Is Ivy Bridge the answer to AMD Bulldozer ?

[996GT2]

We really need some BD performance numbers already to stop all this speculation! At least if we saw a preview from AT we could extrapolate based on some actual data...

AMD wants to prevent all the enthusiasts from jumping ship to Intel...hence no numbers for as long as they can delay

 

[IntelUser2000]

Still not coded very well... most people would be angry if they saw their GPU at only 50% load, and getting bad FPS.

I don't think its fair to call it "poorly coded" based on CPU utilization numbers.

Games should be coded for optimal performance, if that somehow results in 10% CPU utilization, so be it.

 

[Skott]

I'm certainly no expert on all these things about CPUs and how they work and what makes them work but I think once BD and IL are both out the landscape won't change much from how it is now. AMD will have a 8 core but Intel's chips will be faster and more powerful. Intel will claim the high end dominance and AMD will claim a cheaper priced product on the lower end. Pretty much how it is now.

 

[podspi]

It's true provided you are not multi-tasking a bunch of single-threaded apps.

But then having 8-cores (higher throughput) will be beneficial anyway

 

[Phynaz]

The thing that gets me is that it was maybe 2 months ago that the prevailing belief on these forums was that Bulldozer would be inferior to Sandybridge (and probably Nehalem as well), and now it seems that the prevailing belief is that it will be so much better than Sandybridge that Intel will need Ivy bridge to compete with it.

Based on the known architecture improvements, the IPC should improve to the point where it is competitive with current Intel processors, but to be honest we don't know if those improvements will indeed show up in the performance of the processor. We don't have a single benchmark to show the BD's processing power, so I wouldn't put any real effort into figureing out what future processor will be able to outperform it, since we don't even know if current processors are going to outperform it yet.

What can be looked at is comments from AMD.

Look at another thread where JF is trying to downplay top performance by saying things like there are a wide range of niche areas for AMD to place their cpus.

There are also the AMD financial conference calls, where company executives said being the top performer is not a goal of Bulldozer. In their opinion the ship has sailed on having top performance at any cost, and Bulldozer was designed around performance per watt and die size.

People especially seem to ignore the latter.

 

[tweakboy]

Again, Sandy Bridge on socket 2011 is slated to have 6 and even 8 core parts while the first Ivy Bridges will only be for 1155 and likely only 4 core.

And considering the oft rumored high cost of an LGA2011 system, my guess is Bulldozer fits somewhere in between.

So what late 2011 for 8 core or 12 or whatever.. hmmm?

 

[JFAMD]

What can be looked at is comments from AMD.

Look at another thread where JF is trying to downplay top performance by saying things like there are a wide range of niche areas for AMD to place their cpus.

There are also the AMD financial conference calls, where company executives said being the top performer is not a goal of Bulldozer. In their opinion the ship has sailed on having top performance at any cost, and Bulldozer was designed around performance per watt and die size.

People especially seem to ignore the latter.

If you are going to put words in my mouth, please choose the right words.

The "niche" in the market is the $1000 SKU that nobody buys. We will do quite well in the 99.9% of the rest of the market. Hell, we might even do well in the top spot, I don't know, I am a server guy.

But to imply that we are targeting niches is not in sync with what I said. We will do very well in the areas where people are actually spending money. Which is hardly a niche.

 

[StinkyPinky]

If AMD had a winner on their hands they would be leaking some numbers by now to stop people getting SB cpu's. Instead they're super quiet and delay delay delay.

I think a lot of people are going to be disappointed. Hope I'm wrong.

 

[996GT2]

If AMD had a winner on their hands they would be leaking some numbers by now to stop people getting SB cpu's. Instead they're super quiet and delay delay delay.

I think a lot of people are going to be disappointed. Hope I'm wrong.

I don't think you're wrong. I think that BD to SB will probably turn out to something like Phenom II vs. Core 2.

 

[Skurge]

If AMD had a winner on their hands they would be leaking some numbers by now to stop people getting SB cpu's. Instead they're super quiet and delay delay delay.

I think a lot of people are going to be disappointed. Hope I'm wrong.

I think AMD just decided that its better to keep quiet, you never know what could happen, something like the TLB bug could strike again and the CPU might end up slower than intended.

They keep quiet on the GPU side too, even if they have a winner.

 

[Makaveli]

I hope bulldozer is a success but when I look at things on paper.

AMD's new cpu needs alteast a 30% increase in IP just to equal SB, would be 15-20% to equal Nelahem. We are getting closer and closer to the point where you cannot just throw in extra cores to win the battle.

I'm so looking forward to BD vs SB-E but my gut tells me the high performance crown is gonna stay with the blue team.

 

[Idontcare]

If AMD had a winner on their hands they would be leaking some numbers by now to stop people getting SB cpu's. Instead they're super quiet and delay delay delay.

I think a lot of people are going to be disappointed. Hope I'm wrong.

If you look at what happened in those months before Phenom was released you see that AMD had a real problem in that they needed their stock price to stay up as well as their credit rating so they could fund the ATI purchase. They had every reason to hype, funding for the ATI purchase was depending on it. AMD isn't in that position now.

 

[nonameo]

If you look at what happened in those months before Phenom was released you see that AMD had a real problem in that they needed their stock price to stay up as well as their credit rating so they could fund the ATI purchase. They had every reason to hype, funding for the ATI purchase was depending on it. AMD isn't in that position now.

O_O

wow, that actually makes a lot of sense. Thanks for the insight.

 

[mosox]

For 99% of the users out there the Sandy Bridge is already overkill. Yet both AMD and Intel scramble to make them even more powerful. Intel refuses to make cheap quads and it looks like AMD will do the same.

To spend $200+ on a CPU (make that $250-300 in countries with VAT and higher prices) and to see it staying in the task manager at 4-20% during your usual business is not nice. A quad is good to have for some games and applications but it doesn't have to cost an arm and a leg.

The wasted power of all the CPUs in the world could power a small galaxy. It's like everybody driving 16 wheelers to the shop to get groceries.

Will AMD offer $100 quads on the new socket?

 

[RobertPters77]

If you are going to put words in my mouth, please choose the right words.

The "niche" in the market is the $1000 SKU that nobody buys. We will do quite well in the 99.9% of the rest of the market. Hell, we might even do well in the top spot, I don't know, I am a server guy.

But to imply that we are targeting niches is not in sync with what I said. We will do very well in the areas where people are actually spending money. Which is hardly a niche.

He's right john. Amd is avoiding the issue and dodging questions. I used to buy in the A64 days Ati gpu's with Nvidia chipsets because in DX9 ATI was king. That was the combo every gamer needed/wanted. When the first gen phenoms were released every gamer who didn't jump the fence to intel/nvidia now had no reason not too. Amd lied. Now why are they keeping mum about Bulldozer? Is it because they plan to wow us with a CPU 10-20% faster than Sandy Bridge? Or will it be a dissapointment like the first gen phenoms? I don't need more cores. I have 4, I barely use 2. Sure Average Joe Consumer who walks into bestbuy might get suckered into buying an Octo-Core Bulldozer because the blueshirt will say it's fast.

Whatever markets your company is doing well in doesn't mean it's making a profit to stay afloat without lawsuit settlements and government assistance to keep it's head above water. Look you probably think me of as some anti-amd-asshole, but that's not the case. I wan't amd to succeed because I do want to see some competition back in the cpu market. Phenom 2 and Core 2 were very close amongst each other other(less than 5% on some apps) but Phenom 2 was 3 years too late.

For 99% of the users out there the Sandy Bridge is already overkill. Yet both AMD and Intel scramble to make them even more powerful. Intel refuses to make cheap quads and it looks like AMD will do the same.

Will AMD offer $100 quads on the new socket?

My cousin maxes out his i3-2120 playing Games.

The X4 640 is forwards compatible with AM3+. But I don't know how long they'll keep making those when BD is released.

 

[greenhawk]

We are getting closer and closer to the point where you cannot just throw in extra cores to win the battle.

True, but the point of a little faster clock speed to be worth a new computer is well past.

Not many ways to improve going forward once you get to a reasonable amount of cores.

(power and MMX/SSE expantion is all that come to mind short of the System on a chip or even 1GB of cache)

I'm so looking forward to BD vs SB-E but my gut tells me the high performance crown is gonna stay with the blue team.

BD vs SB-E is not going to happen as a viable comparison personally. BD vs SB might, but then BD vs IB is also proberly not a comparison worth doing.

note: looking at performance

 

[SickBeast]

From the new leaks at xbit, I find it disappointing to see that it takes an 8 core bulldozer to compete with an i7. I know it's a leak and it might not be legitimate, but really it would be nice if AMD could compete with intel core-for-core and clock-for-clock. It's really hard to see the need for 8 cores for most people, unless they do a ton of rendering or video editing.

 

[Makaveli]

True, but the point of a little faster clock speed to be worth a new computer is well past.

Not many ways to improve going forward once you get to a reasonable amount of cores.

(power and MMX/SSE expantion is all that come to mind short of the System on a chip or even 1GB of cache)



BD vs SB-E is not going to happen as a viable comparison personally. BD vs SB might, but then BD vs IB is also proberly not a comparison worth doing.

note: looking at performance

You have a good point there about BD vs Regular SB.

 

[piesquared]

From the new leaks at xbit, I find it disappointing to see that it takes an 8 core bulldozer to compete with an i7. I know it's a leak and it might not be legitimate, but really it would be nice if AMD could compete with intel core-for-core and clock-for-clock. It's really hard to see the need for 8 cores for most people, unless they do a ton of rendering or video editing.

Do you find it dissapointing that Radeon is using at least 2x as many cores than GeForce?

 

[SickBeast]

Do you find it dissapointing that Radeon is using at least 2x as many cores than GeForce?

With a GPU it doesn't really matter how many cores you use; games are always able to take advantage of as many cores as you give them. With CPUs, you can't always take advantage of an endless number of cores.

 

[RobertPters77]

Do you find it dissapointing that Radeon is using at least 2x as many cores than GeForce?

Don't you mean 4-5x as much?

And yes, yes I do.

 

[Blitzvogel]

Don't you mean 4-5x as much?

And yes, yes I do.

They are not directly comparable. AMD's performance/transistor, watt, and die area is currently much superior to Nvidia's. It's especially ridiculous when the full Fermi GF100 die has 3.2 Billion transistors at 529 mm^2, and runs the shaders twice the speed as the rest of the chip. Over on the AMD side of things we have Cypress XT with about 2.2 Billion transistors on a 334 mm^2, running a single clock domain over the entire chip that is running much slower than Nvidia's shaders need to be at.

While generally the GTX 480 is considered faster, not to mention 4x the tessellation capability, larger memory bus, and CUDA capability which adds to the transistor count most surely, Cypress holds a good position near it. With some pieces of software, it equals Fermi with 2/3 the transistors, much lower power consumption, and smaller die area which is good for heat output.

I would liken AMD's graphics as to like many Japanese cars. Highly efficient, well engineered for their intended purpose. Nvidia's graphics are more like an American car. Heavy, and large-muscled with a very high theoretical capability through sheer brute force, but at the cost of fuel and efficiency.

 

[itsmydamnation]

Don't you mean 4-5x as much?

And yes, yes I do.

so you find all VLIW based designs dissapointing? no offence but if after your thread on this issue you dont understand the differences and how it really doesn't matter, then you have some comprehension issues.


edit:
CUDA adds very little to die size, read the beyond3d review on the G100 uarch.

 

[piesquared]

With a GPU it doesn't really matter how many cores you use; games are always able to take advantage of as many cores as you give them. With CPUs, you can't always take advantage of an endless number of cores.

Or put another way, a lower core count isnt't always an advantage over higher core count. The only reason one would argue against more cores is if one doesn't have more cores. And the point is completely moot if 8 cores do in fact out perform 4 cores, no? They are different architectures and different design choices just like Radeon and GeForce.

 

[Phynaz]

If you are going to put words in my mouth, please choose the right words.

The "niche" in the market is the $1000 SKU that nobody buys. We will do quite well in the 99.9% of the rest of the market. Hell, we might even do well in the top spot, I don't know, I am a server guy.

But to imply that we are targeting niches is not in sync with what I said. We will do very well in the areas where people are actually spending money. Which is hardly a niche.

I'm using your words. Finding someplace where Intel doesn't have a presence is a niche.

The goal is to find parts of the market where they aren't that you can own.... Trying to take on a bigger competitor by matching their products is not a great strategy

As far as doing well in areas where people are spending money, what's your enterprise market share today?