AMD "Bulldozer" Interactive Series Is Live

Well it's certainly nice that they included a bulldozer in the video so there's no need for anyone to post one in this thread.

Right...?
 
I haven't found it yet, but I'm pretty sure there's got to be an AMD version of the turbo encabulator video in there somewhere.
 
srangara said:
Well it's certainly nice that they included a bulldozer in the video so there's no need for anyone to post one in this thread.

Right...?
Click to expand...


LOL, except that's not a bulldozer, but a wheeled loader. Now this is a bulldozer: :p

IDF%20Bulldozer.jpg
 
spacing guild said:
AMD really needs to start phasing out pins on CPUs...looking forward to Bulldozer.
Click to expand...
Little good would come from AMD changing sockets any more than absolutely necessary. Hard to call sticking with PGA packages wrong though. Even Intel still uses PGA for socketed mobile CPUs.
 
canna said:
So if I'm calculating that right, it looks like Bulldozer is about a 13% performance gain clock for clock?
Click to expand...
Depends on the clock speeds of each one. ;)

If both were run at the same clock speed (and no turbo), that would be a 12.5% performance gain in IPC in whatever that benchmark happens to be. But the clock speeds are pretty important for an IPC calculation.
 
Here's to hoping that it will be a good cpu, competition is always great.
 
There is a lot resting on the desktop performance. I hope they get the IPC up to Intel's numbers or just call me a Intel owner this Christmas.
 
griff30 said:
There is a lot resting on the desktop performance. I hope they get the IPC up to Intel's numbers or just call me a Intel owner this Christmas.
Click to expand...
Rumors (leaked slides with comparisons and prices) point to them being compared to i7-2600 level performance with their highest end desktop chip. That is pretty respectable considering how much ground they have to make up. However, keep in mind that 2600 is still only upper mid-range of the i7's Sandy Bridge CPU's, they have yet to release the high end LGA 2011 chips.

Of course, it is all rumor and speculation at this point. Until I see actual hard numbers, I'm taking this with a grain of salt. Bulldozer could be amazing or it could be another "Phenom" -enal failure. My bet is that it is somewhere in-between, a really decent bang for buck desktop cpu that falls short of Intel's high end, but it's unique architecture owns everything on the server side until Intel's next set of major releases.
 
griff30 said:
There is a lot resting on the desktop performance. I hope they get the IPC up to Intel's numbers or just call me a Intel owner this Christmas.
Click to expand...
If that "50% faster" is a best case (and likely includes a clock speed bump on Interlagos over Magny Cours), it's pretty unlikely single threaded performance is anywhere near SB.

For servers, Interlagos is fine. More fast cores in many cases is a great way to scale (software licensed by the core OTOH has disadvantages with that). But I just don't see how throwing a bunch of cores at the types of desktop software and games most people use solves any problems for AMD on the desktop. There are no BD mobile chips until 2012, which is when Fusion gets a new CPU core.
 
canna said:
http://www.hardocp.com/news/2011/04/06/amd_bulldozer_interactive_series_live

^at the end of the video, click performance.

He says a 16 core BD will be 50% faster than a 12 core Operton.

So if I'm calculating that right, it looks like Bulldozer is about a 13% performance gain clock for clock?
Click to expand...

Yeah.. 12.5% to be exact...


That has me concerned. He doesn't say that it is clock for clock, but lets assume for a moment that it is.

If this translates into similar speeds for dekstop models this means that Bulldozers core for core, clock for clock performance will only be 12.5% faster than K10 Phenom II's

This would mean that bulldozer on the desktop will not even catch up with Lynnfield Core i5's and i7's on the IPC front... And there is no way they'll be able to make up for that in clock speed....

Looks dire to me.

That being said, for games and 3D, this will be more than enough to keep up with a quad SLI GTX580 either way...
 
Zarathustra[H];1037082130 said:
This would mean that bulldozer on the desktop will not even catch up with Lynnfield Core i5's and i7's on the IPC front... And there is no way they'll be able to make up for that in clock speed....
Click to expand...

I probably don't need to say that this means AMD more or less completely is down and out as far as the enthusiast consumer market goes...

More cores simply don't scale well on most desktop applications. You need high clocks AND high IPC to complete in this market.
 
I think they know that but damn if they ever do anything about it in this century.
 
Is it just me, or was he in the last 10sec of the video, bobbing his head like: "Yeah bitches....."" :)
 
Zarathustra[H];1037082142 said:
I probably don't need to say that this means AMD more or less completely is down and out as far as the enthusiast consumer market goes...

More cores simply don't scale well on most desktop applications. You need high clocks AND high IPC to complete in this market.
Click to expand...

But the only way to produce performance gains is more cores. The ceiling ghz wise is limited to the manufacturing process.

Badly written software (like last bunch of console -> PC ports) will always cripple performance ....
 
griff30 said:
There is a lot resting on the desktop performance. I hope they get the IPC up to Intel's numbers or just call me a Intel owner this Christmas.
Click to expand...

supposedly they have. and on par with intel if the rumors hold true. Of course this still boils down to being a half cycle BEHIND Intel rather then leap frogging them. and that is architecture not fabrication. bulldozer should be in a class of its own in a few areas but I think we are going to have be happy with "competitive" in the desktop area.
 
Zarathustra[H];1037082142 said:
I probably don't need to say that this means AMD more or less completely is down and out as far as the enthusiast consumer market goes...

More cores simply don't scale well on most desktop applications. You need high clocks AND high IPC to complete in this market.
Click to expand...

Why do you think JF-AMD has been running around screaming that single-threaded performance no longer matters and multi-threaded is where it's at now? I think he's basically admitting that IPC hasn't really increased much :/
 
Pieter3dnow said:
But the only way to produce performance gains is more cores. The ceiling ghz wise is limited to the manufacturing process.

Badly written software (like last bunch of console -> PC ports) will always cripple performance ....
Click to expand...

...I guess that's why AMD's top end Phenom II X6's are getting their asses handed to them, in some cases even by dual core i3's...
 
Logi_c@1 said:
Why do you think JF-AMD has been running around screaming that single-threaded performance no longer matters and multi-threaded is where it's at now? I think he's basically admitting that IPC hasn't really increased much :/
Click to expand...

JF-AMD is correct in his assertions....


...for the server platform.


For desktop use, in the vast majority of cases you stop seeing significant scaling benefits over 2 cores, and really stop seeing any scaling benefits over 4 cores...

If Bulldozer is unable to challenge Intel's last gen (pre-Sandy Bridge) processors core for core, clock for clock, I fear they are completely out of the enthusiast desktop market.

This is very sad. I was really hoping for a major AMD comeback. I was a huge fan in the K7-K8 years when P4's were getting their asses handed to them...
 
Logi_c@1 said:
Why do you think JF-AMD has been running around screaming that single-threaded performance no longer matters and multi-threaded is where it's at now? I think he's basically admitting that IPC hasn't really increased much :/
Click to expand...

I don't know if that is true or not (I read that they have upped it quite a bit) but it WILL be true of both intel and amd. there is a limit to the amount of power per area^2 that is feasible. or in other terms the amount of circuitry necessary to increase the IPC becomes impractical. even with die shrinks. intel is currently ahead but at some point it isn't going to matter as they will both hit that limit sooner then latter. Intel is banking on multi-threading just as much as AMD is for the future. you just don't hear it as much from Intel as you do AMD due to Intel running at a considerable advantage currently. They even have some projections on this (I couldn't follow the math though :eek:).
 
Valset said:
I don't know if that is true or not (I read that they have upped it quite a bit) but it WILL be true of both intel and amd. there is a limit to the amount of power per area^2 that is feasible. or in other terms the amount of circuitry necessary to increase the IPC becomes impractical. even with die shrinks. intel is currently ahead but at some point it isn't going to matter as they will both hit that limit sooner then latter. Intel is banking on multi-threading just as much as AMD is for the future. you just don't hear it as much from Intel as you do AMD due to Intel running at a considerable advantage currently. They even have some projections on this (I couldn't follow the math though :eek:).
Click to expand...

Undoubtedly this is true. The parallels are similar to the Pentium Pro. In the WIndows 95 days it was a huge disappointment for desktop users and cost a good bunch more than a regular Pentium, because at that time everyone was running mostly 16bit code. The Pro excelled at 32bit code. The genius of the Pentium Pro wasn't really realized by desktop users until much later on, when it was mostly obsolete anyway, when it was apparently how much better it ran modern code than the old Pentiums.

Same think could be said for AMD's Athlon 64's when they were first launched. 64 bit code was pointless on the desktop then. Now, looking back, we see how useful it is, especially if you need more RAM, but at the time, it was kind of a pointless feature.

I have no doubt that at some point software programmers will be forced into writing good multithreaded code that scales well on multiple cores. The problem is that it is challenging to write good multithreaded code. it costs more to do something challenging than something easy, so in the near future, only tsoftware that absolutely needs to will be written to do this well. (and the oddball small game developer that likes giving themselves a challenge, like Firaxis with Civ 5.) Because of this. I see it as being at least 5 years out on the desktop platform.

The problem with this is that we don't buy CPU's because of how well they will perform in 5 years. We buy them for how well they will perform now, and now dual core (and to a lesser extent, quad core) performance is where the multi-core setups tend to peak out in scaling. 5 years from now when we will have more intelligent desktop software that scales well on an 8 core Bulldozer most of us will already have upgraded to our next CPU and platform, making it pointless today.
 
Zarathustra[H];1037085625 said:
Undoubtedly this is true. The parallels are similar to the Pentium Pro. In the WIndows 95 days it was a huge disappointment for desktop users and cost a good bunch more than a regular Pentium, because at that time everyone was running mostly 16bit code. The Pro excelled at 32bit code. The genius of the Pentium Pro wasn't really realized by desktop users until much later on, when it was mostly obsolete anyway, when it was apparently how much better it ran modern code than the old Pentiums.

Same think could be said for AMD's Athlon 64's when they were first launched. 64 bit code was pointless on the desktop then. Now, looking back, we see how useful it is, especially if you need more RAM, but at the time, it was kind of a pointless feature.

I have no doubt that at some point software programmers will be forced into writing good multithreaded code that scales well on multiple cores. The problem is that it is challenging to write good multithreaded code. it costs more to do something challenging than something easy, so in the near future, only tsoftware that absolutely needs to will be written to do this well. (and the oddball small game developer that likes giving themselves a challenge, like Firaxis with Civ 5.) Because of this. I see it as being at least 5 years out on the desktop platform.

The problem with this is that we don't buy CPU's because of how well they will perform in 5 years. We buy them for how well they will perform now, and now dual core (and to a lesser extent, quad core) performance is where the multi-core setups tend to peak out in scaling. 5 years from now when we will have more intelligent desktop software that scales well on an 8 core Bulldozer most of us will already have upgraded to our next CPU and platform, making it pointless today.
Click to expand...

I really ought to proof-read for typos before I hit "post"...
 
I think everyone's concern here is that AMD might be putting 8 weak cores on Bulldozer, as opposed to maybe 4 strong cores like i7-2600K. Nobody wants a many core Atom, it will still suck in the end. Just an analogy :p

The problem here, I think, is that while yes, programming paradigms will change and writing multi-threaded code becomes more important, not everything can be parallelized. Some things are inherently serial, and that's where IPC and clock speed comes in, I guess.

I'm also not sure what the current trend of PC games being either multi-platform or console ports will mean with regards to parallelism. Will a console port be inherently single-threaded? If so, we might not see much advantage with regards to gaming with Bulldozer.

Still, the most intensive usage of my PC is video encoding with x264, which is very nicely multi-threaded. So I still look forward to Bulldozer :)
 
Logi_c@1 said:
I'm also not sure what the current trend of PC games being either multi-platform or console ports will mean with regards to parallelism. Will a console port be inherently single-threaded? If so, we might not see much advantage with regards to gaming with Bulldozer.
Click to expand...

In an ironic twist, Console ports may actually become better at taking advantage of parallelism, considering the PS3's heavily parallelized Cell architecture. Also considering the Xbox 360 has (compared to the PC) a pretty weak 3 core design, code here would also likely be optimized for parallelism.

The question is whether or not these multi-threaded optimizations will survive the porting process or not.

A current problem here is that neither AMD nor Nvidia support the necessary functions for multithreading per the DX11 spec in their drivers. I believe the key missing feature currently is "Driver Command Lists".
 
Zarathustra[H];1037085728 said:
A current problem here is that neither AMD nor Nvidia support the necessary functions for multithreading per the DX11 spec in their drivers. I believe the key missing feature currently is "Driver Command Lists".
Click to expand...

Actually, I take that back...

The latest Nvidia 270 beta drivers apparently support this. I wonder how much this impacts performance in Civ 5, seeing that it is a fully multithreaded game, other than the DX11 driver problems...

My guess is that the benefit will be most noticeable on systems with multiple cores with low IPC and clock speed, but with a high end video card. In other words, it would ahve been CPU bottlenecked before, but with this feature enabled, the rendering will be spread more evenly over the CPU cores.
 
You must log in or register to reply here.
Back
Top