• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

AM2 memory controller Vs C2D memory controller

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

pak

Member
Joined
Mar 21, 2005
Location
Tracy, ca
Now I know C2D will eat Am2 up all day, but please. That is not what this thread is about.

My OC'n experience has been all AMD. I think I have a fairly good understanding of how it works on the AMD side. This is assuming 939 is the same/similar to the way AM2 works.

Before I take my step in to DDR2, I want to know how intel works. I have seen alot of new things(words phrases) in regaurds to C2d OC'n in regaurds to memory. One example being *interleaving*(I think thats the word I am looking for. Something about intel's taking advantage of using 4 memory sticks vs 2 where AMD doesnt like to have more than 2.

Teach me

--pak
 
The 2 sticks VS 4 sticks has to deal with running at 1t command rate or 2t command rate. I don't know if Intels can run 1t any better than AMD does or not. However AM2 seems to like high speeds and 2t better so its really a non-issue anymore.

Interleaving can be done with either platform and I believes it acts something like raid 0 does for HDs.

Really Intel is simpler to overclock(atleast IMHO) than AMD becuase you don't have to play with 3 million memory parameters if you don't want to(unless you decide you need an Nvidia chipset then it might get complicated.).
 
speed bump said:
Really Intel is simpler to overclock(atleast IMHO) than AMD becuase you don't have to play with 3 million memory parameters if you don't want to(unless you decide you need an Nvidia chipset then it might get complicated.).
I wouldn't consider complete control on the memory timings making OC'ing any more complicated.
On the contrary.

Also Hammer does actually have 100% lossless dividers which isn't the case with Intel. That's if your Intel MCH actually supports dividers.

I dunno what exactly is the deal with upward dividers anyways when there's FSB on Intels which practically neuters any gains from memory speeds higher-than-FSB...
 
I recall hearing that AM2 kicks *** with 4 sticks (4 bank interleaving) no knowledge of Intels usage of it.

Most OCers wont run 4 sticks though as you increase the odds of getting a bum clocker
 
Last time I checked, On-Dye memory controllers were generally more efficent and much more responsive. ;)
C2D May be a fast chip, but C2Q is starved of memory thanks to Intel's 15(+)-year-old design.
Better dividers, lower latency, less overhead, and a shorter distance for data to travel makes a happy processor in the AM2 socket.
And what Neur0mancer said.

Rant: And WTH is this LGA thing? Broken pins on the mainboard bring me a lot more hell then the odd bent one on the processor... Intel proably just went for LGA just because it's cheaper, and they don't take the blame for bent/broken pins.
 
Shell said:
Rant: And WTH is this LGA thing? Broken pins on the mainboard bring me a lot more hell then the odd bent one on the processor... Intel proably just went for LGA just because it's cheaper, and they don't take the blame for bent/broken pins.
Don't make a mistake by thinking Intel nor AMD cares one bit about enthusiasts struggling with the pins on their mobos.

LGA simply offers a superior signal/electrical interface. LGA would, and, does work better than PGA for the 99.9% of people that use computers. What I don't understand is why AMD didn't go LGA with AM2, the new socket has no ties to earlier sockets so compatibility was not the point.
Hence AMD is stuck on the inferior legacy PGA until AM3 becomes obsolete in early 09.
 
This is all really new info to me. Please share more! Anyone know anyone who has ran 4 sticks on an AM2? In an ideal situation how much more performance can you gain by running 4 sticks of memory?

And another question I would like to throw out there. Lets "try" to compare two similar CPU's. One being am2 and one being C2d. Using the exact same memory, 2 and/or 4 sticks, who wins the memory bandwidth fight?

--pak
 
pak said:
This is all really new info to me. Please share more! Anyone know anyone who has ran 4 sticks on an AM2? In an ideal situation how much more performance can you gain by running 4 sticks of memory?

And another question I would like to throw out there. Lets "try" to compare two similar CPU's. One being am2 and one being C2d. Using the exact same memory, 2 and/or 4 sticks, who wins the memory bandwidth fight?

--pak


amd will winn the bandwidth fight probably... because they have on die memory controler and 2000mhz link between the cpu and ram... the path of the memory on am2 goes directly between the cpu and ram, on a intel it goes from the cpu to the northbridge via the FSB, then NB to ram, via the ram buss.
 
nd4spdbh2 said:
amd will winn the bandwidth fight probably... because they have on die memory controler and 2000mhz link between the cpu and ram... the path of the memory on am2 goes directly between the cpu and ram, on a intel it goes from the cpu to the northbridge via the FSB, then NB to ram, via the ram buss.
Yes this is true but the fact that C2D can beat the AM2 in most all benchmarks that require heavy bandwidth from CPU to RAM (like SuperPie and others) seems to me that they shortened up some thing to accommodate this . Why are conroes getting 13 seconds and lower scores with the same RAM used in an AM2 system ? I just don't get it . :bang head
 
Lets keep this civil please ;). No one needs a flame war claiming one widget is supreme because of x pin and y year old design. The question at hand is of which is faster under the configurations the OP is asking for. Please share your experiences and not your little fireballs. Thank you ;).
 
trickson said:
Yes this is true but the fact that C2D can beat the AM2 in most all benchmarks that require heavy bandwidth from CPU to RAM (like SuperPie and others) seems to me that they shortened up some thing to accommodate this . Why are conroes getting 13 seconds and lower scores with the same RAM used in an AM2 system ? I just don't get it . :bang head

They didn't shorten anything, they simply have an interal cpu design that can achieve more with less clocks. Mem Bandwidth limitations can be outdone with a fast enough cpu, and thats what intel has. simple. To explain better, if intel had AMD's mem controller on the C2D its would be a million times better than it already is because it would have the bandwidth to back it up. AMD mem system is capable of more bandwidth than the cpu's can use, hence intel can get high scores.

As to the LGA/PGA debate, i don't think LGA is more superior because ultimately its all just electrical connections, nothing more. I think they need to design a truly pinless design (not pin on the cpu like PGA, and not pins in the mobo liek LGA). And to claim it makes a difference to people who use computers is silly, neither way dose. They achieve the same task in almost identical ways.
 
pak,
Using 4 dimms will gain you few hundred MB/s in bandwidth and a marginal performance bump. link
nd4spdbh2 said:
amd will winn the bandwidth fight probably... because they have on die memory controler and 2000mhz link between the cpu and ram...
Hypertransport bus - if that's what you'r referring to - has nothing to do with memory bus in AMD Hammer architecture. K8 memory controller is an integral part of the CPU and that is operating at CPU frequency, but the memory controller is connected to RAM with the exact same frequency the RAM is running at.
trickson said:
C2D can beat the AM2 in most all benchmarks that require heavy bandwidth from CPU to RAM (like SuperPie and others) seems to me that they shortened up some thing to accommodate this . Why are conroes getting 13 seconds and lower scores with the same RAM used in an AM2 system ? I just don't get it.
Memory bandwidth is only one factor that affects performance. There are many, many other factors too.

Conroe is getting such low SPi times because the architecture is wider and smarter (= newer) than K8. Conroe has wider L1/2 cache pathways, instruction decoding and tons of others features not found on K8. K8 is not bandwidth bottlenecked, it's limited by clockspeed and architectural issues (instruction fetch, etc.). That's why even dual channel doesn't yield that much to system performance for K8.
Hipcrostino said:
Mem Bandwidth limitations can be outdone with a fast enough cpu, and thats what intel has. To explain better, if intel had AMD's mem controller on the C2D its would be a million times better than it already is because it would have the bandwidth to back it up.
Like AMD K8, Conroe is not bandwidth bottlenecked. link. C2D would be better with an IMC. ..That's granted, but the gain wouldn't exactly be stellar because bandwidth isn't a problem for Core.
AMD mem system is capable of more bandwidth than the cpu's can use, hence intel can get high scores.
That doesn't make any sense.
i don't think LGA is more superior because ultimately its all just electrical connections, nothing more. And to claim it makes a difference to people who use computers is silly, neither way dose. They achieve the same task in almost identical ways.
LGA offers larger contact area and higher contact density compared to PGA, thus allows lower contact resistance yielding more efficient and stabler power distribution interface.

Ever dismantled a PGA socket? The method pins are bonded with the socket is much more complicated than LGA, PGA actually require 2 pins inside the socket to connect a single pin of the CPU. More solder contacts automatically means more signal distortion -> worse for high frequency.
This is what the insides of a PGA look like. LGA is much simpler.
 

Attachments

  • IMG_2650_rc.jpg
    IMG_2650_rc.jpg
    35.7 KB · Views: 151
  • IMG_3563_rc.jpg
    IMG_3563_rc.jpg
    37.5 KB · Views: 167
oooo man if intel put an on die memory controler on the C2D's O MY GOD that would be AMAZING!... heck throw in some reverse hyperthreading and we got ourselves a deal :beer:
 
Back