is Xeon only option to have CPU without GPU?

[IntelUser2000]

See the part that says Processor Graphics ?? it takes time, resources and money to design it. Higher end user cost.

Such BS. Intel has over 50% of the marketshare in graphics and is #1. Those people pay for all development costs, because the benefits are worth it for the extra $5 or so integrated graphics adds.

They themselves have said that improved media and graphics in Sandy Bridge allowed them to increase ASPs(Average Selling Prices).

Let's put it this way. What if AMD didn't have integrated graphics? They would have had to sell at much cheaper prices than they do now. That makes the iGPU taking up close to 50% of the 220-240mm2 die space worth it, because some people buy the chips just because of the iGPU.

The same applies to Intel. Those people that's fine with running Sandy Bridge/Ivy Bridge graphics would have to pay the penalty of extra cost and power use if it didn't have the iGPU(it also means they may choose not to buy it because of those negatives). Maybe you have an enthusiast-desktop-user thinking. That's fine, lot of people in these forums have that.

You didn't get my point using the example I gave about the empty space on the full quad core Sandy Bridge die. That's fine, and its my fault for having long posts.

 

[ShintaiDK]

Or just look back on prices for chipsets with and without iGPU. At best we talk peanuts.

 

[pantsaregood]

If Intel removed their IGPs, what would they drop their prices to? From $30 to $300, they have every single price point covered.

Lynnfield i7s were released at $284 for the i7-860 and $562 for the i7-870. The i7-2600 launched at $294 and the i7-3770 launched at $278.

The presence of an IGP had no apparent impact on the cost of these units.

AMD's CPUs would be worth considerably less without the IGP. That isn't because the IGP is expensive to manufacture, but because the IGP is literally the only feature of AMD's current lineup that looks good.

On-die IGPs DO increase production costs, but those costs are irrelevant to the consumer. You pay for a segment of the market, not for the production costs.

 

[ShintaiDK]

iGPU is also nice if you dont have a discrete card for whatever reason (Like frying it after OC.). Or retire the CPU into a HTPC life

Forgot to add in my previous post. Non iGPU chipsets costed less than some non iGPU chipsets. AMD seems the same way too. 890GX cheaper than 890FX.

 

[Ben90]

Or just look back on prices for chipsets with and without iGPU. At best we talk peanuts.

I find it interesting the amount of anger that churns up over 10 dollars on these forums. We are talking about around 3 dollars worth of silicon and possibly the remainder of 10 on fab savings if our CPUs didn't have IGPs.

Something better to complain about would be the free drinks offered to employees as that hits the bottom line a lot harder than lost sales from the cost of an IGP.

 

[AstroManLuca]

It's nice to have an IGP even if you don't always use it. What if your graphics card dies? You have a backup so your computer stays usable while you wait for a replacement. If you don't have a spare, using the onboard graphics could be the only way to confirm the problem is with your GPU and not something else!

Motherboards come with numerous connections I'll never use. Why do I need onboard surround sound when, like many folks, I only use stereo? I don't use PS/2 either. And all those extra SATA and PCIe slots that just sit unused.

 

[Yuriman]

The problem is hot spots generated in small logic areas inside the core.

CPU cores (not CPU die) are becoming smaller and smaller and thus the logic blocks inside each CPU core, like Integer Execution Units are becoming smaller and thus generating more heat per mm2. You can have a bigger CPU die (large L3 that occupies more space than the four(4) CPU Cores) and yet you could still have temperature problems because your CPU cores would be very small and you will have hot spots and a non uniform power map.

All im saying is that a bigger die because of dead silicon (powergated iGPU) will not help the hot spot problem that manifest in the CPU Cores that much. Having the iGPU as dead silicon will not have any significant effect in the thermal characteristics of the CPU when you already have the IHS. Intel didnt install the iGPU to act as dead silicon for thermal reasons.

Also to note, the topography of the CPU Cores and L2, L3 etc plays a big role in thermal management and uniform/non uniform power maps.

With my overclock, the full load temps of the 2nd core (middle core away from the iGPU) is 12c higher than the 4th core (next to the iGPU).

The core at the other edge of the chip runs 5c higher than the one next to the iGPU.

It's worth something.

 

[Kenmitch]

The 2550K still got the transistors for the iGPU.

These are defective chips....Mine wont boot at 57x but 56x works at insane voltages. I feel cheated

Glad I got it for $180 out the door with Intel protection plan!

 

[ninaholic37]

With my overclock, the full load temps of the 2nd core (middle core away from the iGPU) is 12c higher than the 4th core (next to the iGPU).

The core at the other edge of the chip runs 5c higher than the one next to the iGPU.

It's worth something.

So having 5 iGPUs between each 4 cores would be even better? :sneaky:

 

[pantsaregood]

So having 5 iGPUs between each 4 cores would be even better? :sneaky:

Actually, yes. This would drastically increase the die size without too much increase in power draw. In the case of the IGPs being off, practically no increase in power draw.

 

[mikek753]

in this case
what IB would be most cost efficient to get?
yes, I'll disable GPU on it
I'll use air only - to get it close to 4.5 GHZ on about same voltage - no increase as IB doesn't like over voltage on air

so many numbers so easy to get lost

yes IB-E would be good alternative, but
1. will be avail the next year
2. might cost too much - not easy to justify that premium

tnx

 

[ShintaiDK]

in this case
what IB would be most cost efficient to get?
yes, I'll disable GPU on it
I'll use air only - to get it close to 4.5 GHZ on about same voltage - no increase as IB doesn't like over voltage on air

so many numbers so easy to get lost

yes IB-E would be good alternative, but
1. will be avail the next year
2. might cost too much - not easy to justify that premium

tnx

There is no IB-E, never will be.

Gaming 3570K, computational work 3770K.

 

[pantsaregood]

Overvolting Ivy Bridge on air works fine. Ivy Bridge dies are rated to run safely up to 105 C. You probably won't hit 4.5 GHz on stock voltage. It usually takes Sandy Bridge/Ivy Bridge a little extra kick to make it to 4.5.

I don't know why people are freaking out and making up conspiracy theories about Ivy Bridge running hot. It isn't a bad design - Intel just reduced die size by a lot more than they reduced power consumption. Bloomfield ran pretty hot too, and I don't recall anyone flipping over it.

Then there's the whole "bad paste job" mess. I'd say that theory had some ground if it weren't for the fact that Ivy Bridge still runs hot when delidded.

 

[coffeejunkee]

It's because we got spoiled by Sandy Bridge temps. Which can actually can get pretty toasty themselves with big oc, but Ivy definitely gets hotter.

Compared to Bloomfield/Lynnfield Ivy does pretty good, 920 at 4Ghz would usually get well in the 80's, 3570K at 4GHz stays under 65.