introduction of AMD's Phenom II X4 955 Black Edition, 3.2GHz - 245$
08 / 2009 - Phenom II X4 965 Black Edition, 3.4GHz - Which assumes the 245$ price point
09 / 2010 - Phenom II X4 970 BE, 3.5GHz at $185
01 / 2011 - Phenom II X4 975 BE, 3.6GHz @ 195$
05 / 2011 - Phenom II X4 980 BE, 3.7GHz keeps the $195 price, but increases the clock
The only time AMD charged more for one of their incremental upgrades in two years was with the 975. Otherwise, they would usually replace their top SKU with something slightly faster, every 4-5 months or so, and keep the price the same. (the exception being the 970 which came a full year after the 965, but did drop the price 60$)
So Intel introducing a new part with only the very slightest of speed bumps after nine months, while raising the price point of their top SKU, feels like stagnation. If not downright regression.
All in keeping with the "Welcome to the Dark Ages" theme of the thread.
I think innovation has been stifled for at least 2-3 years.
If we go back to the late 1990s, cpu speeds doubled every couple of years,,,, and then slowed down,,,,, and now technology is inching its way forward.
From the time I bought my first computer around late 1994 or early 1995, until the year 1999, we went from a 75mhz cpu, all the way to a 450mhz pentium iii.
From 2006 to 2011, we have not seen that same advance in speed.
You seem to be missing a few important details.
1). Have you looked at the GCN architecture, or watched Eric Demers keynote at AFDS? Bulldozer with GCN is a pretty potent combination. Bulldozer isn't at the end of it's life, it's a little over a week old.
2). AMD has been clear that their future is the Fusion ecosystem.
3). GCN is just around the corner, and Trinity just behind that.
4). intel isn't even in the same league for GPU develpment as AMD and nvidia. Good luck competing there, they've tried and failed every single time.
5). Bulldozer achieves it's target, competition in the high volume market and a forward looking architecture with CPU/GPU integration in mind.
6). "Enthusiasts" are a dying breed. At least the 'spend money at all costs to compete with a few thousand others in a handful of benchmarks' are. The REAL enthusiast, the ones that build their systems as a hobby or toy to play around with on a budget will always be the big market, and is growing. Mainstream being the biggest of all. The number of people wanting a $1000 desktop CPU are waning fast.