Intel Architecture vs AMD architecture, which is better?

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
what GHZ??? the slowest it can get is going to be at least 1.5 THZ (1THZ = 1000Ghz)

and thanks for wishing me good luck, i need that.

btw the cpu im going to make is going to be 1 in all, what i mean is you dont need a mobo/hard disk/ram/gpu, everything will be in that cpu on die, first week is going to be on sale for only ummm 450 $ NOT BAD!!!!!
 


All right, I'll try to do my best here. First of all, AMD is not doing badly at all. Their CPUs are solid and even though they are a little slower in some programs than the fastest Core i7s, the overall price of an equivalent AMD system is noticeably less than an Intel system and it is hard to recommend Intel for any desktop system except for higher-end systems. Also remember that AMD's CPUs do very well in gaming and are frequently faster than Intel's CPUs. It is my opinion that Intel's fastest CPUs are generally a little faster than AMD's because of the following:

1. Compile-time optimization targeting Intel CPUs is far more common than compile-time optimization targeting AMD CPUs on Windows. This is especially true for programs compiled with Intel's own ICC compiler, which has a cpuid() check that can disable certain optimized code paths on AMD CPUs but leaves them enabled on Intel CPUs. This is apparent if you look at benchmarks of the same programs compiled with a vendor-neutral compiler like GCC running on Linux. Go look through the Phoronix database and the Phenom IIs are considerably more competitive with the Core i5s and i7s than they appear to be on Windows.

2. Greater clock speed on lightly-threaded tasks due to the use of Turbo Boost, particularly on Lynnfield CPUs.

3. Greater memory bandwidth due to the 192-bit IMC on the Bloomfields.

4. HyperThreading allowing for a potential ~20% boost in highly-threaded applications.

5. Typically larger caches. Intel was bolting huge caches on the Core 2s as the Yorkfields had up to 12 MB of cache onboard. This is a smaller issue today as both the Core i5/i7s and the Deneb Phenoms have about 8 MB of usable cache- Intel has an inclusive cache strategy with 8 MB of L3 and AMD has exclusive caches of 6 MB (L3) and 2 MB (L2s).

6. The Core-based chips are four-issue, while the Phenoms are three-issue. This is more of a theoretical advantage as I've only ever seen peak performance of 2.x IPC retired in x86 CPUs. if somebody has better examples, tell me.
 
Finally, thank u MU Engineer, something along the lines of what i was asking for

now i just need more people to back up or contest ur info to get a good knowledgeable thread going here
 
All K8-K10 are 3 issue. Ever since Core, Intel has been using 4-issue. Just one of the many reasons Athlon 64s lost to the Core 2s...

That and the fact, with i7, Intel has gone with an extreme speed L1, speedy L2 and giant L3. It is just much more efficient than what Intel used to have (double chezburger). The L1 and L2 caches on the i7 are independent and information is shared between the cores. This is the same with the Phenom. With the Core 2 Quad, it is really two Core 2 Duos. For cores to communicate, they had to go through the FSB to the other cores. i7s and Phenoms have a shared L3 to communicate. For that reason, Core 2 Quads were handily beaten by Phenoms in server-like tasks which required huge amounts of data.

 


Theoretically speaking, if the AMD cpu's were to also 4-issue nothing else changed how much of a performance increase could that lead to if any?
 


I don't think that has a whole lot to do with it as from what I have seen, actual x86 code doesn't lead to more than 3 instructions being retired per clock cycle, so the extra issue width of the Core 2 doesn't really help much.

The biggest reasons why the K8 lost to the Core 2 were that the Core 2 had much larger caches and could execute 128-bit SSE instructions in one clock cycle versus 2 instructions every 3 clock cycles for the K8. The K10 has the same SSE processing abilities as the Core 2s, which is why the K10 and Core 2 are much closer in per-clock performance. The original 65 nm K10s mostly just suffered from a lack of clock speed and somewhat from a lack of cache size and speed. The current 45 nm K10s do very well against the Core 2s as they have a similar clock speed and fair-sized caches.

That and the fact, with i7, Intel has gone with an extreme speed L1, speedy L2 and giant L3.

I don't think the cache structure on the Nehalem is anything special. IIRC the L1 cache on the Nehalem is slower than the L1 on AMD CPUs- the Nehalem's is 4 cycles versus 3 cycles for the K10. It's also half the size, but it has a higher associativity. The L2 cache is relatively quick and I think maybe a touch faster than AMD's L2, but it is a 256 KB inclusive L2, so it is pretty small with only 192 KB of data that is not a mirror of L1 cache data. The L3 is a little larger than AMD's, but it has a lower associativity and it is also inclusive, so that 8 MB of L3 has only 7 MB of usable space, since it has L2 and L1 cache contents in it. The Phenom II has a greater effective cache size as it has exclusive caches that gives it 8.5 MB of non-duplicated cache capacity. The one thing that Intel does have an upper hand on is the speed of the L3 cache as the fastest L3 cache speed on any AMD CPU is 2.2 GHz, while Intel has some that go up to close to 3 GHz.

It is just much more efficient than what Intel used to have (double chezburger). The L1 and L2 caches on the i7 are independent and information is shared between the cores. This is the same with the Phenom. With the Core 2 Quad, it is really two Core 2 Duos. For cores to communicate, they had to go through the FSB to the other cores. i7s and Phenoms have a shared L3 to communicate. For that reason, Core 2 Quads were handily beaten by Phenoms in server-like tasks which required huge amounts of data.

Yes and no. The FSB was used to send data between the two different dies in a C2Q, but in a single-socket setup, the higher clock speed of the C2Qs generally allowed them to outrun the Phenoms as the FSB was enough to handle this traffic. It took four dies (two quad-core Xeons) to be noticeably crippled by the FSB bandwidth bottleneck, where considerably lower-clocked Opterons outran FSB-equipped Xeons.
 
AMD may have competitive products in the sub-nehalem market. Yet their price advantage means nothing to Intel's brand weight and OEM availability.
First of all, Pentium 4 sold more than Athlon 64 because all the "Intel Inside" campaign made non-tech savvy buyers look for the logo; even thiough AMD had better CPUs. AMD had poor marketing schemes.
A couple of years back, when I fist got interested in technology and CPUs, I was dissapointed to know my machine had an AMD processor. I thought it was just a crappy budget alternative to Intel (because I had not heard of it before), and was desperate to switch to the "great" Pentium 4.
Added to this, Intel's CPUs have been outperforming AMD's since Core 2, so solid product + good marketing = Intel's huge success.
So, even if the Athlon II X4 or some Phenom IIs offer better price-performance than Intel's CPUs, their lower price doesn't make up for Intel's well known name.
Just an example, I ask people about their CPUs, and they say they "think" it's an Intel dual core; or Intel "core dual 2" or "2 duo core"; and in some cases, they actually have an Athlon X2!!!
Lastly, I don't see much AMD prescence in HP or Dell, and since most people buy their PCs from OEMs, I can see why AMD's market share has plummeted. I don't know whether it is poor AMD-OEM relationships or Intel's illegal practices, but AMD CAN compete with lower end products, and they are the important ones (make most sells).
AMD, nevertheless, NEEDS a "conroe" architecture. Their hope is Bulldozer if they don't go bankrupt by 2011. 2010 seems to be a tough year for them.
 


no sir that isn't the competition here, we are competing to see who has the most knowledge and who can prove you, jennyh, or fazers wrong on ur facts, or vice versa

which means we all bow down to you o great one
 

Not everything is a competition, meaning you don't have to make everything one.
 
Hey

Didnt read whole topic but what I have notice yesterday ... that...
Intel Q6600 is rendering scene in 3ds max 100% use pc and it can stream video thorught skype withput no problem

The amd quad core pc died, still did the render but didnt stream any video.

Yes it can be priority foult etc etc but all was on normal... so since yesterday I ges Intel>AMD by me ... :s
 

I'm just curious which AMD CPU did you use and did both CPUs have similar system specs? My girlfriend uses a stock Q6600 and I use a Phenom II 955 and I run things much faster than her (using stock settings).

Just to run a similar test, I am using a Phenom II 955 (stock clock/voltage for this test) and a slightly OCed 4890 Toxic (1000mhz core/1100mhz mem) while testing Single/Multi-core 3D image rendering and OpenGL rendering with Cinebench R10 (64-bit). I will also be playing an MMORPG, having Ventrilo and AIM on, playing music through iTunes and having Firefox streaming a HD Youtube video all at the same time.

Results: I have no slowdown in 3D rendering, playing my game, watching the video, talking on AIM and Ventrilo, and listening to my music, even at 100% CPU load during the Multi-core 3D rendering.

Intel > AMD, in terms of i5 and i7 over AMD products, but the Phenom IIs from AMD are rock solid. Phenom IIs are designed to be Core 2 Quad killers and they do a fine job at a much lower price.

i7>i5/PII>C2Q is really more like it.
 
kokin humh well it was not about speed in our clase... was more about how good processor can multi task... AMD died with single live upload cam converencion (sry if I spelled it wrong) + Rendering simple scene in 3ds max when Q6600 handle it without problems... I'm not saying AMD is bad etc etc ! I used it for ages..it just came to my atention yesterday that AMD couldnt do what Intel could <lol>

I was not even a test... he just skype me and show me his descopt to ask for help in rendering and everytime he render the screen just freeze... never had that kind of things with Intel... or meybe its just Windows 7 foult coz hes using it so I cant be really sure... :]
 

I see. I also run Windows 7, but I haven't encountered any problems yet. It does seem strange that his computer freezes for simple stuff like that, does he have a GPU or is he just using the motherboard's onboard graphics, if he has that? That might be what's causing his system to hang. Your friend should post up his problem and maybe people here can help solve it. I really doubt that an AMD quad will crap out with such few tasks. I want to say it's his graphics, but I don't know the situation.

Also, I understand it wasn't your intention to say AMD is bad, but I just wanted to defend that AMD CPUs aren't crappy by running a similar situation.
 
Sorry, I don't have any benchmark examples - but the quad issue capability is important in hyperthreaded CPUs as it provides more resources for two instruction streams to utilize.
 
intels architecture is superior, and intels marketing team is vastly superior, amd needs to understand how to market dragon which is an increadible platform, and they also need to learn how to use cache, if they can pull these off, intel even with superior arch wont be able to stop them