How will the bulldozer be better then the i7?

Red Squirrel

[H]F Junkie
Joined
Nov 29, 2009
Messages
9,211
I've always been an AMD fan, but I find in the past few years, Intel seems to have gone way ahead. My last build was actually intel based as it ended up being cheaper for a better cpu. At the time, the core2quad was mostly mainstream, and I don't recall if AMD even had a quad core processor under a grand yet, at that point.

I was reading on the bulldozer, and to me, it looks like it's about equivalent to the i7, but maybe I'm just missing something. I would love to see AMD get back on the top again.
 
It will most likely be better than i7 since AMD can match the i7 940 with their current X6 chips. Bulldozer may even be faster than the i7 980. Will it be faster than the top end socket 2011 next year? That I do not believe so. One thing for certain AMD will have to improve single core IPC by at minimum 20%.
 
I've always been an AMD fan, but I find in the past few years, Intel seems to have gone way ahead. My last build was actually intel based as it ended up being cheaper for a better cpu. At the time, the core2quad was mostly mainstream, and I don't recall if AMD even had a quad core processor under a grand yet, at that point.

I was reading on the bulldozer, and to me, it looks like it's about equivalent to the i7, but maybe I'm just missing something. I would love to see AMD get back on the top again.


AMD's going a completely different route with the bulldozer architecture from what its currently been doing and completely different from anything intel has done.. but its all guessing at this point until they actually have some real ES models being tested.. until then we can only speculate and guess.. but depending on how well the bulldozer does it could completely change how cpu's are made and we might actually see the true death of hyper threading finally..
 
Yeah I've been following the posts on XS (apparently an AMD employee). But yeah, the Intel guy comes into his thread to start some shit, no surprise. All in all, he obviously can't let on much info, at least that isn't already public info already. They need to close the IPC gap, not just throw more cores at it.
 
since AMD can match the i7 940 with their current X6 chips

they can?

From all i heard they can only beat out the i7 in apps that can use all 6 cores, but for everything else the i7 is still beating out the x6 easily.

i would think AMD's new goodies could be out i7, how old is that platform now... over a year?
 
AMD's going a completely different route with the bulldozer architecture from what its currently been doing and completely different from anything intel has done.. but its all guessing at this point until they actually have some real ES models being tested.. until then we can only speculate and guess.. but depending on how well the bulldozer does it could completely change how cpu's are made and we might actually see the true death of hyper threading finally..

That sounds like a departure from AM3 as well. Where did you hear all that?
 
Definitely a good read in those threads.

Getting extremely anxious about seeing Zambezi in action.
 
if IPC doesn't go up to close to parity with intel I'll be dissapointed.
 
Interesting read did not go through all of it but I will when I have more time. Looking forward to hearing more official news and seeing benchmarks on this, once it's out. So sounds like the big thing is architecture change.
 
but how big of a change and for what aims...

It's AMD's first completely new architecture in years, last being the K7. AMD is betting that future workloads will scale better with more cores/threads than the usual bumps in IPC and such. It looks like it should make the server market interesting, but it's a big unknown on how it will handle the workload of home users since the move to multi-threaded apps is still slow aside from a few specific areas(video encoding and off-line rendering).
 
Man...that was confusing as hell to read. I think I man have learned something about cpu architecture though. :confused:

So it seems that AMD is revolving the BD's around a totally new architecture. I also gather that as far as clock for clock performance Intel may still end up being ahead. The way the one AMD guy put it is that they are focusing on "overall" chip performance rather than single thread speed...he used the illustration of a family car with a "v6" or "v8", saying that you wouldn't focus on the speed of just one cylinder but how all cylinders perform together. I may have misunderstood, that was an eyeful/brain full. :(
 
I think the Bulldozer line will be faster than the 1366 and 1156 i7's. But when the 1155 and 2010 i7's or i9 or whatever come out, theyll regain their crown. With the utter flop of the original Phenom line, AMD is a generation behind. The Phenom II line was a home run but unfortunately the i7's were coming out. Kinda the same thing thats going on with ATI and Nvidia. Yeah Nvidia has finally put out some great cards but the 5000 has been out for nearly a year and the 6000 are right around the corner.

In this business, it seems to be a game of leap frog so any head start you can get is a big help and right now ATI has one in the graphics and Intel has one in the procs.
 
I think the biggest advantage to concentrating on multi core is virtualization. I'm sure an ESX server on such processor would simply rock. Heck, get a board with 4 sockets. *geek boner*
 
they can?

From all i heard they can only beat out the i7 in apps that can use all 6 cores, but for everything else the i7 is still beating out the x6 easily.

i would think AMD's new goodies could be out i7, how old is that platform now... over a year?

I guess I also missed where the AMD chip can outperform the i7980 or the i7 lineup as a whole.
 
i think AMD is wrong trying to think that more cores will be the future. in any case even if IPC is not the key, intel will just throw more cores and since they are more efficient per core it will be a moot point. I see bulldozer flopping really -- I hope I'm wrong though
 
The MOST important things for AMD are

1. Stay alive with sustainable products.

2. x86 Compiler and libraries that can take full advantage of its offering. No need to waste time on whether other compiler should or should not do something. Then choose critical software projects to demonstrate the results. BTW, do not waste time arguing with resistive software writers, Life is too short, find those with good technical capability and human synergy receptive to your idea. If you have a bit of spare funds available, consider sponsoring some software projects.

3. OpenCL and GPGPU. This is the area where AMD is on slightly more positive position due to extended experience in GPU-related coding. nVidia is slightly more advance I think in terms of generic GPGPU development/applications for now but the topic is Bulldozer. VERY LITTLE existing codes to worry, link Bulldozer and Radeon for quick new HYBRID software ecosystem establishment

note: you will hear counter-argument about better in separate things here and there for upgrade and other opinions. Not wrong. my view is Life is too short. You either execute or you don't. Nature accommodates many different things, choose your path, focus the resources and move on.
 
i think AMD is wrong trying to think that more cores will be the future. in any case even if IPC is not the key, intel will just throw more cores and since they are more efficient per core it will be a moot point. I see bulldozer flopping really -- I hope I'm wrong though

To each their own but everyone else in the game believes multi processing is better that mega ultra fast processing. Think Cuda cores and ATI stream Processors, this is where AMD and Intel are going and if you ever owned an SMP machine VS a single core 1p machine you would see the merits in multitasking and overall system stability.

As for me Im looking at the Bulldozer Interlagos! Although there is a 2p board right now that will support it and it has 2 PCIe slots, its only dual channel, Interlagos is going to really shine with quad channel. Just hope motherboard vendors see that some enthusiastic players will look to the Interlagos for desktop chores.
If someone is going to talk smack about prices though, Intel's desktop i7 unlocked quad core is over $1,000 You can get an 8 "real" core Mangy Cores for $280 from Newegg, and a 2p motherboard for $440. So for $1,000 or the price of one Intel CPU, I can get 16 "real" cores. Yes, when Interlagos comes out there will be a price premium for that chip but then I would have 32 bulldozer cores and nothing about Sandybridge gives me wood like this can.
 
If someone is going to talk smack about prices though, Intel's desktop i7 unlocked quad core is over $1,000 You can get an 8 "real" core Mangy Cores for $280 from Newegg, and a 2p motherboard for $440. So for $1,000 or the price of one Intel CPU, I can get 16 "real" cores.

You are not gaining anything (except for an increased power bill) if those 16 real cores each are around 1/3 as powerful as a single real core on the 6 core 12 threaded i7 980.

Note that all of the Magny-Cours chips have their single core IPC being around 1/3 to 1/2 the IPC of a single 3.33GHz i7 core.
 
Last edited:
if any of you actually read through the thread at xs, it looks like bulldozer will do fairly well, 16core chip in the same power space as their current 6core, with a higher IPC per core? I don't know why anyone would complain :D
 
@ Griff30 I've owned smp rigs but I'm all about efficiency. Intel gets more done per core and that I like. I would want 12 intel cores vs 16 amd cores but that could change if AMD does this bulldozer launch right.
 
To each their own but everyone else in the game believes multi processing is better that mega ultra fast processing. Think Cuda cores and ATI stream Processors, this is where AMD and Intel are going and if you ever owned an SMP machine VS a single core 1p machine you would see the merits in multitasking and overall system stability.

As for me Im looking at the Bulldozer Interlagos! Although there is a 2p board right now that will support it and it has 2 PCIe slots, its only dual channel, Interlagos is going to really shine with quad channel. Just hope motherboard vendors see that some enthusiastic players will look to the Interlagos for desktop chores.
If someone is going to talk smack about prices though, Intel's desktop i7 unlocked quad core is over $1,000 You can get an 8 "real" core Mangy Cores for $280 from Newegg, and a 2p motherboard for $440. So for $1,000 or the price of one Intel CPU, I can get 16 "real" cores. Yes, when Interlagos comes out there will be a price premium for that chip but then I would have 32 bulldozer cores and nothing about Sandybridge gives me wood like this can.

Intel chips don't need to be unlocked to oc them well into 4 Ghz range. And as far as zomg i have 12 cores in cpu factor goes those cores are something like 2,2 Ghz so even if's most uber multi threaded code it won't be that much faster than X6 1090T running on desktop with 4-4,2 Ghz (and if you find code which cannot use more than 2 or 4 cores you are in deep shit with those 2 Ghz opteron cores).

Personally if intel brings IPC improvement at the levels they promise and i3 can oc to 4,5 like current generation I'm dumping this x6 and going with 2 ultra fast cores with HT.
 
if any of you actually read through the thread at xs, it looks like bulldozer will do fairly well, 16core chip in the same power space as their current 6core, with a higher IPC per core? I don't know why anyone would complain :D

I will have to look more carefully. I didn't see much about how many modules, what frequency or powerdraw. I do agree that the IPC will be improved. My speculation is an 8 core / 4 module bulldozer at 3.3 GHz would be as powerful as the i7 980 at the same frequency when using 8 threads.
 
Last edited:
How was K8 not a completely new architecture?

The move from K7 to K8 was a large jump, but the underlying architecture and pipeline design was very similar. The biggest change was moving the IMC/Northbridge onto the CPU die. I guess you could compare the move from K7 to K8 to Intel's move from Core 2 to Nehalem .Both are called new architectures, but they have strikingly similar foundation but had some noticeable differences. Bulldozer on the other hand is a move to a completely new way of building a chip, from either company.
 
I hope it beats up on Sandybridge like the Athlon beat up on the P3.
Why?
It's AMDs turn and because it drives down the cost of Intel CPUs.
If the IPC of AMDs goes like I think it will, they will dominate Intel this time around. Im just wanting desent 2P Interlagos motherboards with quad channel memory and Crossfire.
 
1. In order to survive, AMD must have volume shipment.
2. Volume shipment means must sell a lot to commoners, in addition to rocket scientists or engineers
3. Commoners mean they are NOT going to "optimize their code/software" just to suit AMD system. Most of them just want to use the applications or do something using the computer.
4. If commoners are not going to "optimize the software", then the software must mostly come already optimized with AMD in mind to sustain any impressive improvement claim.
5. All superlative proclamation means nothing if the software involved are NOT optimized for it. Take a simple example, the compiler "optimized path" and the standard "legacy path", whatever that means.

6. With both camps progressing roughly on the same enhancement rate with half to one generation gap, any superlative claims will be matched or offset by generic software environment issues.

7. The only critical space will be, yep, scientist/engineers types where they are willing to "optimize their software" to show a difference.

8. And Linux Server environments, where coders are progressing at tremendous pace to absolve any incoming new-techs. However, small catch, mostly relevant on the server-related portion. For client applications, need to assess further.

So the questions are, will Bulldozer EXCEL in an non-optimized environment (as claim by many in the compiler debate" as it is ? Will software writers "Optimize their software" with AMD in mind when Bulldozer arrives? How easy AMD or development software solutions make it possible for software writers to perform the necessary optimization?

Without such efforts, you may find you are perpetually working in the shadow of compiler's "legacy path", you may make the "legacy path" flies, it does mean software writers will then disregard whatever new features, by the time the new features are really needed for performance reason, the alternative-side new code-paths are already firmly ingrained in the ecosystem. And you are back either "as claim"-try to follow exactly like what the standard does to benefit from optimization"--(difficult to do as you will always one generation down to chase), or else to brute-force core-counts to wear the alternative down, but then this strategy requires superior multiprocessing/multithreading capabilities, which goes back to IT software current problems of how to teach the generic software to make effective use of more than 4-core, which then comes back to OpenCL

Just an opinion, hopefully not treated as argument. Cheers
 
Last edited:
they can?

From all i heard they can only beat out the i7 in apps that can use all 6 cores, but for everything else the i7 is still beating out the x6 easily.

From what I understand, the i7 simply slaughters the x6 in consumer applications like photoshop or gaming. It's only when you approach it from an enterprise perspective that it can match or exceed the i7. This is all well and good, but I don't really care about widescale number-crunching - all I want is a CPU under $150 that can play games and transconvert music files very, very quickly.
 
I think current i7 will still have a slightly higher IPC than buldozer. There is too much ground to make up.

+1 to this. Especially with sandy bridge and other newer chips being released.
 
What I got from both threads was that BD should beat or equal Nehalem on single thread performance but will likely lose to Sandy Bridge for single threaded apps. For multi threaded apps BD may very well meet or beat Sandy Bridge while having similar per core TDP/die size. Everyone who actually seemed to know anything was being very cagey though, lots of reading in between the lines... have to wait and see.

Prolly won't know for sure until Q1 11' I bet.
 
That would be really nice.

But still some confirmation that it will work on am3 mobos would be nice to have. Total silence on this part from amd suggest it might be not and they don't want to scare buyers who at the moment think am3 is more future proof than 1156
 
Back
Top