[SnyperTodd]

waldojim, on 26 April 2011 - 02:27 PM, said:

For a laptop - no complaints!...

Nice scores! Especially HDD and RAM. That's an impressive laptop!

[waldojim]

SnyperTodd, on 26 April 2011 - 06:03 PM, said:

waldojim, on 26 April 2011 - 02:27 PM, said:

For a laptop - no complaints!...

Nice scores! Especially HDD and RAM. That's an impressive laptop!

Thanks! I am very impressed with how this machine behaves. Props to Sony for an excellent blend of performance, and props to Intel / AMD for such a wonderful blend of performance, power, and battery savings.

[SnyperTodd]

waldojim, on 26 April 2011 - 07:17 PM, said:

Thanks! I am very impressed with how this machine behaves. Props to Sony for an excellent blend of performance, and props to Intel / AMD for such a wonderful blend of performance, power, and battery savings.

That right there really shows the benefits of an integrated memory controller. You're running 1333MHz @ 9-9-9-24 and getting over 10K on the memory score in Novabench. I'm running 1900MHz @ 8-9-7-22 1t and getting almost 7K in Novabench. I'd like to see Everest's Cache & Memory Benchmark on your laptop, I bet it's a lot higher than 10K in reality. I'm over 10K for everything (read, write, copy) in that, which is most likely more realistic than Novabench's rudimentary benchmark. I wish Novabench measured memory latency, that would be really interesting to compare.

[waldojim]

SnyperTodd, on 27 April 2011 - 10:14 AM, said:

waldojim, on 26 April 2011 - 07:17 PM, said:

Thanks! I am very impressed with how this machine behaves. Props to Sony for an excellent blend of performance, and props to Intel / AMD for such a wonderful blend of performance, power, and battery savings.

That right there really shows the benefits of an integrated memory controller. You're running 1333MHz @ 9-9-9-24 and getting over 10K on the memory score in Novabench. I'm running 1900MHz @ 8-9-7-22 1t and getting almost 7K in Novabench. I'd like to see Everest's Cache & Memory Benchmark on your laptop, I bet it's a lot higher than 10K in reality. I'm over 10K for everything (read, write, copy) in that, which is most likely more realistic than Novabench's rudimentary benchmark. I wish Novabench measured memory latency, that would be really interesting to compare.

Have the results at home, will post tonight... work is killer right now.

[waldojim]

http://img717.imageshack.us/img717/3001/cachemems.png

http://img687.imageshack.us/img687/3528/harddriveread.png

[SnyperTodd]

Wow. My results are back a few pages, but at the speeds mentioned above my latency is still at least 10ns higher. Intel definitely did the right thing following AMD's lead in integrating the memory controller. I figured you would score better in Everest than Novabench. I'm convinced that it's time for me to build a whole new machine...

And I lol'd at your double Socket 370 processors. Time to update Everest...

[coastie65]

SnyperTodd, on 28 April 2011 - 06:30 AM, said:

Wow. My results are back a few pages, but at the speeds mentioned above my latency is still at least 10ns higher. Intel definitely did the right thing following AMD's lead in integrating the memory controller. I figured you would score better in Everest than Novabench. I'm convinced that it's time for me to build a whole new machine...

And I lol'd at your double Socket 370 processors. Time to update Everest...

Not so sure of that integrated memory controller. I think I liked it better the other way ( non integrated), as it gave you more flexibility as to the memory speed. As it is, they integrated it and then locked the speed to whatever. I realize that is not the case with the Sandy Bridge processors as those things are unlocked. Because of that, I suspect that everything between the LGA 775 based Processors and the Sandy Bridge will go by the wayside in time, at least in the sense of those who do not wish to build and still want some upgrade flexibility. I had plenty with the eMachines ( both processor and memory ), not so much with this Gateway as the only performance update options available were the graphics card and processor. The memory was locked at 1066 Mhz although the MOBO will support up to 1600 Mhz. I blame that primarily on Intel's rush to get the 1366 based Core i7 out and not enough testing to determine what could comfortably be run ( they had no idea and locked things down to what had been tested ). It was those who bought them and started tinkering around with them that found out what they could do and informed Intel that their assumptions were wrong. Intel has unlocked some 1366 based i7's, as well as some 1156 based stuff, to run the memory at 1333 Mhz. No prebuilts ( with the exception of the custom builds ) will run at 1600 Mhz to my knowledge. Anyway, I am no fan of the integrated Memory controller, but think the Sandy Bridge has taken a step in the right direction and will offer the flexibilty.

This post has been edited by coastie65: 29 April 2011 - 07:43 AM

[waldojim]

SnyperTodd, on 28 April 2011 - 06:30 AM, said:

Wow. My results are back a few pages, but at the speeds mentioned above my latency is still at least 10ns higher. Intel definitely did the right thing following AMD's lead in integrating the memory controller. I figured you would score better in Everest than Novabench. I'm convinced that it's time for me to build a whole new machine...

And I lol'd at your double Socket 370 processors. Time to update Everest...

Yeah, I guess it is...

[waldojim]

coastie65, on 28 April 2011 - 09:13 AM, said:

Not so sure of that integrated memory controller. I think I liked it better the other way ( non integrated), as it gave you more flexibility as to the memory speed. As it is, they integrated it and then locked the speed to whatever. I realize that is not the case with the Sandy Bridge processors as those things are unlocked. Because of that, I suspect that everything between the LGA 775 based Processors and the Sandy Bridge will go by the wayside in time, at least in the sense of those who do not wish to build and still want some upgrade flexibility. I had plenty with the eMachines ( both processor and memory ), not so much with this Gateway as the only performance update options available were the graphics card and processor. The memory was locked at 1066 Mhz although the MOBO will support up to 1600 Mhz. I blame that primarily on Intel's rush to get the 1366 based Core i7 out and not enough testing to determine what could comfortably be run ( they had no idea and locked things down to what had been tested ). It was those who bought them and started tinkering around with them that found out what thecould do and in formed Intel that their assumptions were wrong. Intel has unlocked some 1366 based i7's, as well as some 1156 based stuff, to run the memory at 1333 Mhz. No prebuilts ( with the exception of the custom builds ) will run at 1600 Mhz to my knowledge. Anyway, I am no fan of the integrated Memory controller, but think the Sandy Bridge has taken a step in the right direction and will offer the flexibilty.

Intel has some very good reasons for what it does sometimes. If you will recall, during the Pentium, P2, and P3 days, Intel would never use a 100Mhz base clock (as they do now). They claimed they didn't use it because it was unstable, there was no natural way to get to that clock. Now, I am not sure how true that is, but sometimes they are right on with these things. I would prefer slightly slower and stable over stupidly fast, and unstable.

With memory controllers though, I have to admit that I do like the external controllers, as that allows you to buy better boards with better controllers. At the same time, with it integrated, you don't have to worry about cheap boards having crap controllers either.

I have to admit- I do like the latency numbers.

[coastie65]

waldojim, on 28 April 2011 - 09:30 AM, said:

coastie65, on 28 April 2011 - 09:13 AM, said:

Not so sure of that integrated memory controller. I think I liked it better the other way ( non integrated), as it gave you more flexibility as to the memory speed. As it is, they integrated it and then locked the speed to whatever. I realize that is not the case with the Sandy Bridge processors as those things are unlocked. Because of that, I suspect that everything between the LGA 775 based Processors and the Sandy Bridge will go by the wayside in time, at least in the sense of those who do not wish to build and still want some upgrade flexibility. I had plenty with the eMachines ( both processor and memory ), not so much with this Gateway as the only performance update options available were the graphics card and processor. The memory was locked at 1066 Mhz although the MOBO will support up to 1600 Mhz. I blame that primarily on Intel's rush to get the 1366 based Core i7 out and not enough testing to determine what could comfortably be run ( they had no idea and locked things down to what had been tested ). It was those who bought them and started tinkering around with them that found out what thecould do and in formed Intel that their assumptions were wrong. Intel has unlocked some 1366 based i7's, as well as some 1156 based stuff, to run the memory at 1333 Mhz. No prebuilts ( with the exception of the custom builds ) will run at 1600 Mhz to my knowledge. Anyway, I am no fan of the integrated Memory controller, but think the Sandy Bridge has taken a step in the right direction and will offer the flexibilty.

Intel has some very good reasons for what it does sometimes. If you will recall, during the Pentium, P2, and P3 days, Intel would never use a 100Mhz base clock (as they do now). They claimed they didn't use it because it was unstable, there was no natural way to get to that clock. Now, I am not sure how true that is, but sometimes they are right on with these things. I would prefer slightly slower and stable over stupidly fast, and unstable.

With memory controllers though, I have to admit that I do like the external controllers, as that allows you to buy better boards with better controllers. At the same time, with it integrated, you don't have to worry about cheap boards having crap controllers either.

I have to admit- I do like the latency numbers.

Yeah, the key word here is "Sometimes". I don't think this time was one of them. The Multiplier on the CPU has always been higher than their specs. On the 920 the multiplier was supposedly 20, but the one in here was at 21. on the 960 it was supposed to be 24 and the one in here is at 25. The result was that the 920 was running at 2.8 Ghz ( the same as the 930 ) and the 960 is running at 3.34 Ghz. I'm not complaining mind you, just pointing it out. Too bad they didn't leave the memory multiplier up to 6 or whatever. With the eMachines, the memory was wide open, or at least to what the MOBO would support, so you had some flexibilty to upgrade. I was told by both eMachines AND Intel ( It is an Intel D102GGC2 MOBO in there ) that it only supported 533 Mhz DDR2 memory. I installed 667 Mhz Memory ( the processor had an 800 MHz FSB ) and it showed as and ran at 667 Mhz. Didn't try 800 Mhz ram, as I figured that it probably would only run at 667 Mhz. Anyway, I think Intel was a bit off the mark when they released the first series of Core i7's and probably should have researched and tested them a bit further. As I recall they had some worries about heat, which proved to be unfounded. Price not withstanding, I think a lot of folks stayed away from the 1366 Based i7's because of the things being hampered by being locked down. Oh well, It looks like the Sandy Bridge is going to be the one that will become the standard. When you consider the price / performance aspect, it seems to be a winner.

[SnyperTodd]

coastie65, on 28 April 2011 - 09:13 AM, said:

Not so sure of that integrated memory controller. I think I liked it better the other way ( non integrated), as it gave you more flexibility as to the memory speed. As it is, they integrated it and then locked the speed to whatever. I realize that is not the case with the Sandy Bridge processors as those things are unlocked. Because of that, I suspect that everything between the LGA 775 based Processors and the Sandy Bridge will go by the wayside in time, at least in the sense of those who do not wish to build and still want some upgrade flexibility. I had plenty with the eMachines ( both processor and memory ), not so much with this Gateway as the only performance update options available were the graphics card and processor. The memory was locked at 1066 Mhz although the MOBO will support up to 1600 Mhz. I blame that primarily on Intel's rush to get the 1366 based Core i7 out and not enough testing to determine what could comfortably be run ( they had no idea and locked things down to what had been tested ). It was those who bought them and started tinkering around with them that found out what they could do and informed Intel that their assumptions were wrong. Intel has unlocked some 1366 based i7's, as well as some 1156 based stuff, to run the memory at 1333 Mhz. No prebuilts ( with the exception of the custom builds ) will run at 1600 Mhz to my knowledge. Anyway, I am no fan of the integrated Memory controller, but think the Sandy Bridge has taken a step in the right direction and will offer the flexibilty.

You're right about the prebuilts. The memory multiplier isn't locked in the processor, it's locked in your motherboard. If you had gotten that EVGA board working, you'd be able to run whatever memory multiplier you want to on either of your LGA 1366 processors. The memory multiplier isn't locked in the processor. Simple as that.

Edit: Not sure what happened, but below is from Coastie65:
My Point is that you can get prebuilt Gateway DX's ( P55 1156 Socket ) running DDR3 Dual Channel @ 1333 Mhz. In fact, a lot of 1156 stuff is coming out running at 1333 Mhz. There were a couple of FX's in this series running the 965 & 975 with the memory running at 1333 Mhz. When they talk about the Sandy Bridge and mention that the processor is unlocked, it is the only one that they have unlocked. In my case, the memory multiplier is locked at, I think ,4 instead of 6. That is within in the controller which is integrated in the processor, which is locked. Sometime down the road, I guess I'll move to the Sandy Bridge ( i7 2600k is on my list at the moment ), but I'll stay with the i7 960 for the time being as it is doing well. I wasn't real happy with that EVGA mobo as it didn't have enough USB headers. I liked everything else though as things ran much cooler than the one in here ( had better hardware such as heat sinks ). I am a bit surprised that they haven't backed up and corrected the issues with the 1366 as far as the memory speed goes. There is no reason for them to leave them hobbled, except for maybe getting people to move to something else, which may well be the case.

This post has been edited by SnyperTodd: 07 May 2011 - 06:27 AM

[SnyperTodd]

waldojim, on 28 April 2011 - 09:30 AM, said:

With memory controllers though, I have to admit that I do like the external controllers, as that allows you to buy better boards with better controllers. At the same time, with it integrated, you don't have to worry about cheap boards having crap controllers either.

I have to admit- I do like the latency numbers.

I agree with you to an extent. It was more of an issue with AMD's notoriously weak IMCs in AM2/AM2+ processors, but the LGA 1366/1156/1155 IMCs are very robust. I've seen them run some insane numbers that my X48 DDR3 controller can't even come close to. Obviously the memory itself makes some difference, but Intel's IMCs as a whole are consistently better than Intel's previous external memory controllers. Not only the obvious throughput advantages, but faster and tighter as well.

[waldojim]

SnyperTodd, on 28 April 2011 - 12:48 PM, said:

I agree with you to an extent. It was more of an issue with AMD's notoriously weak IMCs in AM2/AM2+ processors, but the LGA 1366/1156/1155 IMCs are very robust. I've seen them run some insane numbers that my X48 DDR3 controller can't even come close to. Obviously the memory itself makes some difference, but Intel's IMCs as a whole are consistently better than Intel's previous external memory controllers. Not only the obvious throughput advantages, but faster and tighter as well.

As I am fairly new to the i5 market, I cannot comment too much on Intel's controller. But I will agree, the AMD controller gives very little room to play.

[waldojim]

OK, I would like to state that I have never been a person who gave Turbo Boost any credit... at all. I actually never liked the premise of it, nor the way you seemed to be robbed by the locked base clock of current generation chips.

That said, I must give Intel credit where it is due. Intel is true to their word, as long as conditions are met, they will let you enjoy massive overclocked speeds.

I decided to try something tonight to see what the chips really do with a hefty load. That said, I will have to make a point of increasing that load later - more on that in a minute. Right now, I am running Handbrake in what I would consider a worst case scenario. To be honest, I am trying to recompress a few movies for an upcoming trip (don't need 1080P - so why bother wasting battery life decoding it?). SO what I have going here is a resize of video from 1920x1080 to 720x320 (give or take - I hate anamorphic movies), and this is being compressed to x.264, uneven multi-hexagon MD searching, with an area of 20, CABAC coding, 8x8 transform, weighted P-frames, etc outputting to a video file with an average (video) bitrate of 700Kb/s. I am mentioning this, because I have yet to come across methods that do a better job with the video output - try it guys, you will be impressed. I also bring it up because I get a whopping FIVE frames per second recoding. Well... 5.7 to be exact.

I will not get into Hyper Threading, or the merits thereof, as I have never met the right conditions to actually take advantage of it. But I will get into some numbers that I can comment on. Right now, two hours into the recoding session, I am currently running at 3392.8 Mhz. I am running 69C per core, and the fan appears to be at about 60~70% speed right now. The CPU has been maintaining 3.192Ghz and 3.392Ghz this entire time. Never dipping below 3.192. For this, I give Intel credit where credit is due. I LOVE seeing 3.4Ghz on a laptop!

More importantly though, is that I am typing on the same laptop that is recoding, and I was watching a movie during a good portion of the recoding. The i7 is a multitasking beast that feels a lot like my desktop Core 2 Quad. Quick, responsive, and uncaring what load there was, it will find a way to balance it.

For those people unsure of why I am getting into all this: It is because of how well the Intel chips take stress that I prefer them over the AMD counterparts. My AMD machines never run this smooth...

[SnyperTodd]

coastie65, on 28 April 2011 - 09:13 AM, said:

I am a bit surprised that they haven't backed up and corrected the issues with the 1366 as far as the memory speed goes. There is no reason for them to leave them hobbled, except for maybe getting people to move to something else, which may well be the case.

They haven't corrected that issue because it's not an issue. They are working as designed. They're only hobbled by the BIOS of whatever motherboard they're in. It is absolutely NOT built into the processor.

This post has been edited by SnyperTodd: 07 May 2011 - 06:30 AM

[coastie65]

SnyperTodd, on 07 May 2011 - 06:29 AM, said:

coastie65, on 28 April 2011 - 09:13 AM, said:

I am a bit surprised that they haven't backed up and corrected the issues with the 1366 as far as the memory speed goes. There is no reason for them to leave them hobbled, except for maybe getting people to move to something else, which may well be the case.

They haven't corrected that issue because it's not an issue. They are working as designed. They're only hobbled by the BIOS of whatever motherboard they're in. It is absolutely NOT built into the processor.

I don't know. I'm basing my assumptions on the fact that you see the Sandy Bridge being advertised as an "Unlocked Processor" and the fact that I remember seeing and article way back when, when the Core i7 ( 1366 ) first came out, the fact that Intel chose to lock down the memory in the processor s well as the QPI. I believe it had something to do with a presumed Heat issue, which was later proven to be erroneous. Anyway, tried again to unlock the BiOS in here but nothing. Booted to the flash drive, typed in the command and hit enter. After a pause all I got was another prompt. I went back to square one and reformatted and redid the flash drive, even went so far as to try it in FAT32 format instead of FAT to see if that would work, and got the same results. Oh well, at least I can get the Flash Drive to boot.

This post has been edited by coastie65: 07 May 2011 - 07:05 AM

[SnyperTodd]

coastie65, on 07 May 2011 - 06:59 AM, said:

I don't know. I'm basing my assumptions on the fact that you see the Sandy Bridge being advertised as an "Unlocked Processor" and the fact that I remember seeing and article way back when, when the Core i7 ( 1366 ) first came out, the fact that Intel chose to lock down the memory in the processor s well as the QPI. I believe it had something to do with a presumed Heat issue, which was later proven to be erroneous. Anyway, tried again to unlock the BiOS in here but nothing. Booted to the flash drive, typed in the command and hit enter. After a pause all I got was another prompt. I went back to square one and reformatted and redid the flash drive, even went so far as to try it in FAT32 format instead of FAT to see if that would work, and got the same results. Oh well, at least I can get the Flash Drive to boot.

The only thing unlocked on an unlocked SB is the multipliers. That's the only way to effectively overclock a SB processor. Raising the reference clock won't work on them because they only have one main clock generator that is tied to everything. That is their major weakness, and the "K" (unlocked) processors are a crutch to get around that massive weakness. The LGA1366 processors, on the other hand, are not locked down themselves. If you want to raise the memory speed on a Nehalem, you either raise the uncore clock or raise the memory multiplier. Neither are locked in the processor, nor is the QPI. They are only locked in crippled BIOSes. The article you are referencing above was based on an Engineering Sample, and yes, those were locked, but the production processors are not. The guys that wrote that article even printed a correction to the above referenced article after they realized that.

[coastie65]

Ok. Yeah, I know the memory isn't locked per se and that you can raise the multiplier and increase the memory speed. I think for 1600 Mhz, 6 should do it. Now that you mention it, I believe I did see the correction. The BiOS in here is no help for sure.

[Kazmatron]

After doing some messing around here's my CPU-Z and Novabench stuff.

877 seem about right for that setup?

This post has been edited by Kazmatron: 29 May 2011 - 12:09 PM

[waldojim]

Kazmatron, on 29 May 2011 - 12:05 PM, said:

After doing some messing around here's my CPU-Z and Novabench stuff.

877 seem about right for that setup?

That is actually pretty good.