AMD/Intel Comparison

AMD/Intel Comparison
​

Always trying to help people understand technology
​

by Serg​

1. Introduction

Explaining what a CPU is, is rather obvious at this point. But if you need to know, here is the definition by Wikipedia explaining it. The main thing about this article is to explain how Intel and AMD differ in technology, mostly focused on where they allocate the cache.

Most of the laptop market is ruled by Intel, being HP the only major brand using AMD mostly to power their HP Pavilion notebooks, which are consumer-oriented. But why has this happened? Is it really that AMD is so far behind Intel as far as technology and efficiency goes? Is it that one is that far superior than the other? Well, I am going to explain some differences between the two, most importantly how they "search" for instructions.

2. Intel Cache Architecture

As some of you know, I wrote a few months ago a thread where I explained about Intel CPUs. It was called the Intel Processor Family, so to be fair, I know my way around Intel CPUs, their nomenclature, numbers used, what those numbers mean in the performance area and the letters, what do they stand for and TDPs. I understand the basic Architecture and how it works. The way Intel works is completely different as AMD's way of building CPUs.

Intel uses L1 Cache for each core and a shared L2 cache between all cores. For example, the P9700 clocked at 2.8GHz has a 1GHz FSB and uses a 6MB shared L2 cache between the 2 CPUs and has 2x32+2x32KB L1 cache+instruction, therefore each core uses 32KB L1 cache and a 32KB instruction for that cache, right? This is easy to follow so far. Some extra rough info on this here.

​

For those who don't know.
Cache memory is the fastest memory type there is in a CPU. Wikipedia defines it as "The cache is a smaller, faster memory which stores copies of the data from the most frequently used main memory locations." ( link)

This means that the cache is where the CPU searches for bits of information there before going to the RAM and in the last case, to the HDD. So it works this way: CPU needs to run a thread. The thread gives an order which the CPU needs to process. It goes first to L2 shared, secondly L1 core-dedicated, next RAM and finally HDD looking for the instructions to run it. (this is in Intel).

Is this clear enough so far?​

3. AMD Introduction

Now it comes to AMD. Although AMD is not quite as popular in the laptop world, it is still a very important brand that we must consider, specially inside a budget. I know this sounds rather cruel to AMD brand, but AMD/ATi is known to be less expensive than the competition, and that is why some people think they are not as good performers. This couldn't be any further from the truth.

Even though AMD does not perform as well as Intel in a clock-to-clock comparison, overall performance is rather close to Intel's. In fact I use a 2-year-old AMD Sempron 3600+ at 2.0GHz on my laptop (no, I am not biased) and I can safely say it runs as fast as a similar-date Intel Celeron running at the same clock speed.

​

Disclaimer: I am not saying that one is better than the other.

4. AMD Cache Architecture

As aforementioned, Intel uses a L1 core-dedicated and a L2 shared Cache system. AMD on the other hand uses a somewhat different system where both L1 and L2 are core-dedicated and they have up to a L3 shared memory. Even with mid and high-ranged dual-cored CPUs. A good example is AMD Athlon x2 64. This is a 65nm version of the CPU, therefore it is quite old compared to the actual 45nm versions.

Why does AMD use this system? Their approach to faster processing power relies on the fact that they place the cache closer to the core, or so called on-die, allowing it to access info faster, while Intel uses a massive shared center of cache. This difference is quite important, and most people (like me before) don't know this. Back to the placement of the L1 and L2 cache, AMD uses a smaller one, due to it being on-die, and on the latest CPUs, such as Athlon II x2, a L3 is present, and all Phenom I and II have L3 as well, but shared. More info on Phenom II and performance here.

NOTE: Tigris platform is based on Phenom II, therefore it should utilize the same L1, L2, L3 configuration, but I cannot find any info on this.

​

Note how the L1 is 2x64KB like versus Intel's 2x32KB, there is a 512KB dedicated L2 Cache for each core and there is a 3MB L3 shared between the 2 cores. Phenom CPUs are based on the same system of L1 and L2 dedicated and L3 shared.

​

On the other side, Intel uses L1 dedicated and dual-cored-shared L2 cache for quad cores, while AMD uses L1 and L2 dedicated and L3 shared between the 4. This is now being used on the latest Intels, but until Yorkfields, L3 was not present.

So AMD may not be so far behind on this...Intel might have a better energy management and produce less heat, but AMD performs very well too. Unfortunately, AMD has the reputation to overheat laptops, and most people dont recommend it, saying that "if you choose AMD, be prepared to return posting about overheat". And believe me, I have read this types of posts tons of times here in NBR, not just making that up.

5. AMD Future

Unluckily, AMD is behind the "die-shrinking race" as I call it. Intel is far ahead here, having working models of Westmere, and tons of posts around the web. I honestly haven't seen any on the AMD Copperhead (?) based on the same 32nm process. Some roadmaps say it is due to 2011, but some others say 2009...this is rather confusing. But AMD claims they would be ready to start producing them, but not until Q3 2010 they could mass produce them.

​

Apparently, by 2010, AMD and Intel will be very much on par...News about 32nm AMD CPUs here. So I hope not to see people saying "Intel is far superior" or "Intel is ahead".

6. Conclusion

If you are looking for a good CPU for your laptop, don't try to compare AMD and Intel by the same terms. Remember they use different architecture and speeds to achieve the same goals. Cache might be larger on Intel, but AMD places it on-die, closer to the core, so it can access the bits of info a lot faster. And HyperTransport technology on AMD allows CPUs to achieve even faster communications.

So bottom line is that it is not useful to compare both based on the same specs, without considering the technology they use. For example comparing ATI and NVIDIA in the shaders count, if done by rough numbers, well, we all know the answer...

So, please, don't go talking bad about AMD nor Intel

7. Disclaimer

I am neutral to either Intel or AMD. As both can satisfy my computational needs and both offer excellent performance, I will not say that one is better than the other, nor speak in favor of either one of them. This is merely a way to show people that AMD and Intel are NOT the same, and that is also good to know something else from time to time.

The order of AMD and Intel is due to alphabetical order, not to be confused with preference over any brand. Same case as ATI and NVIDIA, alphabetical order.



*If you feel I need to add something else, please, post it, and I will gladly add it. This is a good way to help people around know a little more.

 

Intel is ahead, see this.

And this is AMD's future roadmap:



Caspian is code name for the cpu on the Tigris platform. It still uses DDR2?

Intels roadmap for comparison:
Q1 2010: 32nm Westmere
Q3/Q4 2010: 32nm Sandy Bridge new architecture
2011: Ivy Bridge shrink of Westmere to 22nm (quad core becomes mainstream)
2012: 22nm Haswell new architecture

 

That was a fast reply...

 

Nice post, was thinking about pulling together my posts on GPUs & CPUs in the same sort of manner, but just haven't found the time.

 

The advantage that Intel holds in the mobile space right now isn't due so much to their superior architecture as it is tot their superior manufacturing technology, Intel has been making 45nm mobile chips for more than a year and a half while AMD has inexplicably failed to transition from their weak 65nm process to their relatively strong 45nm process. As a result Intel holds a significant performance per watt advantage (~40%) despite the relatively similar published TDPs. Of course the extra performance won't be noticed by most users but the decreased battery life will. I am really cheering for AMD and recently built an AMD desktop but they are not very competitive in the mobile space right now.

 

Unfortunately I must agree with you deputc26. AMD has fallen behind, but if rumors are true, Phenom II Mobile is very competitive, as much as Tigris.

It is just a matter to wait and see. And remember that the L1 and L2 dedicated suppose a slightly more complex design, so I could say this is holding them back, figuring out how to cramp so many things in a smaller size. (I could be talking nonsense nonetheless).

I find that Intel's design is quite simpler, therefore shrinking it was simpler as well...

 

Honestly when has AMD ever competed with Intel in the mobile sector in performance/power consumption?

AMD has a clear advantage when it comes to platform with the GPU compared to Intel but their CPUs still lag behind. Most people will not give up a couple of hours of battery for a faster GPU when most laptop buyers rarely care enough about gaming to make it a deciding factor(I am not taking into account desktop replacement notebooks).

 

One thing that I've never quite understood is how die/size/process play into performance. I think smaller means more energy efficient, but does it mean faster?

And, if so, why? Less distance to travel>

 

Something else that is worth mentionning.

Intel keeps a copy of the L1 cache inside the L2 cache.

AMD's L2 cache is kind of the extension of L1 cache. The data is not doublicated between the caches.

 

The only way process shrinks increase performance is by allowing the chip to run at a higher clock speed in the same power envelope. Of course this can also be used to operate at the same clock speed in a lower power envelope.

 

Interesting guide...

I like AMD. They're cheaper, and really, not much slower. I care more about my graphics card than I do my processor. But it's too bad no one uses AMD.

 

Thanks a lot to all!

If you think something needs to be added/modified let me know.

------------------------------------------------------

What I still wonder is if the rumor that an ATI GPU/IGP will work better if paired with an AMD CPU...
AMD+ATI>Intel+ATI? True or False?

------------------------------------------------------

About the shrinking question. When the parts are smaller, they consume less battery, and can be placed together in a smaller form factor. This smaller form factor uses less energy, therefore you get more battery life. Now, if you want to keep consumption, you can cramp more smaller transistors in the same area, and the added transistors make the CPU work faster (faster clock-speed for same Watts).

What is different is that the Cache is on-die on both. But Intel only has L1 on-die, while AMD has L1, L2 on-die and even some support L3 shared. This means they have more components to shrink and make them use less energy.

Forgot to say that AMD HyperTransport is the FSB, normally smaller because (based on what I read) it can be overclocked and is shorter, making info travel faster. Not so sure about this last point, so I wont add it till this is confirmed.

 

I think you are being much to kind to AMD. Its mobile processors are currently not competitive with any Intel processors on the high end or even in the mid range. Furthermore, even on the low end, Intel's processors win out and as far as anyone can tell, they get quite a bit more battery life too.

The one good thing about AMDs processors is that they're dirt cheap, but you're pretty much getting what you pay for. This isn't necessarily a bad thing (e.g. some people just need a processor -- any processor -- and want a notebook larger than netbooks), but AMD needs to port its newer architecture to laptops and shrink the die size to be competitive.

 

Intel always had the advantage over AMD in mobile, even in the worst times. Now that Intel is far ahead of AMD in other segments, they are even more pronounced in mobile.

Anyway take a look at this: http://www.anandtech.com/mac/showdoc.aspx?i=3582

Same Macbook achieves better battery life under OSX than Vista. I believe that might be due to the problem that Windows don't really allow threads to go to idle. Core i7 with its hardware PCU might be able to take care of that problem with better control over it. I think the advantages are going to get even wider in October...

I just noticed:

I'm sorry but this is plain wrong. They are both using on-die caches. Do you even know what that means? On-die means when the manufacturer designs the CPU, they consider the cache part of the CPU design rather than making a seperate L2 cache and putting them next to the CPU die. The term you are looking for is Dedicated cache vs. Shared Cache. And by the way, the cache measurement shows that the Intel caches are faster than AMD caches making the point moot.

Also, I think you are confusing L1 with L2. In Intel processors(actually ALL processors), L1 cache is accessed first then L2 is accessed. But in case of Intel, the contents of the L1 cache is duplicated in the L2 cache. That doesn't mean its not on-die.