Ivy Bridge 3700K @ 77W TDP

[tatertot]

It appears Intel's 22nm process is coming along nicely...

Top bin Ivy desktop 4C/8T only uses 77W, down from 95W Sandy.

A bit different than GlobalFlounderings.

http://pctuning.tyden.cz/component/...-procesory-ivy-bridge-core-i7-3700k-pouze-77w

(Use google translate)

 

[pantsaregood]

Intel's naming convention is becoming such a clusterfuck.

3700K? Really?

First generation Core CPUs were Nehalem/Westmere.
Second generation Core CPUs were Sandy Bridge.
Third generation Core CPUs are Ivy Bridge and Sandy Bridge-E?
Fourth? Ivy Bridge-E?

Wouldn't it have made sense for the entire Sandy Bridge family to fall under second generation, leaving third for Haswell and Broadwell?

 

[IntelUser2000]

That's less of an issue. How do they expect to sell 4-core Sandy Bridge E when Ivy Bridge is right around the corner?

Oh well, this is the problem of basing on roadmaps. Maybe it'll make more sense by then.

 

[Edrick]

That's less of an issue. How do they expect to sell 4-core Sandy Bridge E when Ivy Bridge is right around the corner?

Nov 2011 for SB-E and April 2012 for IB. Yes they are close together, but I would not call it right around the corner. Especially for people who have been waiting almost a year for SB-E already. But I am just impatient.

 

[tatertot]

That's less of an issue. How do they expect to sell 4-core Sandy Bridge E when Ivy Bridge is right around the corner?

4-core Sandy-E = LGA-2011, high-TDP, 4-channel mem, 10MB L3 cache, no IGP.

4-core Ivy = LGA-1155, low-TDP, 2-channel mem, 8MB L3 cache, IGP.

Never understood complaints about model numbering schemes. It's like complaining about someone's phone number. It's just an identifier.

 

[IGemini]

Intel's naming convention is becoming such a clusterfuck.

3700K? Really?

First generation Core CPUs were Nehalem/Westmere.
Second generation Core CPUs were Sandy Bridge.
Third generation Core CPUs are Ivy Bridge and Sandy Bridge-E?
Fourth? Ivy Bridge-E?

Wouldn't it have made sense for the entire Sandy Bridge family to fall under second generation, leaving third for Haswell and Broadwell?

You're doing it wrong. SB-E is 32nm on a larger die. It isn't third generation, that's like saying Bloomfield was SB.

 

[jpiniero]

That's less of an issue. How do they expect to sell 4-core Sandy Bridge E when Ivy Bridge is right around the corner?

Well, my assumption is that the 3820 is broken 6C. They probably aren't expecting to sell many and it will be off the market long before Ivy arrives.

The chart pretty much confirms that clock speeds are going to be the exact same as their SB counterparts. If that's true, 77 W is disappointing given it's a full node shrink and only a very small increase in IPC. Perhaps the GPU is taking up more heat than they expected.

 

[tatertot]

The chart pretty much confirms that clock speeds are going to be the exact same as their SB counterparts.

It does?

77 W is disappointing given it's a full node shrink and only a very small increase in IPC.

OTOH, there's always a full node shrink, a decrease in IPC, and an increase in power... *cough* BD *cough*

 

[jpiniero]

It does?

Well, the 3700 can't have higher stock speeds than the 3820 can. That would be silly.

 

[aigomorla]

now u guys see why i am crying IVY IVY give us our dayam poison IVY~

The chart pretty much confirms that clock speeds are going to be the exact same as their SB counterparts. If that's true, 77 W is disappointing given it's a full node shrink and only a very small increase in IPC. Perhaps the GPU is taking up more heat than they expected.

uhhh... 77W disappointing... ur aware how many cores are on it right?
This aint a dual core cpu... u know how hard it is to shrink TDP, while increasing performance, even though a node reduction.

 

[tatertot]

Well, the 3700 can't have higher stock speeds than the 3820 can. That would be silly.

That's not sound reasoning. Of course it can. Different process, platform, etc.

And why stop with the 3820? Why not argue it can't be higher than the 3930? Which would imply lower than 2600K speeds...

 

[pantsaregood]

You're doing it wrong. SB-E is 32nm on a larger die. It isn't third generation, that's like saying Bloomfield was SB.

I know SB-E is 32nm. That doesn't change the fact that the first digit of the model number indicates generation.

Sandy Bridge is second generation Core-series. Intel is calling SB-E third-generation.

 

[tatertot]

Intel is calling SB-E third-generation.

I haven't seen an Intel document specifying a model number decoder....

 

[jpiniero]

That's not sound reasoning. Of course it can. Different process, platform, etc.

Intel could have chosen to use a different naming scheme for Ivy. They didn't.

And why stop with the 3820? Why not argue it can't be higher than the 3930? Which would imply lower than 2600K speeds...

The 3930 has 6 cores. That's why it's better. The 3700 and 3820 are both QC+HT.

 

[aigomorla]

Intel is calling Sandy-E third generation because its the third series to carry the i7 tag.


lol i miss when intel was still deciding on calling the 980X a i9 and when all cpuid's used to read it as i9.

people would go OMGWTFBBQ i9!?!?!?!?!!?

 

[frozentundra123456]

This is sort of off topic, but I am not clear about Intels core count roadmap. I am very impressed with SB, but when are they going to bring out a 6 core CPU for the high end desktop, that is not priced outrageously? I mean they can sell the 2500K for around 200-250. Couldnt they make a hex core for around 400.00?
Granted, bulldozer was a disaster, but with AMD already having an "8 core" chip on the market, wont intel have to at least go up to 6 cores for the mainstream enthusiast market? It seems like they have nearly maxed out IPC with SB, so wont they have to start increasing core count also?

Edit: while I was writing this someone did post about the upcoming 6 core chips, but I am still not clear if it is a "mainstream" chip or a super high end one. And I agree, the model lineup and naming schemes have me totally confused.

 

[IntelUser2000]

They explain the numbers every time they release it. For example for servers:

E7-8870(Westmere-EX)

8-8 sockets
8-Designation for Socket LS
70-SKU

For Sandy Bridge EP chips, the naming for them is as follows:

"E5-2600"

2-2 Sockets
6-Designation for Socket R
00-SKU

And Ivy Bridge is set as "3rd generation Core" processor, hence 3xxx naming. IMO we focus too much on desktops when we look at things like naming. On laptops, the gain for Ivy Bridge should be decent to huge, justifying the "3rd generation" naming.

Well, the 3700 can't have higher stock speeds than the 3820 can. That would be silly.

Actually I'm gonna pull what I said. Ivy Bridge 3700K can still beat 3820. Like how the Lynnfield beat most of the Bloomfield chips on a much cheaper platform.

3820 clocks:

Base: 3.6GHz
Turbo 2-4 core: 3.8GHz
Turbo 1 core: 3.9GHz

And its supposed to cost $294 or something, cheaper than 2600K. The 4 core clocks are better than 2700K with 3.5GHz Base/3.6GHz Turbo 4 core.

 

[aigomorla]

man 2012... when 4ghz overclock means nothing but squat....

:\

i remember when 3.6ghz was almost viewed as godly...

 

[Puppies04]

If you want 6 cores you will need to buy a 2011 socket mobo. I know you want a 6 core ivy bridge chip that will fit straight into your $150 mobo but it isn't going to happen. There is no reason for intel to produce a cheap hex core as there is nothing the average user does that can't be accomplished very well on a 4c4t or 4c8t chip.

TL: DR If a 3700k doesn't have enough grunt for you prepare to get spanked in the wallet department.

 

[Idontcare]

<- 4GHz Prime95 stable QX6700 in 2006 (under phase of course )

man 2012... when 4ghz overclock means nothing but squat....

:\

i remember when 3.6ghz was almost viewed as godly...

If 3.6ghz was godly then you were holding it wrong, or using puzzie water cooling

 

[tatertot]

Intel could have chosen to use a different naming scheme for Ivy. They didn't.



The 3930 has 6 cores. That's why it's better. The 3700 and 3820 are both QC+HT.

The 3820 has 4 memory channels and more L3. That's why it's better.

But feel free to infer that all 22nm LGA1155 3700 series clock speeds must be lower than a 32nm LGA2011 3800 series part clock speed if that floats your boat.

Clocks are not expected to rise much from SB1155 to IB1155. But it has nothing to do with the clocks of an LGA2011 part and the relative magnitudes of their different-series model numbers.

Using 81&#37; of the previous TDP while bumping CPU IPC & clock a small amount AND doubling IGP performance is pretty impressive... it translates to:
about a 35% increase in perf/W on the CPU side (1.1 / .81), and a huge jump on the GPU side.

 

[Nemesis 1]

Well, my assumption is that the 3820 is broken 6C. They probably aren't expecting to sell many and it will be off the market long before Ivy arrives.

The chart pretty much confirms that clock speeds are going to be the exact same as their SB counterparts. If that's true, 77 W is disappointing given it's a full node shrink and only a very small increase in IPC. Perhaps the GPU is taking up more heat than they expected.

The 77watts is only disappointing if you don't understand which node the microarch, was designed for . The 32nm is a good example . As SB was designed for that 32nm node. Thats were intel picked up the efficiency over the last generation.

I didn't realize you were debugging for intel or another party . Since you seem to know intels IPC for IVB you must have a IVB chip. Than you take and throw the IGP improvements out even considering its considerable improvements and the fact that if Open cl brings anything to the table by March 2012 those benny's will show up than and only than . Ya 77watts doesn't seem like much . But I see it as alot of wiggle room for intel to [create faster scews if Intels needs to] . Haswell is the Chip being designed for 22nm node this is were you will see the SB type improvements on the 22nm node as far as efficiency by likely 2 fold. I haven't looked yet But I have never read anything about IVB-E . As I understood IDF Haswell will cover the entire line . Down to Atom III. As for intel keeping clocks at same level as SB . That is brillant on intels part. I didn't realize another X86 processor was up and coming that would cause intel to be stupid and push clocks higher needlessly . That what being green is all about . It also in reviews if it O/Cs on air to a stable 5.3 ghz . That will be included in the reviews as a &#37; increase over last generation SB . Intel has learned well from the past. You do know Haswell is NOT a Next generation 4 do you not . Haswell will not use the 1155 socket. Likely it will be the The universal socket intel talked about a few years back . I do believe Some intel 10 core products today include ras features. On haswell intel can added or remove IGP cores as they please If intel used 2 IVB IGP cores on a haswell die they would likely need that 4 channel memory that is on the 2011 socket. I doubt that intel will use the 2011 socket on haswell but than again who knows. After intel haswell intel really should be in a position were new sockets aren't necessary as everthing should be ondie. If your AMD Haswell AVX 2 is something to fear as it will use both lower and upper to do Vec on integre apps. To lazy today to find the correct wording for the upper and lower and integre for AVX2 USING Vex prefix. There was a very good video link posted here on a X intel engineer that is now professor . Very good stiuff and the guy is so common sense its refreshing . I mean quiting intel because he refused to do something intel wanted him to do that was stupid and he was a man about it and said screw you I quit. Very refreshing,

 

[IntelUser2000]

I can't see why 77W TDP is such a bad news. It's not. It still leaves room for 10&#37; clock improvement, along with the IPC gains.

77/95 = 0.81

Now what does 10% reduction in voltage in give you? Oh right, 90% the clock speed and 81% of the TDP. And 22nm is supposed to bring 20-22% improvement in clocks at the same power.

1.22x0.9 = 1.1x clock AND 0.81x TDP.

So 3700K can look like this:

Base-3.7GHz
4 core Turbo-3.9GHz
Max Turbo-4GHz

 

[toyota]

Well, my assumption is that the 3820 is broken 6C. They probably aren't expecting to sell many and it will be off the market long before Ivy arrives.

The chart pretty much confirms that clock speeds are going to be the exact same as their SB counterparts. If that's true, 77 W is disappointing given it's a full node shrink and only a very small increase in IPC. Perhaps the GPU is taking up more heat than they expected.

77w is tdp and a stock 2600k only uses about 65-70 watts of its rated 95w tdp. so that means a 3700k might end up using only about 50 watts or so. for a faster cpu with a faster gpu its hardly bad at all.

my question is why such an odd number like 77?

 

[IntelUser2000]

77w is tdp and a stock 2600k only uses about 65-70 watts of its rated 95w tdp. so 15-20 watts less for a faster cpu with a faster gpu in it is hardly bad at all.

my question is why such a random number such as 77?

How is 95W and 130W not random? You take the improvements that's given.

Like I said 77W is what 10&#37; voltage reduction with square in relation to voltage results in. You can reach 95W with 2600K if you use power virus code.