Intel & AMD, Architectural Discussion, How Far Ahead Is Intel ?

[TechFan@ic]

You're missing the whole point. Intel ain't ahead, it's just AMD lagging behind.

 

They thought of modules? wow, great. They are slower than Intel counterparts... Why not just make Phenom III? I'll tell you why. Save die space for APUs.

 

Except they are too confident about APU performance - APUs will never catch up to regular dedicated GPUs. And as of today APUs are useless and overpriced (price to performance is really really bad, 50 bucks more you got a dedicated GPU + 4 core CPU).

 

And to make it clear, you're kinda talking about 8 core 4 module processors like they are something new... Intel can throw more cores at the problem at any time, they got processors with lots of cores for years now... They just dont feel like putting them in regular consumer price range.

 

Look up Intel Xeons, look up Intel Phi.

I think you're missing the point, an intel core is twice as large as an AMD core, so of course it's going to be faster.

However what turned out to be very interesting is that although an intel core is twice as large as an AMD core it is not twice as powerful not even close to that.

APUs are not about a dedicated graphics card replacement, not at all, AMD's goal is to use the onboard GPGPU cores for processing, they will replace floating point units, that's why AMD did not bother to improve floating point performance.

A Carzio APU (Excavator based) will have extremely powerful integer processing capabilities per die area as compared to intel and will have extremely powerful floating point processing performance per die area as compared to intel, that is the ultimate goal.

Intel will still have a very big advantage though through their manufacturing process, they'll be at 14nm while AMD will just have reached 20nm.

[Dravic]

I think you're missing the point, an intel core is twice as large as an AMD core, so of course it's going to be faster.

However what turned out to be very interesting is that although an intel core is twice as large as an AMD core it is not twice as powerful not even close to that.

APUs are not about a dedicated graphics card replacement, not at all, AMD's goal is to use the onboard GPGPU cores for processing, they will replace floating point units, that's why AMD did not bother to improve floating point performance.

A Carzio APU (Excavator based) will have extremely powerful integer processing capabilities per die area as compared to intel and will have extremely powerful floating point processing performance per die area as compared to intel, that is the ultimate goal.

Intel will still have a very big advantage though through their manufacturing process, they'll be at 14nm while AMD will just have reached 20nm.

 

Speculations much. Yes and meanwhile AMD is due 2 years with their Kavieri APUs... I'll believe it when I see it.

[TechFan@ic]

Speculations much. Yes and meanwhile AMD is due 2 years with their Kavieri APUs... I'll believe it when I see it.

Nothing I've stated is speculation, it's all official from AMD.

Compute performance of hetergenous architectures (CPU+GPU)

Progression of single-core performance, compared to multi-core compared to parallel GPGPU designs.

AMD's goals for the APU.

[Dravic]

Nothing I've stated is speculation, it's all official from AMD.

Compute performance of hetergenous architectures (CPU+GPU)

Progression of single-core performance, compared to multi-core compared to parallel GPGPU designs.

AMD's goals for the APU.

 

All i see:

 

 

Mind you im not an intel fanboy but when AMD has nothing good at hand they are going for aggresive marketing, for example:

 

http://www.youtube.com/watch?feature=player_embedded&v=eH6XayaLTw8

[Glenwing]

You're missing the whole point. Intel ain't ahead, it's just AMD lagging behind.

 

They thought of modules? wow, great. They are slower than Intel counterparts... Why not just make Phenom III? I'll tell you why. Save die space for APUs.

 

Except they are too confident about APU performance - APUs will never catch up to regular dedicated GPUs. And as of today APUs are useless and overpriced (price to performance is really really bad, 50 bucks more you got a dedicated GPU + 4 core CPU).

 

And to make it clear, you're kinda talking about 8 core 4 module processors like they are something new... Intel can throw more cores at the problem at any time, they got processors with lots of cores for years now... They just dont feel like putting them in regular consumer price range.

 

Look up Intel Xeons, look up Intel Phi.

 

Read my previous post for the concepts behind AMD's current architecture.  It's slower because it's missing the last piece of the plan.  The stars architecture has already been refined again and again.  There is not much left you can do with it besides die shrinks.

 

An A10-5800K is about $120, if you want a dedicated GPU + quad-core that can beat that you'd have to spend another $100 just for a dedicated GPU that will make a meaningful improvement over the A10's integrated graphics.

 

Bulldozer came before Xeon Phi... Xeon Phi is not a CPU anyway, it is a coprocessor that accelerates floating point operations for the CPU, similar to a GPGPU.

[Dravic]

Read my previous post for the concepts behind AMD's current architecture.  It's slower because it's missing the last piece of the plan.  The stars architecture has already been refined again and again.  There is not much left you can do with it besides die shrinks.

 

An A10-5800K is about $120, if you want a dedicated GPU + quad-core that can beat that you'd have to spend another $100 just for a dedicated GPU that will make a meaningful improvement over the A10's integrated graphics.

 

Bulldozer came before Xeon Phi... Xeon Phi is not a CPU anyway, it is a coprocessor that accelerates floating point operations for the CPU, similar to a GPGPU.

 

As i already said in a different thread, APUs are shitty and their price-to-performance ratio sucks balls of any GPU+CPU combo including Phenom II x4 + 70$ Mobo + 7770.

[Vitalius]

As i already said in a different thread, APUs are shitty and their price-to-performance ratio sucks balls of any GPU+CPU combo including Phenom II x4 + 70$ Mobo + 7770.

I think you are still missing the point. The point of APU's, based on what Glenwing is saying, is not to game, but to offload the Floating Point calculations to the GPU side (where floating point calculations get done muuuuuch faster) and have a dedicated GPU for gaming. This way their CPUs are more powerful because they are focusing on the type of calculations that GPUs are  not good at, specifically, Integer calculations.

Off-loading the Floating Point calculations to the GPU side of an APU will make the CPU side stronger. Basically, they are building a chip that has cores designed for each of the two specific calculations (Integer and Floating Point), this way, they can separately focus on what they are best at rather than having a CPU do the floating point calculations, which it is not particularly good at.

It is kind of like Hyper Threading too, in that both Integer and Floating Point calculations will be done at the same time, by two different sides of the APU. So a more advanced version of it, Kinda.

At least that is my understanding based on what I have read so far in this thread.

[Dravic]

I think you are still missing the point. The point of APU's, based on what Glenwing is saying, is not to game, but to offload the Floating Point calculations to the GPU side (where floating point calculations get done muuuuuch faster) and have a dedicated GPU for gaming. This way their CPUs are more powerful because they are focusing on the type of calculations that GPUs are  not good at, specifically, Integer calculations.

Off-loading the Floating Point calculations to the GPU side of an APU will make the CPU side stronger. Basically, they are building a chip that has cores designed for each of the two specific calculations (Integer and Floating Point), this way, they can separately focus on what they are best at rather than having a CPU do the floating point calculations, which it is not particularly good at.

It is kind of like Hyper Threading too, in that both Integer and Floating Point calculations will be done at the same time, by two different sides of the APU. So a more advanced version of it, Kinda.

At least that is my understanding based on what I have read so far in this thread.

 

Hyperthreading compared to APU? You've lost me man.

 

Nah for now it's all about money, AMD is in troubles because they made a jump to worse architecture instead of pushing Phenoms and now they are facing the results and trust me there's no smile on their face. They are barely making it to get back from their bad state.

[Vitalius]

Hyperthreading compared to APU? You've lost me man.

 

Nah for now it's all about money, AMD is in troubles because they made a jump to worse architecture instead of pushing Phenoms and now they are facing the results and trust me there's no smile on their face. They are barely making it to get back from their bad state.

Well, Hyper Threading is basically when a CPU core does both an Integer calculation and a Floating Point calculation at the exact same time, thus allowing it to do 2 calculations at once which is basically where "virtual cores" come from. 

On an APU, since there is a GPU side and a CPU side, each side can do a calculation (the CPU doing Integer calculations, the GPU doing Floating Point calculations) thus effectively replicating Hyper Threading on an AMD platform.

That is just my speculation though. I am not sure how "separate" the GPU and CPU sides are and so I do not know if they can function independently enough that they can do separate calculations like that. However, I imagine, based on what Glenwing said, that is what AMD is inevitably planning to do. I do not think they have that yet on the current APUs or there would be some mentioning of it.

Being all doom and gloom does not do much. I completely doubt they are losing money still on the APU market. If we ignore everything else that APUs are great for (specifically low budget systems), AMD still has the consoles. Which I am more than certain is a very nice penny in their pocket. 

Think about it. The APUs in the consoles are supposed to be next generation ones (in that they are better than any sold to consumers right now). That is likely AMD putting their ideas into practice and experimenting with what can be done with these new APUs. 

I could be wrong. I doubt it.

[Glenwing]

As i already said in a different thread, APUs are shitty and their price-to-performance ratio sucks balls of any GPU+CPU combo including Phenom II x4 + 70$ Mobo + 7770.

 

A 7770 costs about as much as an APU by itself, that and a Phenom II isn't really in the same price bracket.  For low budget builds that still want some gaming, APUs are really not bad, and in the mobile market they are excellent.  That's where Intel's been focusing its efforts for the past 2 generations and AMD is still looking over their shoulder at them there.

[Glenwing]

Well, Hyper Threading is basically when a CPU core does both an Integer calculation and a Floating Point calculation at the exact same time, thus allowing it to do 2 calculations at once which is basically where "virtual cores" come from. 

On an APU, since there is a GPU side and a CPU side, each side can do a calculation (the CPU doing Integer calculations, the GPU doing Floating Point calculations) thus effectively replicating Hyper Threading on an AMD platform.

That is just my speculation though. I am not sure how "separate" the GPU and CPU sides are and so I do not know if they can function independently enough that they can do separate calculations like that. However, I imagine, based on what Glenwing said, that is what AMD is inevitably planning to do. I do not think they have that yet on the current APUs or there would be some mentioning of it.

Being all doom and gloom does not do much. I completely doubt they are losing money still on the APU market. If we ignore everything else that APUs are great for (specifically low budget systems), AMD still has the consoles. Which I am more than certain is a very nice penny in their pocket. 

Think about it. The APUs in the consoles are supposed to be next generation ones (in that they are better than any sold to consumers right now). That is likely AMD putting their ideas into practice and experimenting with what can be done with these new APUs. 

I could be wrong. I doubt it.

 

Hyperthreading isn't about integer or floating point actually, it just lets each core use idle cycles for a secondary thread whenever the primary thread is stalled for various reasons.  Each core has integer and floating point capabilities, so both threads could be either one.

 

HSA is more like integrated CUDA, but with a little bit broader usage.  With HSA, floating point operations on APUs would gain performance in several multiples of what it would be just using CPU cores, and since games are pretty heavy on floating point math, like I said it should be a good showing.  Additionally, if AMD adopts HMC DRAM as standard on their platforms, which they have shown interest in, that will give even a greater boost.  Of course this is all speculation, until someone actually does it we can't know what the numbers will be.

[Vitalius]

Hyperthreading isn't about integer or floating point actually, it just lets each core use idle cycles for a secondary thread whenever the primary thread is stalled for various reasons.  Each core has integer and floating point capabilities, so both threads could be either one.

 

HSA is more like integrated CUDA, but with a little bit broader usage.  With HSA, floating point operations on APUs would gain performance in several multiples of what it would be just using CPU cores, and since games are pretty heavy on floating point math, like I said it should be a good showing.  Additionally, if AMD adopts HMC DRAM as standard on their platforms, which they have shown interest in, that will give even a greater boost.  Of course this is all speculation, until someone actually does it we can't know what the numbers will be.

Ugh, and I thought I understood what hyper threading was. Thanks for the clarification. 

What is HMC DRAM?

[Glenwing]

Ugh, and I thought I understood what hyper threading was. Thanks for the clarification. 

What is HMC DRAM?

 

http://www.extremetech.com/computing/152465-microns-320gbsec-hybrid-memory-cube-comes-to-market-in-2013-threatens-to-finally-kill-ddr-sdram

[Vitalius]

http://www.extremetech.com/computing/152465-microns-320gbsec-hybrid-memory-cube-comes-to-market-in-2013-threatens-to-finally-kill-ddr-sdram

Good read, I only skimmed, but I will come back when I have more time. Sounds awesome.

As for the whole splitting integer and floating point calculations between CPU and GPU cores respectively, that, imo, goes hand in hand with cores sharing resources (L2 cache) like how is being done in the FX CPUs. 

Imagine a FX APU. Like where they fuse the "lots o' cores" FX CPUs with the APU concept. Just do it. It will blow your mind. I hope. Add in HMC DRAM and bam. Awesomeness.

 

[A/C]

Wow man, just wow. This is the best thread on the subject I have ever read, bar none. Thank you sir for your contribution to my CPU knowledge. Here's a cookie.

[thexzenon]

Great read, thanks for all this. Keep up with these informative threads.

[Vitalius]

Well guys I've been doing my own research & analysis of the performance of AMD & Intel CPUs for sometime, I did not do tests personally however I tried to salvage as much knowledge as possible to try and answer a question that has been on my mind, which is how far ahead is intel ? and that question lead to me to some very interesting realizations.

The discussion will be highly technical and might seem boring to some, but I wanted to gauge the interest of the community before I go on about writing what I had found out, because it's going to be quite long & do not wish to spend my time on something the community is not interested in.

So is anyone interested ?

UPDATE : Ok you seem to be interested guys, I'll start off with the basics & end with the answer to the original question plus several others.

Ok let's begin, first off our discussion will be surrounding AMD's modular Bulldozer/Piledriver architecture & Intel's core series from Westmere to Sandy/Ivy & Haswell.

The Architectures : AMD Bulldozer/Piledriver

The building block of AMD's modular architecture is not an x86 core, it's an x86 module, this module contains everything that a core contains but with a number of components doubled, these components include the integer scheduler, its datapath, 16KB of L1 DCache & it's own load/store unit, so essentially two integer cores.

The doubling of these components is what gives the AMD module two threads.

The integer cores share the early pipeline stages (e.g. L1i, fetch, decode), the floating point unit, and the L2 cache with the rest of the module.

 

How this affects performance :

Depending on the workload the quad module eight core AMD processors can either perform as eight processing units or as four and this depends entirely on the affinity of the workload to use integer calculations or floating point calculations.

If integer calculations are needed there are eight integer pipelines that can be used, however if floating point calculations are needed there are only four.

The difference between an integer processor and a floating point processor is that an integer processor only processes integers, as in whole numbers (including fixed fractions), while a floating point unit can process floating-point values which are fractions that don't end so to speak.

So an integer core can process a value that look like this 5/2 (which equals 2.5), while a floating point unit can process a value that looks like this 22/3 which equals 7.33333333333.... (a number that doesn't end) which in turn is approximated to 733333 ×10^-5 or 7.33333.

Think of the floating point unit as the invention of the scientific notation but for processors.

Why is a floating point unit needed ? well there is a range to which an integer core can process information, a 64bit processor can process 2⁶⁴ discrete numbers, if the process you require falls out of range you approximate it if needed & code it for the floating point unit, this can also help save on-die memory space.

You can read more about the difference between integer processing and floating point processing here & here.

Here are examples of the effect this has on performance.

in 7-Zip the compression workload is very integer core intensive and thus a quad module AMD CPU performs well because it can provide eight integer threads.

However in floating-point heavy workloads like a synthetic render scene in Cinebench the eight integer pipelines are throttled by the four floating point pipelines in the 8350 which holds performance back.

If the 8350 had eight floating point units single-threaded performance would be multiplied by 8 resulting in an 8.8 score.

The sharing of the early pipeline stages like the decode stage can also present similar symptoms to the sharing of the floating point unit.

Elimination of this resource sharing can improve performance by up to 25%.

Here is the in-depth analysis.

Steamroller will double the decode resources, which will minimize resource sharing and make it exclusive to the floating point unit.

AMD's building block v.s. Intel's.

AMD Bulldozer/Piledriver Module (2 integer cores) on top, Intel Westemere core on the bottom.

Both are built on the 32nm process, Sandy Bridge is also built on the 32nm process.

Cores excluding the L2/L3 2MB cache are.

Westmere: 17.2mm^2

Sandy Bridge: 18.4mm^2

Bulldozer Module: 19.42mm^2

So manufacturing process aside, a modern intel core is roughly the same size as an AMD module.

What is a CPU die ? click here

Size comparison :

Ok so now that we've established that two AMD "cores" are roughly the same size as one intel Sandy/Ivy/Haswell core, lets dive in the benchmarks.

We'll be comparing the FX 8350 (315mm^2) & the i7 3820 (294mm^2), the FX 8350 is approximately 21mm^2 larger because of the extra 6MB of Cache memory it has compared to the i7 3820.

Compression : uses all integer threads available.

Hash-rate, same algorithm used to mine bitcoins : uses all integer threads available.

FPU VP8 / SinJulia : uses all floating point threads available.

7-Zip file compression : uses all threads available.

Video encoding : uses all threads available.

Photoshopt Cs6 : uses one thread.

Cinebench : single threaded

Ray-tracing renderer : uses all threads available.

So we're noticing a pattern here, we're all familiar with, single threaded workloads run faster on the larger intel cores, however anything that utilizes all threads available runs faster on the smaller more parallel architecture of the AMD module.

This is nothing new, the 4 extra logical threads of intel's i7s helps it keep up with AMD's eight physical integer cores. What about intel CPUs that don't support hyper-threading though ?

In that case 3 AMD modules become as fast as 4 intel cores as long as they are clocked 500mhz higher, although 500mhz isn't really high considering the haswell i5s usually top at 4.2Ghz while the 6 core AMD processors can reach 4.7Ghz on the same cooling.

FX 6300 @ 3.5Ghz vs i5 4430 constant turbo on all 4 cores @ 3.0Ghz.

FX 6350 @ 3.9Ghz vs i5 4570 constant turbo on all 4 cores @ 3.4Ghz

So without hyper-threading an AMD Module becomes significantly more powerful than an intel core.

 

Back to my original question which is: how far ahead is Intel ?

An AMD module packs more performance than intel core which takes roughly the same die area.

So architecturally speaking, Intel isn't really ahead of AMD nor is AMD ahead of Intel so to speak, the parallel nature of the AMD modules gives them the pure performance advantage over the intel cores, however what intel lacks in total CPU performance makes up in single-core performance.

In an absolute since an AMD module can process more data faster than an intel core, however there is a sacrifice which is single-threaded performance.

Although the majority of CPU-intensive programs are already multi-thread reliant & the ones that aren't are becoming so which will in turn make this sacrifice worth it for AMD.

The reason why Intel stuck to single-threaded performance focused CPUs is because it's easier to program for them but it's becoming extremely difficult to push more performance out of a single thread which makes the move to higher parallelism inevitable.

Where Intel is really ahead of AMD is not the architecture, it's the manufacturing process, intel have had 22nm available to them for over a year, while AMD only now is moving to 28nm, this is because they sold their foundries and so they are forced to use whatever manufacturing process is available by Global Foundries or TSMC.

The fabrication process gives Intel two advantages over AMD, one is power efficiency, the second is higher profit margins and that's where intel is really ahead of AMD.

It was a long-ride I hope you've enjoyed it.

And I did enjoy it now that I have read it.

Just a few things:

1. "In an absolute sense, an AMD module can ...." not since.
2. Based on this, Intel is ahead of AMD in regards to gaming due to two reasons: One being that most games are single threaded, mostly and two being that games rely on Floating Point operations a lot which AMD is weaker at, in comparison to Intel. 
3. This explains why AMD is pushing their APU line, having a dedicated GPU for doing the floating point calculations would take away one of the two major weaknesses of their CPUs (floating point operations).

Those are just my thoughts based on my understanding attained from reading this. I could be wrong.

[TechFan@ic]

And I did enjoy it now that I have read it.

Just a few things:

1. "In an absolute sense, an AMD module can ...." not since.
2. Based on this, Intel is ahead of AMD in regards to gaming due to two reasons: One being that most games are single threaded, mostly and two being that games rely on Floating Point operations a lot which AMD is weaker at, in comparison to Intel. 
3. This explains why AMD is pushing their APU line, having a dedicated GPU for doing the floating point calculations would take away one of the two major weaknesses of their CPUs (floating point operations).

Those are just my thoughts based on my understanding attained from reading this. I could be wrong.

1. fixed

2. added

3. added.

Thank you everyone so much for your interest & valuable feedback.

[Kuzma]

Good read, I only skimmed, but I will come back when I have more time. Sounds awesome.

As for the whole splitting integer and floating point calculations between CPU and GPU cores respectively, that, imo, goes hand in hand with cores sharing resources (L2 cache) like how is being done in the FX CPUs. 

Imagine a FX APU. Like where they fuse the "lots o' cores" FX CPUs with the APU concept. Just do it. It will blow your mind. I hope. Add in HMC DRAM and bam. Awesomeness.

 

Psssst :p I kinda said imagine an FX-APU hybrid earlier on,  this is kinda making me consider going for the FM2+ platform now, It would actually be very helpful for me from  a developers POV and it would be nice to be able to play games with my dedicated GPU while rendering videos with my on board gpu if that's even possible? Should be though.

[Kuzma]

It's things like this that shock me to be honest:

[Cobalt]

Wow, I wasn't excepting that, AMD reallly has a BIG, BIG, BIG plan running in the background, APU, ATI, they all add up if you think it

 

Kaveri could even jump to middle-high range and use full-power steamroller cores, I guess even the rumor of them abandoning FX and AM3+ makes total sense now, and Gigabyte releasing some pretty high-end FM2+ boards (they higest end I have seen in an APU platform at the day), call me conspiracy crazy, but all of this, this is gonna be so ridiculously awesome.

 

In the end AMD is looking for the long run, so awesome!!

[A/C]

Wow, I wasn't excepting that, AMD reallly has a BIG, BIG, BIG plan running in the background, APU, ATI, they all add up if you think it

 

Kaveri could even jump to middle-high range and use full-power steamroller cores, I guess even the rumor of them abandoning FX and AM3+ makes total sense now, and Gigabyte releasing some pretty high-end FM2+ boards (they higest end I have seen in an APU platform at the day), call me conspiracy crazy, but all of this, this is gonna be so ridiculously awesome.

 

In the end AMD is looking for the long run, so awesome!!

 

Yes, and I love it!

 

Always liked AMD a that much more than intel for being "alternative" (in my head at least), since the days of pentium III... And I'm happy they seem to have a plan how to stick it to the man. And frankly I'm tired of having only intels around me, at house, at work... I'd be proud to own an AMD again. 

[Cobalt]

Just a few more things that popped into my conspiracy head:

Think about this thread and those others about Intel using their libraries to sabotage AMD, and adding that up, even then is AMD winning in all those benches?? Just blowed my mind, just imagine with projects like the Yippe!! (that's how it is called right?)libraries!!

and also, are the changes that we are seeing on the desktop with APU similar in the mobile side??

[TechFan@ic]

Just a few more things that popped into my conspiracy head:

Think about this thread and those others about Intel using their libraries to sabotage AMD, and adding that up, even then is AMD winning in all those benches?? Just blowed my mind, just imagine with projects like the Yippe!! (that's how it is called right?)libraries!!

and also, are the changes that we are seeing on the desktop with APU similar in the mobile side??

The situation is truly profound when you look deeply into it with the intel-ARM benchmarking scandal that's going on right now where similarly to what we've seen on the desktop a number of benchmarking software was specifically programmed to use libraries that would favor intel.

AMD is really pushing the heterogeneous architecture in the desktop & notebook spaces Qaulcomm, ARM & Nvidia are doing the same in the mobile space where they're pushing more balanced chips in terms of CPU & GPU performance, Nvidia is also doing something very semilar to AMD's hUMA (unified memory access) on the tegra line with the upcoming Maxwell.

So the whole industry is pushing in this direction & Intel is doing the same thing, where year after year they significantly increase the GPU die area on their processors to increase performance, so even though they don't like to admit it, they're also moving towards APUs.

[leipero]

"The second advantage is higher profit margins and that's where intel is really ahead of AMD." = Sad but true, especially when you have on mind that company who pushed PC technology so far done worse (I'm not saying Intel didn't, because that would be a lie, I'm just saying AMD did more).