Which is better: AMD or Intel?

[kingneptune117]

Ok so heres the deal. Im getting a new motherboard and processor, which means I have to decide if I want an Intel or AMD.

It seems like the higher performing AMD processors are cheaper than Intels, but im not sure why. I was looking at this AMD as it performs amazing for its price and for what im using it for (flight simulator X)

Anyways, what are better? Im looking for nothing below 3.0 ghz and thats without over-clocking it. I also need it to include the fan.

 

[BadTrip]

GHZ mean next to nothing between AMD and Intel, being that Intel's ipc is higher. FSX loves high-end cpus and the more cores the better. If i remember correctly it supports up to 256 cores.

 

[hunter315]

To ask which is better AMD or intel is ignorant, and to ask for nothing below 3GHz shows you are not up to speed with modern CPUs. The i5 @ 2.66GHz performs similarly to the Phenom II 965 @ 3.4GHz, to compare a processor on speed alone is a poor decision. For FSX consider an intel i7 860, FSX will appreciate the additional threads, its a fair bit more than the i5 or any AMD offering and well below your 3GHz requirement, but it will perform better than any AMD offering for FSX. You need to make sure you also pair it with a balanced GPU to get the best experience.

 

[keithlm]

The i5 @ 2.66GHz performs similarly to the Phenom II 965 @ 3.4GHz, to compare a processor on speed alone is a poor decision.

Are you referring to the i5 which is rated minimally at a frequency of 2.66Ghz but is really running at 3.2Ghz in many of the benchmarks used for these benchmark comparisons that people often quote?

It would be not be revealing all facts to to claim that a processor that dynamically overclocks is always running at the lowest rated frequency.

 

[ElMoIsEviL]

Are you referring to the i5 which is rated minimally at a frequency of 2.66Ghz but is really running at 3.2Ghz in many of the benchmarks used for these benchmark comparisons that people often quote?

It would be not be revealing all facts to to claim that a processor that dynamically overclocks is always running at the lowest rated frequency.

That depends on the benchmark and how multi-threaded the benchmark is. In this thread the application in question is FSX (a multi threaded flight simulator).

FSX runs far better on the Intel Core 2 and higher architecture than it does on the AMD Phenom and higher architecture. FSX also loves Hyperthreading therefore an i7 (even the models on the socket 1156) fair much better than AMD Phenom/Phenom II based processors in this particular game.

Put it this way, my buddy Obi has a SuperMicro board with dual Opteron Shanghai Quads running at 3GHz each, my Core i7 920 eats it for lunch (8 cores vs. 4 cores) at anything above 3.4GHz in that game.

If you're into FSX, you're better off investing in an Intel Core i7 based system (even an i7 860) as the performance gains are rather large.

You would also want an adequate GPU for the game. I have it on good word that an nVIDIA Geforce GTX 285 is a great GPU for this game. I tested the game using a CrossfireX utility to enable support with a 4870X2 + 4870 and was able to max out the game with full DX10 features, advanced texture pack and the weather FX add-ons (the system in my sig bellow). That having been said AMD GPUs aren't usually recommended for FSX (I do have a Radeon HD 5870 sitting here waiting for the 5970 to be released so that I can pair them to try FSX to see if the extra TMU and ROP units help).

 

[jimmysmitty]

Are you referring to the i5 which is rated minimally at a frequency of 2.66Ghz but is really running at 3.2Ghz in many of the benchmarks used for these benchmark comparisons that people often quote?

It would be not be revealing all facts to to claim that a processor that dynamically overclocks is always running at the lowest rated frequency.

Lest you forget that its based on a max TDP and when all 4 cores are active it only goes up one step instead of 4 like it will when its only maxing a single thread. When all 4 cores are being used it will not go to 3.2GHz because it would break the thermal limit that is set.

But thats not important to remember at all.

BTW, SMT is a gimmick, remember that yet AMD plans on using it in some of its upcoming CPUs. Whoops.....

OP, I would listen to Elmo since he has first hand experience with this. FSX loves more cores even SMT based ones (virtual cores) so a i7 build would be a better way to go. Plus a ATI 5800 series would be good to go with its Eyefinity for multi monitor support.

 

[keithlm]

Lest you forget that its based on a max TDP and when all 4 cores are active it only goes up one step instead of 4 like it will when its only maxing a single thread. When all 4 cores are being used it will not go to 3.2GHz because it would break the thermal limit that is set.

But thats not important to remember at all.

BTW, SMT is a gimmick, remember that yet AMD plans on using it in some of its upcoming CPUs. Whoops.....

OP, I would listen to Elmo since he has first hand experience with this. FSX loves more cores even SMT based ones (virtual cores) so a i7 build would be a better way to go. Plus a ATI 5800 series would be good to go with its Eyefinity for multi monitor support.

So then you are telling me to just forget that it really isn't running at 2.66Ghz and just drink the fanboy-aid and accept 2.66Ghz as being what it runs at.

Sure. Okay. I'll pretend I'm completely stupid and completely agree with you. Or perhaps I might actually know better.

Also: the new AMD design doesn't really have SMT does it? It has multiple clusters. That's called CMT and is a completely different design which apparently you didn't know or don't understand. So we'll just call it SMT and pretend it is the same thing so that you can pretend that SMT is not just a bit of a gimmick even though it is.

Luckily the CMT can be disabled if needed. But then it won't be suffering from the same problems as SMT will it?

But then we already know that a lot of Intel fanboys will be pretending it is the same thing just so they can claim that AMD "changed their mind". Let's review: one core pretending to be two cores versus two real cores using the same pipeline and processing units and calling itself one cluster or module. Yeah... I can see how somebody that was clueless might actually not be able to understand that it is not the same.

 

[jimmysmitty]

So then you are telling me to just forget that it really isn't running at 2.66Ghz and just drink the fanboy-aid and accept 2.66Ghz as being what it runs at.

Sure. Okay. I'll pretend I'm completely stupid and completely agree with you. Or perhaps I might actually know better.

I didn't say that. I said it goes up by 1 step instead of 4. Reading comprehension.

 

[apache_lives]

look at your budget and see what performs best for your $ and that will tell you which is better, but as a general guideline, if you have the $$$ an Intel, if you have slightly less $$$ then AMD but check benchmarks anyhow.

Me? i prefer the HT on Intel and the 3rd memory channel on the i7 platform for 50% more memory.

 

[jj463rd]

Basically Phenom II like the 955/965 will roughly be on par with the Core 2 Quad Kentsfield so it would be like having an already oc'd Q6600.It's not quite on par with Core 2 Quad Penryn's though in terms of IPC.
That's not to say that it's bad.
Phenom II will run FSX with less stuttering though (more smoothly) than the Q6600.
A year and a half ago many people ran oc'd Q6600 CPU's to run FSX until Core i7 came out.
Rather one has to decide on whether to run FSX in higher resolutions with more eye candy in the simulation in which Core i7 will do much better and the Bloomfield Core i7's will oc to a much higher degree than Phenom II by around 50% as compared with the max oc of Phenom II of 20%.
To run FSX with high levels of detail one wants to oc as much as possible to obtain it.
However AMD will have their 6 core Thuban CPU out next year as an upgrade for the AM3 platform and Intel will have their uber expensive Gulftown Core i9 as an upgrade for LGA 1366 so one could also take that into a possible consideration.
It really depends on your budget.If you decide for a Phenom II 955 or 965 be aware that you will not get as much perfromance out of it in FSX as compared to a Core i7 system but still you could be pretty satisfied with it.Measuring by Ghz is a myth to the OP.Even a slower 2.66 Ghz Core i7 920 will perform better than the 3.2 Ghz Phenom II X4 955 in most benchmarks.Not sure but I think there was one encryption benchmark in which Phenom II did much better than the Core i7 though but that was one exception.
By the way one company is planning on a Microsoft FSX replacement in which they expect to release in 2012.From the discussion threads it looks like a lot of money,time and development will go into it.I hope that they succeed.
http://www.forum.aerosoft.com/index.php?showtopic=29444

 

[ElMoIsEviL]

So then you are telling me to just forget that it really isn't running at 2.66Ghz and just drink the fanboy-aid and accept 2.66Ghz as being what it runs at.

Sure. Okay. I'll pretend I'm completely stupid and completely agree with you. Or perhaps I might actually know better.

Also: the new AMD design doesn't really have SMT does it? It has multiple clusters. That's called CMT and is a completely different design which apparently you didn't know or don't understand. So we'll just call it SMT and pretend it is the same thing so that you can pretend that SMT is not just a bit of a gimmick even though it is.

Luckily the CMT can be disabled if needed. But then it won't be suffering from the same problems as SMT will it?

But then we already know that a lot of Intel fanboys will be pretending it is the same thing just so they can claim that AMD "changed their mind". Let's review: one core pretending to be two cores versus two real cores using the same pipeline and processing units and calling itself one cluster or module. Yeah... I can see how somebody that was clueless might actually not be able to understand that it is not the same.

So let me get this straight.

I poor a glass of Coke and I add more sugar in it and I can resell it as something else than Coke? What if I buy a GeForce GTX 285 and change the heat sink... can I market it as a different card?

Changing the name doesn't change what it is at the heart of it. One would be Coke with more Sugar and the second example would be a GeForce GTX 280 OC or Super Cooled or whatever additional cool marketing buzz word you want to add.

What AMD have done is simply improve SMT by doubling more aspects of the execution engine. If AMD calls it CMT, good for them, but it's still SMT (one could argue improved). In essence one could argue that "when compared against large monolithic SMT processors, a CMT processor provides very competitive IPC performance on average, 90-96% of that of partitioned SMT
while being more scalable and much more power efficient."

What this means is that CMT isn't as good as SMT but is far more energy efficient and obtains nearly the same level of performance.

Don't take my word for it: http://www.csl.cornell.edu/~albonesi/research/papers/ispass05_1.pdf <--- Research paper on SMT vs. CMT.

SMT:
RED = Execution Engines
BLUE = L1 Cache

CMT:
RED = Execution Engines
BLUE = L1 Cache

 

[ryohei]

AMD sells great sub-100 procs, but Intel definitely owns the 200+ market. for fsx, a quad core highly recommended in order to get the sliders all the way to the right

 

[JackNaylorPE]

Define Best.

If best is "best performance".....Intel's best beats AMD's best.

If best is best value ...... we could be here all day as we move up and down the various price points. In short, AMD has an edge at the lower price ranges, Intel at the higher.

 

[saint19]

This is a eternal question that never could be answered. Why? Intel is better is some aspects and AMD is better is others.

 

[mildiner86]

I think u need to change ur title!!!

to maybe whats the best chip i can buy on my budget for this task.

i dont care if my chip is intel ar AMD i care about its performance and as many have said intel and amd both have the best chips at differnt price points.

and yeah kinda stop getting hung up on the Ghz too, gone are the days of the pentium 4.

 

[kingneptune117]

thanks for all the replies guys!

The AMD Phenom II 955 is a quad core though, just like most other intels I saw. The other intel quad cores were WAY more expensive though.

I thought that you need to look at mainly the number of cores and the Ghz. What else should I take into consideration.

Also, ive been hearing a lot of rumors that FSX is single threaded...only uses one core. Is this true?

Also, if i were to get the i7, which one do i get? There are a bunch of different ones with a bunch of different speeds.

 

[cjl]

FSX is multithreaded and uses as many cores as you can give it.

As for what you should look for, GHz matters, but it isn't the only thing that determines how fast a CPU is. How efficient the CPU is at using each cycle matters a lot too. The Intel chips are more efficient than the AMD ones right now, so the slower clocked Intel chips give the same performance as higher clocked AMD ones.

As for what you should get? It depends on your budget. For around ~$200, get either the Phenom II 955 or 965, or the Intel Core i5 750. For most purposes, they will all perform about the same, with the i5 winning sometimes, and the Phenom winning at others. For a slightly higher budget, I'd go with the Core i7 860. It's a good balance of price, performance, and the extra threads and high turbo modes will give it a significant boost over the Phenoms and the i5 750 if you don't plan to overclock.

 

[hunter315]

You cannot compare simply based on number of cores and GHz between architectures, yes it is a fair comparison in the core 2 line as they all perform approximately the same clock per clock, however the nehalem architecture the new intel chips are based off of performs far more IPC's than the core 2 did. Comparing to AMD opens up a whole new can of worms too because of the many different lineups of quads AMD offers. The older phenom series was slow, and hot while the new ones are faster, cooler, and get more IPC's done so they come out significantly faster in the end.

The best way to pick a processor is to look at benchmarks for applications you use. If you were to go for an i7 the 860 is your best choice, its over all platform cost is significantly lower than the i7 920's and will let you put more into your GPU.

FSX is far from single threaded, one of its recent updates allows it to use up to 256 cores.

 

[kingneptune117]

Wow quick responses! Thanks.

Sorry if i sounds noobish, that is what i am. I just two days ago got into all the processor stuff. Im starting to understand it now.

AND WOW @ THE i7 860 http://www.youtube.com/watch?v=QkjLlxBEqgY

 

[keithlm]

So let me get this straight.

<Snipped rant attempting to claim CMT is just another name for SMT>

Go ahead and pretend they are the same thing with a different name. That won't change the fact that they are not the same.

But hey... don't take my word for it... keep reading the document that YOU conveniently linked. It invalidates your ranting: (I love it when people are clever and post links that actually disprove what they arguing.)

The CMT is much more realizable, more energy efficient, and is likely to have either a higher clock or lower internal latencies. The CMPs of more realistic P-SMT and SMT configurations fall far short of the CMT performance. Thus, for future chips supporting four and eight thread workloads, a uniprocessor CMT or a dual processor of simpler CMTs are very attractive options.

Anyway... as mentioned before... Intel has SMT which uses one core and pretends that it is two. AMD's CMT apparently will be using two "mini-cores" that use common core resources to make a single "real core". This was verified in a slide recently presented by AMD. I don't have the link readily avaiable, but I can easily get it if you are actually curious. (Which is doubtful since apparently you just want to argue.) SO... they are completely different hardware designs. A quad core CPU will have 8 physical "mini-cores" and not 4 "real" cores and 4 "fake" cores like Intel's SMT design.

What will be even more interesting to see is how tightly AMD links the OpenCL to the CMT architecture. If the "mini-cores" of CMT can communicate status information independently of each other then OpenCL will be MUCH more efficient with CMT than it could ever be with SMT. In fact SMT could actually be detrimental. Which would be bad because then Intel will actively discourage OpenCL usage.

==============

BUT regardless, none of this changes the basic fact that having either SMT and CMT will not really help much in the average workloads that most people perform. They both will really only skew some benchmark results and foolish people that don't have a clue will believe whatever the most popular benchmarks sites claim.

 

[kingneptune117]

Basically Phenom II like the 955/965 will roughly be on par with the Core 2 Quad Kentsfield so it would be like having an already oc'd Q6600.It's not quite on par with Core 2 Quad Penryn's though in terms of IPC.
That's not to say that it's bad.
Phenom II will run FSX with less stuttering though (more smoothly) than the Q6600.
A year and a half ago many people ran oc'd Q6600 CPU's to run FSX until Core i7 came out.
Rather one has to decide on whether to run FSX in higher resolutions with more eye candy in the simulation in which Core i7 will do much better and the Bloomfield Core i7's will oc to a much higher degree than Phenom II by around 50% as compared with the max oc of Phenom II of 20%.
To run FSX with high levels of detail one wants to oc as much as possible to obtain it.
However AMD will have their 6 core Thuban CPU out next year as an upgrade for the AM3 platform and Intel will have their uber expensive Gulftown Core i9 as an upgrade for LGA 1366 so one could also take that into a possible consideration.
It really depends on your budget.If you decide for a Phenom II 955 or 965 be aware that you will not get as much perfromance out of it in FSX as compared to a Core i7 system but still you could be pretty satisfied with it.Measuring by Ghz is a myth to the OP.Even a slower 2.66 Ghz Core i7 920 will perform better than the 3.2 Ghz Phenom II X4 955 in most benchmarks.Not sure but I think there was one encryption benchmark in which Phenom II did much better than the Core i7 though but that was one exception.
By the way one company is planning on a Microsoft FSX replacement in which they expect to release in 2012.From the discussion threads it looks like a lot of money,time and development will go into it.I hope that they succeed.
http://www.forum.aerosoft.com/index.php?showtopic=29444

Ya man good to see someone else active in the aerosoft forums. This sim is gonna kick ass.

Anyways i think im gonna go with the i7-860. Sound good?

 

[randomizer]

SMT and CMT are only gimmicks if you plan to play poorly threaded games, although I can't speak for well threaded ones. Calling them gimmicks when doing real work is foolish, because they do improve performance, and synthetic benchmarks are not needed to show this.

In any case, the "benchmark" in question here is FSX, so arguing over the how much of a "gimmick" these functions are or aren't isn't important to the OP unless you can show the data for FSX to prove it.

 

[keithlm]

SMT and CMT are only gimmicks if you plan to play poorly threaded games, although I can't speak for well threaded ones. Calling them gimmicks when doing real work is foolish, because they do improve performance, and synthetic benchmarks are not needed to show this.

In any case, the "benchmark" in question here is FSX, so arguing over the how much of a "gimmick" these functions are or aren't isn't important to the OP unless you can show the data for FSX to prove it.

What you said: "Poorly threaded games" actually applies to all applications.

And what you said is backwards: SMT and SMT would not be gimmicks with poorly threaded games/application. If an application is poorly or loosely threaded then you can probably get a performance increase with SMT/CMT. (Sadly this happens way too much even in professional applications.)

However the converse is also true: if an application is tightly optimized to use an exact number of REAL cores... then SMT won't gain any benefit even if you tell the application that the "fake" cores are "real". i.e., You can't get more than 100% out of a core that is pegged at 100%.

In that situation CMT might still get a performance increase in this situation by counting each "mini-core"; but I've not actually worked on a CMT system so I can't say from first hand experience. Which is why if somebody handed me a chip with CMT on it... it would immediately have CMT disabled until more actual data is available.

 

[ElMoIsEviL]

Go ahead and pretend they are the same thing with a different name. That won't change the fact that they are not the same.

But hey... don't take my word for it... keep reading the document that YOU conveniently linked. It invalidates your ranting: (I love it when people are clever and post links that actually disprove what they arguing.)

The CMT is much more realizable, more energy efficient, and is likely to have either a higher clock or lower internal latencies. The CMPs of more realistic P-SMT and SMT configurations fall far short of the CMT performance. Thus, for future chips supporting four and eight thread workloads, a uniprocessor CMT or a dual processor of simpler CMTs are very attractive options.

Anyway... as mentioned before... Intel has SMT which uses one core and pretends that it is two. AMD's CMT apparently will be using two "mini-cores" that use common core resources to make a single "real core". This was verified in a slide recently presented by AMD. I don't have the link readily avaiable, but I can easily get it if you are actually curious. (Which is doubtful since apparently you just want to argue.) SO... they are completely different hardware designs. A quad core CPU will have 8 physical "mini-cores" and not 4 "real" cores and 4 "fake" cores like Intel's SMT design.

What will be even more interesting to see is how tightly AMD links the OpenCL to the CMT architecture. If the "mini-cores" of CMT can communicate status information independently of each other then OpenCL will be MUCH more efficient with CMT than it could ever be with SMT. In fact SMT could actually be detrimental. Which would be bad because then Intel will actively discourage OpenCL usage.

==============

BUT regardless, none of this changes the basic fact that having either SMT and CMT will not really help much in the average workloads that most people perform. They both will really only skew some benchmark results and foolish people that don't have a clue will believe whatever the most popular benchmarks sites claim.

Holy crap!

You didn't understand a single thing from that research paper. Why do you keep commenting on things you obviously know nothing about? Seriously... you're mis-informing people and that is just not cool. I am not Jumping Jack.. I don't have the patience to deal with people like you.

CMT is not two mini-cores and SMT does not "pretend" to be two cores. Both are only a single core doing more work per clk by utilizing unused execution resources. Take SMT for example. SMT works by using the idling Execution engine while an already executed workload is making it's way through the pipeline (but is still being processed). So in essence you're executing TWO threads simultaneously. That's an explanation in laymen terms. If you can't understand that then by all means please stop commenting on my posts.

SMT:

What CMT does differently is that certain aspects of the execution engine are doubled. Those aspects are highlighted in those pictures. So CMT can execute two threads in parallel (not in serial as SMT does). It's perfectly suited for an architecture whose pipeline is not as deep, like AMDs K8/K10 based architectures for instance or in the case of it's intended usage... Bulldozer.

Those two "mini-cores" you keep referring too aren't cores.. they're execution engines. I'll explain how it works again by using those pictures (I figured a picture would be enough I guess I was wrong):

This is SMT Notice it only has a single L1DCache and a single Execution engine (I circled both):

This is CMT Notice it has TWO L1DCaches and TWO Execution engines working in parallel (I circled both):

And what the hell do you mean by "AMD links the OpenCL to the CMT architecture"? OpenCL is a programming language that is meant to create a cross platform set of tools for various compute devices. OpenCL cannot be "linked" to a "CMT architecture" that doesn't even make sense. Since CMT is simply SMT but with parallel execution engines and doubled L1D Caches (yes it executes two threads but rather than doing so in serial fashion it does so in parallel fashion so it is the same general idea as SMT) then CMT, like SMT will appear to Windows as being TWO Cores (One Physical and the other Logical). Just like SMT, any application written to take advantage of multiple threads will work with CMT (the O/S Kernel controls the processor Affinity).

Stop commenting on things you know nothing about. And if what you're doing is quoting Scentia or Abinstein.. you're doing yourself a LARGE disservice.

 

[croc]

Go ahead and pretend they are the same thing with a different name. That won't change the fact that they are not the same.

But hey... don't take my word for it... keep reading the document that YOU conveniently linked. It invalidates your ranting: (I love it when people are clever and post links that actually disprove what they arguing.)

The CMT is much more realizable, more energy efficient, and is likely to have either a higher clock or lower internal latencies. The CMPs of more realistic P-SMT and SMT configurations fall far short of the CMT performance. Thus, for future chips supporting four and eight thread workloads, a uniprocessor CMT or a dual processor of simpler CMTs are very attractive options.

Anyway... as mentioned before... Intel has SMT which uses one core and pretends that it is two. AMD's CMT apparently will be using two "mini-cores" that use common core resources to make a single "real core". This was verified in a slide recently presented by AMD. I don't have the link readily avaiable, but I can easily get it if you are actually curious. (Which is doubtful since apparently you just want to argue.) SO... they are completely different hardware designs. A quad core CPU will have 8 physical "mini-cores" and not 4 "real" cores and 4 "fake" cores like Intel's SMT design.

What will be even more interesting to see is how tightly AMD links the OpenCL to the CMT architecture. If the "mini-cores" of CMT can communicate status information independently of each other then OpenCL will be MUCH more efficient with CMT than it could ever be with SMT. In fact SMT could actually be detrimental. Which would be bad because then Intel will actively discourage OpenCL usage.

==============

BUT regardless, none of this changes the basic fact that having either SMT and CMT will not really help much in the average workloads that most people perform. They both will really only skew some benchmark results and foolish people that don't have a clue will believe whatever the most popular benchmarks sites claim.

Could we stay on topic, please? For FSX, what processor will give the best results.