Intel Core i7 Processor Models and Pricing Revealed!

[thescavenger]

I'll be picking up the cheap 775 stuff when everyone upgrades

From what I've seen the extra memory bandwidth doesn't really help THAT much in real world performance. The 4 extra threads will be nice if you have one of those rare programs that can use all of them. The clockspeeds are meh.

ETA on first 8 threaded game? Long long time.

On the plus side, native SLI on Nehalem boards yay.

I might get my feet wet with Nehalem when the lower socket or the revision comes out in a while.

I completely agree.

How many games or apps can even utilize four threads right now? Not many. My current quad can handle everything thrown it so why would I need to upgrade.

I don't expect the extra memory bandwidth to increase overall performance either. I'll consider upgrading to the new architecture when the nehalem boards get the DDR3 up to 1600MHz+ with really tight timings.

 

[BossBorot]

personally im not impressed. no news as of yet has made me feel that its worth all the money that you are going to end up spending if you need a new mobo, cpu, and ram. But then again I tend to remain budget minded and i7 certainly wont be budget minded for awhile

 

[Badbonji]

ETA on first 8 threaded game? Long long time.

Still waiting on a decent 4 threaded one

Why has the cache gone back down to 8mb? The extreme addition sounds nice, higher QPI speed so hopefully won't bottleneck tri channel DDR3.

 

[jason4207]

ETA on first 8 threaded game? Long long time.

Still waiting on a decent 4 threaded one

Why has the cache gone back down to 8mb? The extreme addition sounds nice, higher QPI speed so hopefully won't bottleneck tri channel DDR3.

All the extra cache isn't needed w/ the increased memory bandwidth.

 

[MR-FIX-IT]

$284 seems good for a quad + HT on each core.

I thought HT required the application to take advantage of it. I was happy HT was removed. In many instances it slowed your computer down because it was using the logical core instead of the physical core. A quad core will show up as 8 cores, 4 physical and 4 logical. I'm hoping that this can be turned off in BIOS like the others did.

EDIT:

I looked up the Core i7, and found this on Wiki:

A 2.93 GHz Core i7 940 system has been used to run a 3DMark Vantage benchmark and gave a CPU score of 17,966.[11] The 2.66 GHz Core i7 920 scores 16,294. A 2.4 GHz Core 2 Duo E6600 scores 4,300.[12]

AnandTech tested the Intel QuickPath Interconnect (4.8 GT/s version) and found the copy bandwidth using triple-channel 1066 MHz DDR3 was 12.0 GB/s. A 3.0 GHz Core 2 Quad system using dual-channel 1066 MHz DDR3 achieved 6.9 GB/s.[13]

Overclocking will be possible with the 900 series and a motherboard equipped with the X58 chipset. However, the Lynnfield and Havendale processors will use a PCH removing the need for a northbridge chipset and requiring a different motherboard design.[14]

 

[Dapman02]

I thought that clock for clock Neha's are faster. I also heard about them being worse for games. I believe Intel brought out neha just to try to beat AMD in the server game.

 

[jason4207]

I thought HT required the application to take advantage of it. I was happy HT was removed. In many instances it slowed your computer down because it was using the logical core instead of the physical core. A quad core will show up as 8 cores, 4 physical and 4 logical. I'm hoping that this can be turned off in BIOS like the others did.

Are you sure about this? I thought it was 4 physical cores, and 8 logical cores. There is no way for the PC to use a logical core, and not use a physical core.

 

[cyberfish]

I thought HT required the application to take advantage of it. I was happy HT was removed. In many instances it slowed your computer down because it was using the logical core instead of the physical core. A quad core will show up as 8 cores, 4 physical and 4 logical. I'm hoping that this can be turned off in BIOS like the others did.

HT is completely transparent to the application. It's not possible to use a logical core and not a physical core. There are 4 physical cores with 2 logical cores each.

The OS needs to know about HT, though (and I think XP and up does), to avoid, for example, scheduling two threads to be executed on the same core, leaving 3 cores idle.

 

[MR-FIX-IT]

Are you sure about this? I thought it was 4 physical cores, and 8 logical cores. There is no way for the PC to use a logical core, and not use a physical core.

That would indicate a "12 core" CPU? (That would be pretty cool..) I thought for every physical there was a logical...or virtual core.

 

[DavidJa]

HT is completely transparent to the application. It's not possible to use a logical core and not a physical core. There are 4 physical cores with 2 logical cores each.

The OS needs to know about HT, though (and I think XP and up does), to avoid, for example, scheduling two threads to be executed on the same core, leaving 3 cores idle.

It's 1/1 ratio for every physical core you have 1 logical.
quad core then gives you 8 threads.

 

[jason4207]

That would indicate a "12 core" CPU? (That would be pretty cool..) I thought for every physical there was a logical...or virtual core.

There are 4 physical (actual) cores. Every operation has to happen on one of those core...no way around that unless you have another CPU. W/ HT the OS sees 8 cores. These are logical cores, b/c as far as the software is concerned there are 8 cores. You don't add them up.

Similarly, w/ RAID1 you have 2 physical drives, but only 1 logical drive. That doesn't mean you have 3 drives.

 

[Evilsizer]

ETA on first 8 threaded game? Long long time.

Still waiting on a decent 4 threaded one

Why has the cache gone back down to 8mb? The extreme addition sounds nice, higher QPI speed so hopefully won't bottleneck tri channel DDR3.

L2 is down to 256KB for each core with a 8MB L3 cache. the L3 cache is what allows the passing of instructions or work to other cores.

All the extra cache isn't needed w/ the increased memory bandwidth.

to a point it is and isnt.... for things like FAH and superpi L2 still rules. the L3 runs slower and wont be as good for those apps.

I thought HT required the application to take advantage of it. I was happy HT was removed. In many instances it slowed your computer down because it was using the logical core instead of the physical core. A quad core will show up as 8 cores, 4 physical and 4 logical. I'm hoping that this can be turned off in BIOS like the others did.

well the app needs to be multicore coded to make use of the extra virtual cores that are showing up via HT... if the program like Cinebench can use more core real or virtual it will use them. HT just inserts a instruction when the cpu is waiting for another to be processed. It does help with cpu EFF, in that reguard. i mean on my Atom 220 i can encode 2 MP3's at once and its took 3m14s for one and 3m15s for the other. i made a bat file that ran 2 instances of LAME 3.89. the file was the same as well 94mb.wav,track 7 from Prodigy "Fat of the land" cd. now if i encoded just one mp3 at a time it would take each one using the same track 2m45s, compressing down to 128bit using the -h flag on Lame. it took slightly longer to encode but i did both at once vs doing each one taking a total of 5m30sec. when i use to use HT on a [email protected] it was more eff to run 2 instance of FAH. slightly longer per frame but able to do 2 wu's at once.

Yes just like with the Atom 220 and older P4's with HT it can be diabled in the bios.

I thought that clock for clock Neha's are faster. I also heard about them being worse for games. I believe Intel brought out neha just to try to beat AMD in the server game.

Neha in Encoding is around 30% faster per clock vs current 45nm cpus. this number came for a "early" review of neha. intel controlled of course so i would like to see "real" review for solid numbers.

Are you sure about this? I thought it was 4 physical cores, and 8 logical cores. There is no way for the PC to use a logical core, and not use a physical core.

it is 4 physical cores/4 logical cores with 4 virtual cores.

 

[jason4207]

it is 4 physical cores/4 logical cores with 4 virtual cores.

Hmmm...I'm still going w/ what I learned in my Computer Engineering classes unless someone can show otherwise.

Physical vs. logical is low-level vs high-level.

At the low level we have 4 physical cores. At the high-level, and how the OS sees it, is that there are 8 logical cores. The OS doesn't worry about the low-level at all.

This is really old, but it does back me up...

Also, I imagine that even
> with dual core Intel CPUs there is a chance that they might do HTT
> inside each core, so that a single physical processor would have 2
> cores each with 2 HTT so a total of 4 logical CPUs from one piece of
> silicon.

Link

to a point it is and isnt.... for things like FAH and superpi L2 still rules. the L3 runs slower and wont be as good for those apps.

You are correct, but I think the reason Intel is cutting down on the cache is for this reason. AMD chips have traditionally had less cache for similar reasons.

 

[Evilsizer]

Hmmm...I'm still going w/ what I learned in my Computer Engineering classes unless someone can show otherwise.

Physical vs. logical is low-level vs high-level.

At the low level we have 4 physical cores. At the high-level, and how the OS sees it, is that there are 8 logical cores. The OS doesn't worry about the low-level at all.

This is really old, but it does back me up...



Link




You are correct, but I think the reason Intel is cutting down on the cache is for this reason. AMD chips have traditionally had less cache for similar reasons.

ok so i never took a comp glass, as from the way i look at it. you physically have one cpu, it has 4 physical cores, which also makes logical. we get 4 more cores thru HT, which i would think is a virtual core. maybe there is some book or website i should read to undstand that part more.

as for the last part of what you said.. it makes sense they cut the L2 size down. it allows for higher yeilds per wafer for the target clocks they want. if i recall right the L3 will be running either half or at the QP link speed. now as you see per the QP bandwidth for the EE in the FP, its a higher speed. my thoughts on amd's less L2 is due to their yeild problems and possibly cost factor. i mean you dont see their dual cores ocing like intels nor thier quad cores. this isnt about AMD though.....

 

[jason4207]

ok so i never took a comp glass, as from the way i look at it. you physically have one cpu, it has 4 physical cores, which also makes logical. we get 4 more cores thru HT, which i would think is a virtual core. maybe there is some book or website i should read to undstand that part more.

as for the last part of what you said.. it makes sense they cut the L2 size down. it allows for higher yeilds per wafer for the target clocks they want. if i recall right the L3 will be running either half or at the QP link speed. now as you see per the QP bandwidth for the EE in the FP, its a higher speed. my thoughts on amd's less L2 is due to their yeild problems and possibly cost factor. i mean you dont see their dual cores ocing like intels nor thier quad cores. this isnt about AMD though.....

Physical, logical, virtual...

it's just semantics. We both understand the basic idea.



Before Conroe, AMD chips didn't need as much cache to compete w/ P4's, and I always thought it was b/c of the direct memory access afforded by the on-die memory controller. I figured the same would hold true w/ i7.

 

[Gautam]

At the low level we have 4 physical cores. At the high-level, and how the OS sees it, is that there are 8 logical cores. The OS doesn't worry about the low-level at all.

Yes...

 

[Evilsizer]

Before Conroe, AMD chips didn't need as much cache to compete w/ P4's, and I always thought it was b/c of the direct memory access afforded by the on-die memory controller. I figured the same would hold true w/ i7.

thats true they didnt but we are also talking to completly different styles in Arch here. well intel took the brute force approach thru high clocks, Amd went more EFF per clock. which does make a difference as we saw, the IMC did just help even more of course. there is a "set" amount of L2 i would like to see on neha per core from going over the numbers of gains of doubling L2. for our purposes 1mb L2 is the right fit anything more is just a waste for us.

As the early benchmarks show poor FPS at higher res's. This could be clockspeed related or L2. it is still hard to say IMO, since the competing cpu vs neha had way more L2. i think it was a 12mb L2 cpu, i dont see how one could compare then to a 512k L2. we all know more L2 makes the cpu more EFF at doing work per clock but to a point. with gaming "we" at least need 1MB L2 and a clock speed of around 3.2-3.4 on current "core 2" cpus. now even though i7 is showing around a 30% increase per clock in encoding that is because of the IMC not L2. L2 has no bearing on encoding, its all about pushing DATA and lots of it. which is why at the time the higher FSB to the cpu for the P4 allowed it to do better at encoding vs gaming. Even with the IMC amd's cpus were not able to keep up in that area.

im sure there are otherthings i am missing but im just going to stop there....