Intel Core i5

[JackyP]

Originally posted by: Idontcare
AMD better get with the program and release their own version of (1) power-gating, and (2) dynamic clocking (turbo).

I don't see them having an entry point until bulldozer but 2010 is gonna be a tough year for them if their X4's are competing price/performance and performance/watt with nehalem based powergated and turbo'd lynnfields/clarkdales.

So you're saying there's no way they can implement it in a modified K10.5 version e.g. Magny Cours or some other version slated for launch in 2010? That might hurt a lot..

 

[taltamir]

power consumption has been one of my major pet peeves with PCs... and one of my major purchase influencer's... Also having done the math to reach the cost of ownership instead of just cost of purchase helps... so I really don't see how AMD can hope to compete without seriously improving their power saving features.

 

[Idontcare]

Originally posted by: JackyP

Originally posted by: Idontcare
AMD better get with the program and release their own version of (1) power-gating, and (2) dynamic clocking (turbo).

I don't see them having an entry point until bulldozer but 2010 is gonna be a tough year for them if their X4's are competing price/performance and performance/watt with nehalem based powergated and turbo'd lynnfields/clarkdales.

So you're saying there's no way they can implement it in a modified K10.5 version e.g. Magny Cours or some other version slated for launch in 2010? That might hurt a lot..

Developing/implementing a PCU equivalent circuit block is not trivial, particularly when it comes to the debug/verification phase.

There is a reason it was delayed from Foxton days until Nehalem debuted.

Technically there is nothing preventing AMD from having put in the time and resources already such that they will be ready to debut their version of this technology on MagnyCours/SaoPaolo...but given that they haven't even uttered a word about it I am very doubtful they are intending to implement it prior to 32nm.

Considering how much they talked about Istanbul prior to its release, if anything like PCU was in the cards for MagnyCours then I'd find it strange for AMD to decide now is the time to play that card close to the chest. But I'll grant you that it is not impossible or infeasible for any technical reasoning.

 

[ilkhan]

Originally posted by: Idontcare
Personally I'm not so much worried about fusion or clarkdale, to me these are cost enablers not performance enablers. It will/should lead to lower-cost mobile compute solutions, but I don't think its going to make anything use less power or be faster versus their chipset-based IGP counterparts would have been with the same process technology.

Can you expand on this IDC?

 

[Idontcare]

Originally posted by: ilkhan

Originally posted by: Idontcare
Personally I'm not so much worried about fusion or clarkdale, to me these are cost enablers not performance enablers. It will/should lead to lower-cost mobile compute solutions, but I don't think its going to make anything use less power or be faster versus their chipset-based IGP counterparts would have been with the same process technology.

Can you expand on this IDC?

Sure, take clarkdale for example...what do you think the performance difference would be if you took that MCM'ed IGP/IMC/PCIe 45nm chip out from under the IHS and slapped it into its own flipchip package and soldered it directly onto the mobo?

(i.e. your traditional NB chipset with IGP)

My contention is that power-consumption and performance are probably about the same regardless whether you MCM the those components or kept them discreet.

In fact it will be interesting to see how much, if any, of a latency hit clarkdale has with its DDR3 accesses versus gulftown given that clarkdale's memory controller is no longer "IMC" in the sense of the word that folks like to use it.

Sure the trace paths are going to be less between the CPU and the NB/IGP with the NB/IGP being MCM'ed, but are we really thinking this is going to amount to more than a 10% improvement in IPC? (basically a speedbin's worth of performance increase)

The only place where I see fusion-type products actually deliver some form of performance-related synergy in which the combination product outperforms a simple sum of the parts involved is if/when the compute resources in the CPU and GPU cores start sharing cache resources at a hierarchy level above the IMC.

 

[taltamir]

isn't always about cost though? a multi socket system with cpus that cost 1000+$ each is not much good to most people. The current huge leaps in GPU are largely a matter of cost, not performance.

 

[Soleron]

Even with all that... there is no reason why they wouldn't sell the s1156 i7 for the same price because it costs them THE SAME to buy it from intel!

I'm not going to argue with the rest, but this is wrong. The list price is for 1,000 unit quantities. If a store buys more, they pay less, sometimes a lot less. That's why when Intel says a chip is $183 it sells for $183 at Newegg despite a 10%+ retail margin: Intel's selling say 50,000 of them for $160 or so each.

 

[ilkhan]

edit: @IDC:
good points, and Ive mentioned the IMC!=IMC thing before too.

But it is nice to see them shrinking the IGP a bit. Saves cost, saves mobo space, it does have benefits. Im still wanting the 2.13Ghz (LV) arrandale...

 

[AdamK47]

Originally posted by: taltamir
the current i7s have a +2 speed increase, the lynfields have a +5 increase.
So I am not sure what you mean by +2

My i7 had a nice increase all the way to 4GHz.

 

[IntelUser2000]

Originally posted by: Idontcare

Sure the trace paths are going to be less between the CPU and the NB/IGP with the NB/IGP being MCM'ed, but are we really thinking this is going to amount to more than a 10% improvement in IPC? (basically a speedbin's worth of performance increase)

10% IPC increase isn't small, if its single thread.

 

[taltamir]

Originally posted by: AdamK47

Originally posted by: taltamir
the current i7s have a +2 speed increase, the lynfields have a +5 increase.
So I am not sure what you mean by +2

My i7 had a nice increase all the way to 4GHz.

we are talking turbo mode not overclocking

 

[Shmee]

I think we should talk overclocking. Personally, IDC how good a proc is @ stock, but rather how it is with a good OC. If the 1166 proc's cant overclock better than a D0, then they arent worth buying IMO. With a $200 D0, 6 GB memory ~$100-$135, and a decent mobo: ~$200-$500
and a good air cooler, you will likely get +4 Ghz on the 920 D0. If the i5's and the 1166 i7's OC as well or better, all the power to them. If not, I don't think I would be nearly excited about it.

I recommend to wait and see how they OC, and then compare how the overclocked processors perform, taking price, features and OVERCLOCKED performance into account.

Again, overclock is the keyword for me, and for many other enthusiasts. Since we have no news of how well 1166 OC, at least that I know of, I would wait and not get excited quite yet.

If anyone has evidence to the contrary, please, present it

 

[taltamir]

i prefer undervolting... overclocking takes up too much electricity (which is expensive).

 

[Shmee]

Originally posted by: taltamir
i prefer undervolting... overclocking takes up too much electricity (which is expensive).

A valid preference

I currently do not have to pay the electric bill, luckily, but if I did, I would certainly not leave the comps on 24/7.

As a hardcore gamer, and someone who simply likes the challenge of overclocking, I mainly just worry about stability, speed, and sometimes voltage migration I try to stay withing relatively safe voltages though.

I just got my C0 to 4 GHz, but that is temporary, I hope to swap it for a D0 when I get a chance, so I can get higher on lower volts.

The cycles per second are in my favor, however, as I have not the use currently for HT, so I leave this off. Therefore, higher clocks are achievable!

Also, another thing that the 1366 has going for it is the features on some of the mobos. I <3 my UD5, 10 Sata ports, plus it supports 3 way SLI or xfire.

Right now I only have 1 GTX 260, but oh the possibilities. Might get a 2nd in the future.
Or get 2 4890's, who knows?

Though, TBH, the gigabyte UD6 in that review did look pretty cool.

 

[taltamir]

pretty cool. It is a rather fun hobby, to wrestle every last ounce of performance and efficiency out of a system. Over or under, you are always seeking to fully actualize the potential of your hardware.

 

[IntelUser2000]

Originally posted by: Idontcare

Sure the trace paths are going to be less between the CPU and the NB/IGP with the NB/IGP being MCM'ed, but are we really thinking this is going to amount to more than a 10% improvement in IPC? (basically a speedbin's worth of performance increase)

Oh right. Did anyone mention that the results won't be directly comparable as the Uncore on the Lynnfield/Clarkdale will be only at 2GHz compared to 2.13GHz on the lower end Bloomfield and 2.66GHz on the higher one?

There's already a 3.06GHz Clarkdale bench out there. For comparison purposes Clarkdale achieved 10GB/s of bandwidth vs 6.5GB/s on the Core 2 Duo. Will that help with IPC considering the triple-channel Bloomfield got around 10% in single thread? Who knows...

 

[exar333]

Originally posted by: Shmee
I think we should talk overclocking. Personally, IDC how good a proc is @ stock, but rather how it is with a good OC. If the 1166 proc's cant overclock better than a D0, then they arent worth buying IMO. With a $200 D0, 6 GB memory ~$100-$135, and a decent mobo: ~$200-$500
and a good air cooler, you will likely get +4 Ghz on the 920 D0. If the i5's and the 1166 i7's OC as well or better, all the power to them. If not, I don't think I would be nearly excited about it.

I recommend to wait and see how they OC, and then compare how the overclocked processors perform, taking price, features and OVERCLOCKED performance into account.

Again, overclock is the keyword for me, and for many other enthusiasts. Since we have no news of how well 1166 OC, at least that I know of, I would wait and not get excited quite yet.

If anyone has evidence to the contrary, please, present it

+1. I could give a crap about the turbo levels of i5 because I would OC the CPU as high as it would go anyway. I still don't think you will get better than a moderately-priced X58 board and a i7 920 right now. Unless the i5 OCs regularly > 4ghz, I just don't think it is worth-it to wait unless you already have a good system (C2Q, PhII X4, etc) and can afford to wait it out for a while longer.

I checked the prices for an i7 920, and between online retailers and ebay, you can get one for anywhere between $200 and $270. There just isn't a better CPU for the money around, and with the plans to discontinue it this year, buy it while it's cheap. I really don't see this CPU getting cheaper than $200, but I could be wrong. Also, the cheap X58 market will probably shift to P55 and leave more expensive alternatives for 1366 in the long-run. For those that care, 6-core chips should be compatible with existing X58 boards after a BIOS update, and this would be a cool upgrade down the road. More cores may not give an appreciable advantage in some situations, but it would be fun to play with.

Edit: SP

 

[IntelUser2000]

Originally posted by: ExarKun333

+1. I could give a crap about the turbo levels of i5 because I would OC the CPU as high as it would go anyway. I still don't think you will get better than a moderately-priced X58 board and a i7 920 right now. Unless the i5 OCs regularly > 4ghz, I just don't think it is worth-it to wait unless you already have a good system (C2Q, PhII X4, etc) and can afford to wait it out for a while longer.

I checked the prices for an i7 920, and between online retailers and ebay, you can get one for anywhere between $200 and $270. There just isn't a better CPU for the money around, and with the plans to discontinue it this year, buy it while it's cheap. I really don't see this CPU getting cheaper than $200, but I could be wrong. Also, the cheap X58 market will probably shift to P55 and leave more expensive alternatives for 1366 in the long-run. For those that care, 6-core chips should be compatible with existing X58 boards after a BIOS update, and this would be a cool upgrade down the road. More cores may not give an appreciable advantage in some situations, but it would be fun to play with.

Edit: SP

The turbo will matter on the real world since most don't overclock. If it did not matter review sites would not show various CPUs of different clock speeds, but rather just a single version. It's a trade off of the days of "unlimited" heat and clock speeds of the Pentium 4 era and the "power-conscious" yet still performance-oriented days of now.

The problems with people touting S1366 compatibilities is that IF in case they need a new version of the motherboard to support it, their point is moot. I believe significant 1366 motherboards will still have issues with the 6 core versions. I've seen it repeated with various sockets before(yes and couple of times compatibility with next gen was mentioned).

 

[taltamir]

but if you DO overclock, than the i3 looks like the best deal.

 

[heyheybooboo]

Originally posted by: taltamir
but if you DO overclock, than the i3 looks like the best deal.

I should know better than to jump into a thread I am not following but my understanding is that the *i3* designation is the rebrand of the s775 Core processors.

The exception in i3 is a mobile Westmere 32nm variant designated socket 989.

I'm not responsible for any inaccuracies in this chart

 

[taltamir]

that completely contradicts every other source on the internet. including anandtech.

http://forums.anandtech.com/me...=2319722&enterthread=y

DESKTOPS:
an i7 for a desktop is defined as a quad core with hyperthreading (8 threads). s1366 and s1156
an i5 is either a quad core with no HT (4 threads) or a dual core with HT (4 threads), s1156 only
an i3 is like an i5, only with turbo mode disabled. s1156 only
Note: Laptop definitions are different than desktop definitions.

 

[three4seven]

This thread is an interesting read - particularly the posts from Idontcare.

While I think many of taltamir's arguments are valid - I think anyone reading his posts (in this thread atleast) can clearly tell his opinion is extremely biased towards i3/i5 - and maybe rightfully so. I am personally waiting to see complete real word tests and pricing before making any judgments.

I personally own a few i7 workstations - and if the i3/i5 is "better" (value or perfomance), I'll consider replacing them. But for now, my i7s are chugging along nicely.

 

[rchiu]

Originally posted by: Shmee
I think we should talk overclocking. Personally, IDC how good a proc is @ stock, but rather how it is with a good OC. If the 1166 proc's cant overclock better than a D0, then they arent worth buying IMO. With a $200 D0, 6 GB memory ~$100-$135, and a decent mobo: ~$200-$500
and a good air cooler, you will likely get +4 Ghz on the 920 D0. If the i5's and the 1166 i7's OC as well or better, all the power to them. If not, I don't think I would be nearly excited about it.

I recommend to wait and see how they OC, and then compare how the overclocked processors perform, taking price, features and OVERCLOCKED performance into account.

Again, overclock is the keyword for me, and for many other enthusiasts. Since we have no news of how well 1166 OC, at least that I know of, I would wait and not get excited quite yet.

If anyone has evidence to the contrary, please, present it

Yeah I doubt you will get better performance with i5+p55 combo, but you should get similar performance with $100~$200 saving compared to i7+x58 combo.

This release is not about ground breaking performance. It's about Intel had enough time squeezing all the money they could charging a premium for i7+x58 combo, and now they are bringing the price down with a mass market platform.

 

[IntelUser2000]

A theory to shatter conventional belief. Lynnfield might perform better comparatively to Bloomfield in dual GPU rather than on a single GPU. A review that I can't find now concluded better dual GPU scaling.

Maybe it makes more sense than the Lynnfield is worse in 2 GPU but better in 1 GPU theory. The latency might become more important than throughput with dual GPU because the GPU needs to communicate with each other. Essentially, with integrated PCI-Express controller the "sync latency" becomes lower.

Mere few % is enough to shift performance by 1 CPU bin. Then people might 2nd think about the Core i7 9xx.

 

[VirtualLarry]

Trying to view this thread on my cellphone, makes it reboot.

It's not just the length of the thread either, because normally, longer threads simply display an error message, "document too large". (I think my phone is limit to 64KB web pages.)