Core i7 LAUNCH CONFIRMED: 11/16

[Sr7]

Maybe I'm missing something here, but from all leaked benches so far, i'm seeing that Nehalem doesn't appear to be much/any faster in games than Intel's previous processors. So my question becomes: Why do people want i7 processors any more than a Penryn or something?

Even if it's truly faster in other things.. what will you do in an average day that justifies having a faster processor than the stuff that's out right now? Open sites/media player faster? Does anyone really run 20 things at once?

I know some people might say "I want to crunch/fold faster!" but GPUs can already do it faster, many times over. So what is it that people want? I honestly don't really understand... maybe someone could clarify this or explain it to me?

[JumpingJack]

Maybe I'm missing something here, but from all leaked benches so far, i'm seeing that Nehalem doesn't appear to be much/any faster in games than Intel's previous processors. So my question becomes: Why do people want i7 processors any more than a Penryn or something?

Because some people do not use their computers for just gaming.

[Jowy Atreides]

Maybe I'm missing something here, but from all leaked benches so far, i'm seeing that Nehalem doesn't appear to be much/any faster in games than Intel's previous processors. So my question becomes: Why do people want i7 processors any more than a Penryn or something?

Even if it's truly faster in other things.. what will you do in an average day that justifies having a faster processor than the stuff that's out right now? Open sites/media player faster? Does anyone really run 20 things at once?

I know some people might say "I want to crunch/fold faster!" but GPUs can already do it faster, many times over. So what is it that people want? I honestly don't really understand... maybe someone could clarify this or explain it to me?

it's a combination of future proofing and elitism
for the longest time companies have been trying to write some sort of code to rewrite any single threaded program into a multi-threaded one

if it ever happens then multi core cpu's just became serious for users other than server builders

[Sr7]

Because some people do not use their computers for just gaming.

(Despite your smart ass reply) that's just my point.. what do you stand to gain performance-wise in your everyday usage, if there are no gains in gaming which is one of the few things that *do* legitimately tax your system to it's fullest?

I think in some cases GPGPU has far more potential for speedup on image processing, and video processing many orders of magnitude over what a slightly faster CPU micro-arch can bring. So what does that leave that makes you really want something faster than a current Core 2 for?

You really think Nehalem is going to be perceivably faster? I think at a certain point, things just open "fast", and a 40% gain from 10ms to 6ms is inperceivable, because it's below a threshold where we can actually discern between for two timings.

I'm willing to bet that currently most people can't even tell the difference between a low end core 2 duo's performance and a high end core 2 duo. This didn't used to be the case with older processors, when booting your system or waiting on some task still felt sluggish.

This is just my opinion, but I'm hoping someone can clarify this logically for me. It seems to me like some people are just hungry for new technology to play with, not necessarily having a use for it. Sure I'd like something 100 times faster than a core 2, but chances are it's not going to be perceivable when I'm opening a photo, or my browser, or running microsoft word, or reading my email, right?

[savantu]

But thats just my point.. what do you stand to gain performance-wise in your everyday usage, if there are no gains in gaming?

I think GPGPU has far more potential for speedup on image processing, and video processing. So what does that leave that makes you really want something faster than a current Core 2 for?

You really think Nehalem is going to be perceivably faster? I think at a certain point, things just open "fast", and a 40% gain from 10ms to 6ms is inperceivable, because it's below a threshold where we can actually discern between two timings.

I'm willing to bet that currently most people can't even tell the difference between a low end core 2 duo's performance and a high end core 2 duo. This didn't used to be the case with older processors, when booting your system or waiting on some task still felt sluggish.

??

Work isn't about opening time ; but how much it takes to get the job done.3 minutes instead of 4 is a huge improvement.For someone who is doing this on a daily basis it could means several days a month.

Nehalem has lots of untapped potential and this will become apparent with new generations of games.Games that are now in the market were designed in the 2004-2006 timeframe.Rest assured that those designed in 2006-2008 will be faster on Nehalem than on Core.

[saint-francis]

What if you don't play video games at all? Does this mean you don't need a fast processor? Sometimes encoding a video can take 24 - 48 hours with a modern core 2 quad. If that could be brought down even by a couple ours it's a big improvement. If all you use you computer for is to browse the web you don't need a beefy processor or an account on xtremesystems.

[Nedjo]

Nehalem has lots of untapped potential and this will become apparent with new generations of games.Games that are now in the market were designed in the 2004-2006 timeframe.Rest assured that those designed in 2006-2008 will be faster on Nehalem than on Core.

where do you see place for game optimizations tailored for Nehalem?

[Hornet331]

I'm willing to bet that currently most people can't even tell the difference between a low end core 2 duo's performance and a high end core 2 duo. This didn't used to be the case with older processors, when booting your system or waiting on some task still felt sluggish.

you lost that bet.

im no on a E5200 and its nothing compared to a QX9650, most noticeable on daily bases on flash an animated gifs, but also on TF2, the cpu pukes itself when theres to much action even though i have a 4870 oced. On my QX9650 everything was smooth a a baby bottom, on the E5200 not... need to oc that lil bastard higher, 3,2ghz dont cut it.

[Nedjo]

Nice sum up of Intel's P6 architecture evolution with Nehalem as as highest level of this progress... naturally Netburst P4 is different spicie:

[Chad Boga]

(Despite your smart ass reply) that's just my point.. what do you stand to gain performance-wise in your everyday usage, if there are no gains in gaming which is one of the few things that *do* legitimately tax your system to it's fullest?

I couldn't agree more with you.

Nehalem's initial launch appears quite underwhelming to me, and it looks like the new computer I get in the first qtr of 2009 will be an E8500.

Perhaps if AMD could become remotely competent again, maybe even Shanghai would make more sense for me.

[Chad Boga]

Basically , Nehalem is every bit as good as an equivalent clocked Quad Core Penryn in games

With them using a mere 9800GTX video card, I wouldn't be paying that review much heed.

Unfortunately I think Nehalem is going to be 5% or so down on Penryn for gaming.

[Shintai]

Nice sum up of Intel's P6 architecture evolution with Nehalem as as highest level of this progress... naturally Netburst P4 is different spicie:

Your chart is so dumbed down for PR purpose it makes kitties cry. You could add dozens more points to i7. Even just take power states as a tiny example. So busy in showing C0-C4 on the others. Yet C6 states are...where?

And thermal sensor under Core 2? That sensor was in long before.

And they missed an entire CPU line with Core.

I give that chart a nice fud rating.

[Nedjo]

I give that chart a nice fud rating.

feel free to mail to mydivers.com

[450tantrum]

So which X58 motherboards will be readily available on Nov. 16th ?

[Sr7]

it's a combination of future proofing and elitism
for the longest time companies have been trying to write some sort of code to rewrite any single threaded program into a multi-threaded one

if it ever happens then multi core cpu's just became serious for users other than server builders

Yes this is true, and its very very complex. If Intel had the answer they'd have it in the market, but right now they're mostly selling quadcores which have little use in everyday life (due to lack of quad-core aware/threadable applications/workloads) to people who don't know that they're not actually twice as fast as dual-cores, and may very well cost the same as a higher clocked dual core (which WILL directly benefit you today).

I believe most people who buy quadcores are a) enthusiasts trying to "future-proof" themselves or b) people who are uneducated about how processors and threading work, and assume 4 cores = 4x the speed of 1 core for everything. It's kind of a lie by omission for the average consumer, especially with sales guys at stores like best buy being comped to push the quads and lie to people who don't know the difference.

I was in a Microcenter recently, and I knew exactly what I wanted to buy. I wanted a fast dual core, as opposed to a slower quad core for the same price. This sales kid swore up and down he ran prey on a quad core and it threaded across all 4 cores and that I need to get a quad core! I said I'm pretty sure the game was single or dual threaded, but not quad threaded for 4 cores, so there's no way it could've done what he said. Every time I'd ask questions about something else like mobos for my prospective purchase, he'd bring it back to "hey btw I know you said you don't want a quad core but..." and "BTW the extreme processors have extra special chipset stuff that makes it faster".. these guys are unbelievable.

??

Work isn't about opening time ; but how much it takes to get the job done.3 minutes instead of 4 is a huge improvement.For someone who is doing this on a daily basis it could means several days a month.

Nehalem has lots of untapped potential and this will become apparent with new generations of games.Games that are now in the market were designed in the 2004-2006 timeframe.Rest assured that those designed in 2006-2008 will be faster on Nehalem than on Core.

Of course work isn't about opening time, I was just making a point that I remember back in the day when you'd sit there waiting for a program to launch, and a faster CPU would load it up so much quicker and it was noticeable and you'd feel good about getting a fast processor and spending $400, because you immediately felt you got your money's worth. Now to me, my normal everyday experience feels the same with a 2.33 as with a 3.0 as with a 3.2 quad core.

Also, of course speeding up things that take a long time on the CPU is noticeable like video transcoding, but why do you care about that when a parallel processor via GPGPU is more suitable for that workload anyway, and already is much faster than even Nehalem at the task? The solution is already here TODAY for those heavy parallel loads. It's almost like you're saying "hey I made a faster hard disk drive!" after a new technology that is already faster like SSD has come out.

No one cares how fast the new version of something is if it's already slower than another technology at that same task because it's not as well suited to the task.

I guarantee you that despite the fact that Intel isn't saying this now because they want to continue selling processors, they'll start telling you how larrabee is better than any multi-core CPU once it's out.

Lastly, how could Nehalem be "untapped" except through some currently unused instruction set? Is there something like this added via nehalem?

you lost that bet.

im no on a E5200 and its nothing compared to a QX9650, most noticeable on daily bases on flash an animated gifs, but also on TF2, the cpu pukes itself when theres to much action even though i have a 4870 oced. On my QX9650 everything was smooth a a baby bottom, on the E5200 not... need to oc that lil bastard higher, 3,2ghz dont cut it.

Okay, maybe so, but the point is you can get a fast as hell dual core (3GHz) for $150 now and be running those games insanely fast. I remember I used to have to buy processors that were considered "fast but not extreme" for around $300-$400 (I bought an AMD Barton 3000+ with 400FSB ) but now I get something smoking fast in the $150 range.

I believe CPUs are amazingly more affordable now, largely because Intel is banking on quad-cores to make them money, even though they don't scale in most games or everyday tasks. People always argue that "the same thing happened with dual cores at first and now they're dual threaded, so that means we'll see games using 4 and 8 cores soon" but they're missing the point.. the more cores you have, it gets exponentially harder/more complex to thread efficiently (completely independent threads which don't stall each other).

So we can't safely assume that developers will be coding with ease on 128 cores a decade from now.. I believe the prevalence of dual-threaded applications is a phenomenon you can't just infinitely extrapolate on.

[Ghigo]

http://translate.google.com/translat...118/118851.htm

[xVeinx]

I was in a Microcenter recently, and I knew exactly what I wanted to buy. I wanted a fast dual core, as opposed to a slower quad core for the same price. This sales kid swore up and down he ran prey on a quad core and it threaded across all 4 cores and that I need to get a quad core! I said I'm pretty sure the game was single or dual threaded, but not quad threaded for 4 cores, so there's no way it could've done what he said. Every time I'd ask questions about something else like mobos for my prospective purchase, he'd bring it back to "hey btw I know you said you don't want a quad core but..." and "BTW the extreme processors have extra special chipset stuff that makes it faster".. these guys are unbelievable.

The guys in the Tustin, CA Microcenter are pretty good. They give good, honest advice when needed, but many know less than me and tend to leave me alone .

As to who really needs this: I'd love to get some of my (biological) data analysis done faster for work stuff. Whether single threaded or multi-threaded, anything that helps me get stuff done sooner is good. Folding also will have gains, especially WCG which does not have a GPU-accelerated version. People working with video/photo manipulation and encoding will be able to use it. People that programs and use a C Compiler will need it. Some of this may be used more for work applications than play, but people can use the extra power.

At some level, do we really have to justify EVERY aspect of our purchase? We're technophiles. It's how we live .

[savantu]

...

Of course work isn't about opening time, I was just making a point that I remember back in the day when you'd sit there waiting for a program to launch, and a faster CPU would load it up so much quicker and it was noticeable and you'd feel good about getting a fast processor and spending $400, because you immediately felt you got your money's worth. Now to me, my normal everyday experience feels the same with a 2.33 as with a 3.0 as with a 3.2 quad core.

How do you know that Nehalem's IMC won't offer a snappiness not present with Core CPUs ? Many AMD people talk about being smooth due to the IMC.Maybe the smoothness will arrive in the blue camp then too..

Also, of course speeding up things that take a long time on the CPU is noticeable like video transcoding, but why do you care about that when a parallel processor via GPGPU is more suitable for that workload anyway, and already is much faster than even Nehalem at the task? The solution is already here TODAY for those heavy parallel loads. It's almost like you're saying "hey I made a faster hard disk drive!" after a new technology that is already faster like SSD has come out.

I'm not sure there is a SW stack and know-how widespread enough to make the GPGPU a reality yet.So a faster CPU still matters.

Lastly, how could Nehalem be "untapped" except through some currently unused instruction set? Is there something like this added via nehalem?

Every CPU has some quirks when it comes to tunning.With Core you wanted to minimize FSB traffic , this becomes irrelevant with Nehalem.
Accordingly , inter-core communication is much faster on Nehalem , another thing to worry less about.Then you have macro-op fusion at 64bit with Nehalem , not only that , but new combinations are possible too.
Lastly , there is HT and making proper use of it.

Tunning for Nehalem will probably be quitte different than for Core and this brings up a lot of new possibilities.The fact that Nehalem does so well with SW tuned for Netburst and Core is a testimony to the designs all around stellar performance.

where do you see place for game optimizations tailored for Nehalem?

Most games aren't able to properly use more than 2 threads.This is fine with Core generation because it helps with locality , going to the other die is costly in terms of performance.
Games which will appear next year and in 2010 will be designed with 4 threads or more in mind.This is already a win ( a boost will be visible on Core products too ) but I expect Nehalem to gain the upper hand.

Games specifically tailored for Nehalem will probably appear in late 09/2010 I'd say.

[Nedjo]

I think it's obvious that general purpose CPU as we know it has become pointless... what we really need is solution for accelerating specific workloads in a way that GP CPU isn't capable of doing it.

[Shintai]

I think it's obvious that general purpose CPU as we know it has become pointless... what we really need is solution for accelerating specific workloads in a way that GP CPU isn't capable of doing it.

Its certainly not the GP CPU thats dead soon. Larrabee is an example on the CPU beating the GPU completely.

And I wouldnt call the generel purpose CPU pointless. You couldnt do nothing really without it. Its the heart of everything

Physics on a GPu is also a temporary "PR stunt". AMDs multicore approach with 6, 8 and 12 core CPUs is also a demonstration in that the CPU is here to be the winner. i7 with 4, 8 and 16 threads and so on.

[Chad Boga]

Its certainly not the GP CPU thats dead soon. Larrabee is an example on the CPU beating the GPU completely.

When did Larrabee get released?

[Helmore]

When did Larrabee get released?

You should know that Shintai comes from the future....

[Shintai]

You should know that Shintai comes from the future....

Or go visit SIGGRAPH.

It should also be of no surprise to anyone. You can argue about rastergame performance. But anything like CTM/CUDA Larrabee is lightyears ahead due to the obviously issue and lacking features of the GPUs.

But its not like we dont have any numbers on Larrabee.

i7 aint released either, yet we still know everything about it.

[Chad Boga]

But its not like we dont have any numbers on Larrabee.

We have no credible numbers on Larrabee

i7 aint released either, yet we still know everything about it.

i7 is weeks away from being released, Larrabee is over a year away.

You embarass yourself making such a comparison.

[Drwho?]

ok, let s look at the GPGPU , CPU and GPU, and let's pay a little attention to what they do, and what it does gives you.
Please forget 5 minutes about that is your favorite brand of PC parts, and stick to the philosophy here.

I am speaking as "Francois", I am not representing my company here. If you response, please just be polite, and we will be able to debate nicely. thanks ...

SO, the CPU. Did you ever asked yourselves, why the PC toke over the Amiga, or over the powerPC for personal computers? Why the x86 got adopted more than the Z80 (I loved it) or the 65xx or 68xxx? The reason is not a business reason, neither a trust reason or what ever. It is because x86 has 1 primary goal: Legacy!

each new processor deal with the previous 10 generation and can run in compatibility mode, you never break away from x86 compatibility, Intel,VIA ,AMD,IBM all are or were careful about this, this is how it become the format. (Plus a lot of hard work in my employer and the other guys)
The competition between x86 CPU providers made them run too, but all of them share the same x86 basement. now, we share x64, thanks to AMD. (Let's be fair, hooooo that hurts my keyboard ;-) )

Now, take a look at what the graphic world offer to you today ... Did you ever get to a possition that your computer will refuse to run a game? I think so, and it is exactly the point, the GPU world is a nightmare of incompatible drivers, via DirectX, optimized drivers, and nVidia, ATi, Intel, Matrox,S3 and all the GPU providers are not really responsible for it, neither Microsoft, DirectX is something that is extremely complexe and is a real nightmare to developpe for, to program a driver, or to run the games ... everybody is frustrated with this ... and we should be.
The video game industries spend more time investing into the Quality insurance of the games, more than developping it ... outch!

Now, take a look at what is a head ... GPGPU: nVidia has its solution: CUDA, AMD has its own, Intel has its own too. In the case of CUDA, you already has version 1 and 2, incompatible between each other, same for all other hardware, you need a driver in between to deal with the hardware bugs and design incompability between GPU generations ... If we keep going this way, we are going for a royal mess ...
Imagine this, today, you keep installing DirectX from Jan 08, or Sep 07, and you need this driver for this, and this one for that ... you got to upgrade windows here, and there ... grrrrrrrr ... it does frustrate the hell out of me when I got to spend 30 min before I can run the game I just bought. It is one of the reason people go for PS3 and XBOX ... no s.it like this there.

Do you agree with this? it is a mess isn't it? Now, let s say that you got to start doing the driver game too, when you want to play a video ... hoooo, I forgot, it is already the case if you try to play blue ray ... download the HDMI secure certified driver ... but .. noooooooooo , this is not the same driver as my game need ... Now, I got to ping pong between drivers ... grrrrrrrrrrrrrrrrrrrr

so, now, you got badaboom, with this driver ... or that driver ... well, I bid 50$ that in a year, the version of badaboom you have today on your hard drive will not work with the next year new driver and GPU ... it will require a patch of badaboom. Same for most of the other apps ported to CUDA. remember ... you can still boot Dos 6.2 ... hummm

Something is fundamentally different here, and it will take time to fix this. a nightmare of drivers is not the solution, we got to put everything back to more native, try to use the legacy of x86 to solve all of those terrible issue that poison our PCs.

We all have to remove our partisan hat, and try to do what is right for the legacy of the PC. CUDA and co will be driving the PC to the shape the cell phone market is, the incompatibility world. I do this very often, I ask people in a room to take out their cell phones from their pocket. Happyly, we usually discover that out of 10 cell phones, 0 are compatible with each other, if you are lucky, 1 or 2. You are hostage of your cell phone provider/maker to provide you a store for your application, and this is not a good thing, it does fix the price of the software.
The fundamental question the consumers will have to answer soon is simple, do you want to be hostage of one or 2 GPU providers?
I don t think so!

it will take time to create a legacy for the GPU, the way the 8080 , 8088 and 8086 did, but we are going to get there. And when we are there , say bye bye to your daily issues with drivers.
This is how i see the future myself.

It is OK to disagree ,

What do you think?

Francois