AMD = to What Intel

i fell like we are getting closer to an agreement.

we now agree on the intergraded gpu and about the pins. (still nothing agenst pins. they work as thay should)

ok.... i am not sure if they are using intergraded memory controller, what i meen is thet controller that used to sit on the motherboard.
you might be right..... But then i just dont understad why thay still have a speed limitation on the motherboard then. the FSB.
when intel got rid of the mem controler on the motherboard, thay also got rid of the FSB. can you explane that?
i am just asking corz i seriosly dont know.
Intel got rid of the traditional front side bus, but there are still parts of the system that serve the same function. They just moved most of it to the CPU die. AMD also replaced most of the FSB (with Hypertransport..more on that later), but they still have a more-or-less traditional north bridge/south bridge setup. However, in both AMD and Intel systems, the memory controller isn't attached to the HT/QPI at all. Both companies' processors have a dedicated memory controller in the CPU itself. In other words, the CPU is linked directly to the RAM. Motherboards have limitations because there's still a limit in the hardware somewhere, but it has nothign to do with the FSB.

i just saw a damm cheep amd bord with an FSB at 2700 mhz... that suprised me. that is kinda fast for FSB and for that lov prise.
but Intels speed via the bord is QPI. an intel core i7 920 has i QPI at 6,4 GT/s. thats 3,2 ghz. and when you overclock it, it goes up.
THATS why it confuses me that you are saying that amd has intergraded mem controller.
No, AMD's default base clock (FSB speed) is always 200MHz, and Intel's I believe is 133MHz up until Sandy Bridge, which drops it to 100MHz. Doesn't matter on sandy bridge anyway though, since you can't change it without screwing stuff up.

What you were probably looking at is Hypertransport speed. Hypertransport, as used by AMD, is more or less what Intel was trying to emulate when the designed QPI. (Hypertransport isn't an AMD-only technology)

Default HT speed for AMD systems varies by what CPU and socket you have, but I think their highest so far is 2000MHz on AM3. However, current Hypertransport technology is capable of up to 3.2GHz. It isn't directly comparable with QPI, though.

Also, HT speed does go up when you overclock the system with the base clock, but not if you do it with the CPU multiplier. Same with RAM speed.



the remaining this is the thing that startet it all.
BENCHMARKS........

cpu benshmarks are different from gpu benchmarks. games ARE optimized to ether ati(amd) or nvidia. also some 3d gpu rendering benchmarks are optimized for amd or nvidia.
like "Unigine Heaven" is optimized for nvidia. "Unigine Heaven" is a benchmark that you should NOT count on. it should be banded.
but you can always count on 3dmark.

anyway... the cpu benchmarks in the other hand is not optimized.
i can't see how you can optimize a non driver needed hardware.
so what you are saying is that they optimize there cpu's for better real life tasks...... nice.....

so is it optimized for like winrar. awesome... i use winrar alot.
is it cinebench optimized.... nice.. i have photoshop cs3... cpu rendering.
about games.... i like games.

cpu's is NOT optimized.
AMD and Intel do not have perfectly identical instruction sets. It's absolutely possible to optimize for one or the other. Intel's current architecture is more efficient clock-for-clock than AMD's though...no argument there.


about your amd 6 core..... i dont need to find a 6 core intel to match up with it.
what about a 4 core core i5.

AMD's Six-Core Phenom II X6 1090T & 1055T Reviewed - AnandTech :: Your Source for Hardware Analysis and News
(just go to some of the benchmarking test in the review.)

i know that in some of the benchmarks core i5 gets its ass handed to it, but thing about that the amd 1090t has a higher mhz (500 mhz) then core i5 AND 2 more cores. total unfare fight.
It's NOT an unfair fight. They're nearly the same price, and as such are direct competitors to one another.

you could also just have a look at the core i7 920 and 860. those are also 4 cores (8 threads) with still 400-500 lower mhz then the amd 1090t. those are really giving the amd 1090t a challange.
this is what i meen about that intel mhz is worth more then amd mhz.
this should be a total unfare fight. 4 core vs 6, lower mhz vs higher. amd has avantage but still looses in most benchmarks.
Again, your percieved advantage is negated by the fact that they're the same price. Doesn't matter what the specs are. Performnce per dollar is what matters in computing. And if you look, in the benchmarks that actually use all of the cores, the x6's are up to 20% more powerful than the i5 and about on par with the i7's, which have eight threads and, key point here, are $100 more. Given the proper workload, you pay 2/3 as much for the AMD and get the same performance. The gap is even wider when you consider the 1055T there.

The Phenom II x6 chips definitely need a heavily multithreaded workload to shine, but when they do, it's pretty obvious that they kick arse compared to the similarly-priced 1st gen Nehalem chip. Pretty good for an architecture designed to combat Intel's old Core 2 architecture.

None of them can hold a candle to the ~$200 Sandy Bridge offerings, though.
 
i feel like we are agreeing in alot of things here.
you are way better to chat with then OG is.

No, AMD's default base clock (FSB speed) is always 200MHz, and Intel's I believe is 133MHz up until Sandy Bridge, which drops it to 100MHz. Doesn't matter on sandy bridge anyway though, since you can't change it without screwing stuff up.
What you were probably looking at is Hypertransport speed. Hypertransport, as used by AMD, is more or less what Intel was trying to emulate when the designed QPI. (Hypertransport isn't an AMD-only technology)
Default HT speed for AMD systems varies by what CPU and socket you have, but I think their highest so far is 2000MHz on AM3.

However, current Hypertransport technology is capable of up to 3.2GHz. It isn't directly comparable with QPI, though.

Also, HT speed does go up when you overclock the system with the base clock, but not if you do it with the CPU multiplier. Same with RAM speed.

yeah its a new thing sandy bridge is doing. you overclock via the multiplyer. but the multiplyer is unlocked in most sandy bridge cpu's. i think it takes some of the fun out in overclocking though. -.-
but it takes less pressure on the rest of the system by using multiplyer. thats why core i7 extreme gets i way higher max overclock. (also becouse of the modded architecture "Westmere")

yeah i have seen a guy get the HT at 2600 mhz (thats was the limit an that board). pretty cool. still not 3200 is is the core i7 standard. as you where saying "HT isn't directly comparable with QPI, though"
but lets see if amd does'ent strike back some day. technology allways changes.


AMD and Intel do not have perfectly identical instruction sets. It's absolutely possible to optimize for one or the other. Intel's current architecture is more efficient clock-for-clock than AMD's though...no argument there.
i see. but what i ment was more like, that you cant optimize to a specific testing software the same way nvidia did with dx11 unigine heaven benchmark. that was just cheating.
but i understand the optimize clock-for-clock. same thing with watt-for-clock. and there is more



It's NOT an unfair fight. They're nearly the same price, and as such are direct competitors to one another.
yeah.... i know. my mistake.
it makes perfect sense that it should be all about the prise range.


Again, your percieved advantage is negated by the fact that they're the same price. Doesn't matter what the specs are. Performnce per dollar is what matters in computing. And if you look, in the benchmarks that actually use all of the cores, the x6's are up to 20% more powerful than the i5 and about on par with the i7's, which have eight threads and, key point here, are $100 more. Given the proper workload, you pay 2/3 as much for the AMD and get the same performance. The gap is even wider when you consider the 1055T there.
i did never say that it is all about the specs. becaus it is NOT(a realy inportent factor is also the architecture and more).
it was more to demonstrate that the core i5 according to the specs should complitly loose. but it did not loose in all benchmarks.
i see that the 6x owne in more heavy multi-threaded benchmarks. but the 4x hyperthreaded core i7 beets the x6 core in most of the heavy multi-threaded benchmarks.

but here is where the x6 realy shines. as you where saying: it can compare to the core i7 for the 2/3 the money. now thats something. AMD IS cheeper, and that a realy important fact when you are buyin a computer.


The Phenom II x6 chips definitely need a heavily multithreaded workload to shine, but when they do, it's pretty obvious that they kick arse compared to the similarly-priced 1st gen Nehalem chip. Pretty good for an architecture designed to combat Intel's old Core 2 architecture.
yeah respect to AMD for winning the prise category in that serie. (they use to win that category)

None of them can hold a candle to the ~$200 Sandy Bridge offerings, though
yeah. but the sandy bridge requres a new motherboard.
that is what intel is knowen for. out with the old and in with the new.
some people like that and others just hate intel for doing so.


i hope you can see that i am not an intel fan boy. i just like it better. you like amd better.
OG just cant understand that. i hope you can. i was thinking about getting an amd radeon card, even though that i am more of an nvidia fan.
i am getting the gtx 570. but i did really considered the AMD Radeon 6950. i could be bios flashed to an 6970.
 
yeah. but the sandy bridge requres a new motherboard.
that is what intel is knowen for. out with the old and in with the new.
some people like that and others just hate intel for doing so.

Again, I'll say that it used to be the other way. It's what turned me away from AMD a while ago. THe simple fact is as processor technology changes one socket will not always work. Intel used it's 775 line for years and now AMD has done the same with the AM2+/AM3 socket. I'm sure in five years we'll see the pattern has flipped again.
 
Again, I'll say that it used to be the other way. It's what turned me away from AMD a while ago. THe simple fact is as processor technology changes one socket will not always work. Intel used it's 775 line for years and now AMD has done the same with the AM2+/AM3 socket. I'm sure in five years we'll see the pattern has flipped again.

hmm. ok i did not know. i got into computers kinda late. less than 3 years ago. so thats before my time.
btw: disturbed rules.
 
i feel like we are agreeing in alot of things here.
you are way better to chat with then OG is.



yeah its a new thing sandy bridge is doing. you overclock via the multiplyer. but the multiplyer is unlocked in most sandy bridge cpu's. i think it takes some of the fun out in overclocking though. -.-
but it takes less pressure on the rest of the system by using multiplyer. thats why core i7 extreme gets i way higher max overclock. (also becouse of the modded architecture "Westmere")
lol :)

Actually, overclocking with the multiplier has been around for quite some time. All of AMD's "Black Edition" processors, and Intel's "Extreme Edition" ones, have an unlocked multiplier. The difference with Sandy Bridge is that you can only overclock with the multiplier.

yeah i have seen a guy get the HT at 2600 mhz (thats was the limit an that board). pretty cool. still not 3200 is is the core i7 standard. as you where saying "HT isn't directly comparable with QPI, though"
but lets see if amd does'ent strike back some day. technology allways changes.
I don't see how they can really "strike back"...Hypertransport and QPI are different technologies. They do serve a similar function, but they don't do exactly the same thing.




i see. but what i ment was more like, that you cant optimize to a specific testing software the same way nvidia did with dx11 unigine heaven benchmark. that was just cheating.
but i understand the optimize clock-for-clock. same thing with watt-for-clock. and there is more
I believe what happened with Nvidia was that they tweaked their drivers, yeah?

What I'm talking about with CPU's is the software itself being coded to use specific features that only one brand has,. For example, Sandy Bridge has AVX, while AMD won't feature it until Bulldozer comes out.


yeah respect to AMD for winning the prise category in that serie. (they use to win that category)
They still do, most of the time. They have to...otherwise they won't stay competitive. The underdog always has to have lower prices to stay afloat.


i hope you can see that i am not an intel fan boy. i just like it better. you like amd better.
OG just cant understand that. i hope you can. i was thinking about getting an amd radeon card, even though that i am more of an nvidia fan.
i am getting the gtx 570. but i did really considered the AMD Radeon 6950. i could be bios flashed to an 6970.
I'm not a fanboy either...the parts in my rig just happened to be the best performance for the price when I bought them. Given the money I would have bought an i5 750 and a GTX 460, which at the time had about the best bang for the buck you could get anywhere. That would have added a good $200 or more, though.
 
Actually, overclocking with the multiplier has been around for quite some time. All of AMD's "Black Edition" processors, and Intel's "Extreme Edition" ones, have an unlocked multiplier. The difference with Sandy Bridge is that you can only overclock with the multiplier.

ohh.... i DID know about that, sorry my english is not that great.
what i was ment to say was that it was a new thing that they ONLY overclock via the multiplyer and had a base clock at 100. i have overclocked my baseclock from 133 to 170. so i know about it.

I don't see how they can really "strike back"...Hypertransport and QPI are different technologies. They do serve a similar function, but they don't do exactly the same thing.

again my english might be a bit sluppy.
what i ment here was amd might do something to compite with QPI or DMI in a neer future. in one of my later post i was in a bad mood and that had effect on the post, so i might have it sound like that i think amd has done nothing right.
so thats why i was trying to give amd some credit in that post.


I believe what happened with Nvidia was that they tweaked their drivers, yeah?

yes. as i was telling OG, nvidia tweaked the driver to get extra score in that dx11 benchmark

What I'm talking about with CPU's is the software itself being coded to use specific features that only one brand has,. For example, Sandy Bridge has AVX, while AMD won't feature it until Bulldozer comes out.

yeah thats the thing OG don't understand. i did tell him that you can't optimize a cpu(non driver hardware) to one specefic benchmark.
this was what started this arguing

ofcoars (as you where saying) is it possible to optimize testing software to test exacliy what the one brand are using.
but try asking 3dmark if they want to do that. i dont think they will. also cinebench are a fair testing tool as well.
most of them is. because if thay are not, the testers that are giving money to fairly test hardware will tell in there review that the test is optimized.
they do that with nvidia and the dx11 unigen heaven benchmark.


They still do, most of the time. They have to...otherwise they won't stay competitive. The underdog always has to have lower prices to stay afloat.

a like that you are saying it like it is. thay are cheeper. and thats what is so awesome about AMD.
 
EchoNatek said:
yeah thats the thing OG don't understand. i did tell him that you can't optimize a cpu(non driver hardware) to one specefic benchmark.
this was what started this arguing

I understand things plenty. I know how the industry works, I know how different CPU architectures work. If you want to try and convince me otherwise, good luck. I use both platforms, I know each one intimately, and I know what I'm talking about.

I have been doing this for damn near two decades. No, you can't write drivers for CPUs. Yes, you can write software to perform a specific way in a given task. Software gives you the ability to make the hardware do anything you want. There are all kinds of scandals over the years where nVidia or ATI (at the time) were being accused of fabricating their drivers to make results appear differently. I was also pointing to the fact that AMD and Intel will always show their CPUs in a positive light, skewing their results to how they think consumers should see them. AMD lately has been very truthful on their benchmark charts (well, the last time I looked at their site that is, about a month ago) and they show Intel on top in some things, while their stuff performs just a little better in others.

My point was never about software being written FOR a CPU in the way that a driver is written for a GPU. If you make a parallel there, I don't know why, but I apologize if that's what you thought.

Also,
EchoNatek said:
what i ment here was amd might do something to compite with QPI or DMI in a neer future. in one of my later post i was in a bad mood and that had effect on the post, so i might have it sound like that i think amd has done nothing right.
so thats why i was trying to give amd some credit in that post.
AMD has no need to upgrade the HyperTransport protocol because it is already similar and in some ways superior to QPI. And that's kind of a misnomer anyway because HTT doesn't have one specific "type" as they keep upgrading it. The current iteration is HTT 3.0 Also, another thing worth noting is that Intel's QPI is proprietary, and developed under just Intel's IP. HyperTransport is an open standard, much like PCI Express is, and there are several companies that develop it. HTT as it is called in writing to differentiate it from HT (Hyperthreading) is used as a system bus interconnect technology for AMD CPUs (and several scientific applications) because of how fast it is, and it's flexibility.
 
Back
Top Bottom