ITT: We propose our own Intel SKUs

[Fjodor2001]

You keep successful brands around. Everyone knows the Pentium name.

They remember it from the days when the top end CPUs were named Pentium. Today most people have no clue where Pentium fits in with the Core iX CPUs. For example I know a lot of non-enthusiast people who think the Pentium CPUs are based on some old architecture from the pre-Core era.

 

[Dave3000]

I would like to see an i7-4890k with a base clock of 4.2GHz and a stock 4-core turbo boost ratio of 4.4GHz. The i7-4790k's stock 4-core turbo ratio is 4.2GHz and it's 4.4GHz for only up to 2 cores.

 

[jpiniero]

Yeah I definitely think Intel is due for a brand overhaul. Maybe soon enough.

 

[SAAA]

+1 to brand overhaul and modify the lineup like this:

i3 with 3 cores and differentiate with clockspeed or HT;
i5 with 5 cores and same segmentation;
i7 with 7/7+ cores and segmentation based on clockspeed/tdp, ht on all CPUs.

The i7 brand is used only for server or hedt, leaving the mainstream and performance segment to i5 brand.
This would give us much more performance at every pricepoint with minimal die size/tdp increase requirements. Also it will finally kill off the pure dual cores beside for any possible crippled part (using Pentium/Celeron brand) or mobile chips only.

I'm pretty sure that with 10nm something like this should be done else competition could catch up on MT performance or users will start complaining after 10 years on the same core count on mainstream...

 

[Atreidin]

Reduce the annoying segmentation. Same instructions across all lines so they have a better shot at actually being used, plus full virtualization and ECC support on all processors. Maybe just differentiate between non-server product lines with other things, like clockspeed, HT, graphics, TDP, and core count. Keep unlocked enthusiast processors but make the "K" version be definitively better without having to make stupid tradeoffs. It backfires anyway when people opt for one processor over another because of a feature that ends up being disabled. Keep multiple sockets and high memory support for Xeon line, maybe some other things I'm not thinking of, to keep their cash cows up there.

 

[jason166]

This++;

Stop with the instruction set level segmentation so developers actually have incentive to use these new capabilities. It's kind of like Tesla, making all versions of their car with the same battery but then artificially limiting it via software. If the only way you have to differentiate your product line is to deliberately cripple features that would otherwise work as is from the factory, then you are doing it wrong!

Reduce the annoying segmentation. Same instructions across all lines so they have a better shot at actually being used, plus full virtualization and ECC support on all processors. Maybe just differentiate between non-server product lines with other things, like clockspeed, HT, graphics, TDP, and core count. Keep unlocked enthusiast processors but make the "K" version be definitively better without having to make stupid tradeoffs. It backfires anyway when people opt for one processor over another because of a feature that ends up being disabled. Keep multiple sockets and high memory support for Xeon line, maybe some other things I'm not thinking of, to keep their cash cows up there.

 

[VirtualLarry]

They remember it from the days when the top end CPUs were named Pentium. Today most people have no clue where Pentium fits in with the Core iX CPUs. For example I know a lot of non-enthusiast people who think the Pentium CPUs are based on some old architecture from the pre-Core era.

Yeah. They need something new. Like "Quantium". Might be too close to "QuantiSpeed", which I think is one of AMD's trademarks.

 

[AtenRa]

One thing from me, remove the iGPU and add more cores at the same die size.

99% of Core i5/7 K series users never used the iGPU the last 3-4 years.

 

[jpiniero]

One thing from me, remove the iGPU and add more cores at the same die size.

99% of Core i5/7 K series users never used the iGPU the last 3-4 years.

Do people not understand the die validation cost issue? It wouldn't make sense from Intel's perspective, especially when OEMs wouldn't really want it. Probally 99% of the requests in this thread would fail the die validation cost test.

That, and of course Intel's desire to kill off the dGPU.

 

[AtenRa]

Do people not understand the die validation cost issue? It wouldn't make sense from Intel's perspective, especially when OEMs wouldn't really want it. Probally 99% of the requests in this thread would fail the die validation cost test.

That, and of course Intel's desire to kill off the dGPU.

This topic is "We propose our own Intel SKUs", I dont give a flying shit for what OEMs or Intel wants

 

[DominionSeraph]

This topic is "We propose our own Intel SKUs", I dont give a flying shit for what OEMs or Intel wants

"Everything in existence, and Intel pays me to receive it."

There. /thread.

 

[pw257008]

"Everything in existence, and Intel pays me to receive it."

There. /thread.

you forgot about all the things not in existence!

 

[cbn]

The 6C/6T Haswell-E should be priced at the same level as the i7-4790K.

Why don't you just say you want an i7 5960X for free?

In all honesty, I think the 6C/6T Haswell-E should actually be priced below the i7-4790K.

(This assuming the Intel Ark price for i7-5820K is $379.)

P.S. This wouldn't be the first time the lowest LGA 2011 socket processor was priced below the top LGA 115x i7 processor either. (eg, i7-3820 was actually cheaper than either the LGA 1155 i7-3770K or i7-2600K)

http://ark.intel.com/products/63698/Intel-Core-i7-3820-Processor-10M-Cache-up-to-3_80-GHz ($305 box, $294 tray for i7-3820)

http://ark.intel.com/products/65523 ($342 box, $332 tray for i7-3770K)

http://ark.intel.com/products/52214/Intel-Core-i7-2600K-Processor-8M-Cache-up-to-3_80-GHz ($326 for i7-2600K)

 

[DominionSeraph]

In all honesty, I think the 6C/6T Haswell-E should actually be priced below the i7-4790K.

(This assuming the Intel Ark price for i7-5820K is $379.)

P.S. This wouldn't be the first time the lowest LGA 2011 socket processor was priced below the top LGA 115x i7 processor either. (eg, i7-3820 was actually cheaper than either the LGA 1155 i7-3770K or i7-2600K)

The i7 3820 was a quad core and priced with them. You want a hex core priced with a quad core. Not only that, you want the additional features of X99 at Z97 prices. So I ask again, why don't you just say you want an i7 5960X for free if you're going to go that route and skip all the BS?

 

[cbn]

The i7 3820 was a quad core and priced with them. You want a hex core priced with a quad core.

i7-3820 was a quad core (without iGPU) priced below LGA 1155 quad core chips (with iGPUs).

Now the situation is even more lopsided as far as die area goes because Intel keeps adding even more iGPU to those quad core desktop chips.

So therefore, at this point, I think it is more than reasonable for a 6C/6T chip without IGPU (built in the future) to cost less than a 4C/8T processor (with iGPU bloating die size) that was released a couple months ago.

In a nutshell, I want my benefit from Moore's law to go towards adding cpu cores rather than adding iGPU.

 

[DominionSeraph]

i7-3820 was a quad core (without iGPU) priced below LGA 1155 quad core chips (with iGPUs).

Now the situation is even more lopsided as far as die area goes because Intel keeps adding even more iGPU to those quad core desktop chips.

So therefore, at this point, I think it is more than reasonable for a 6C/6T chip without IGPU (built in the future) to cost less than a 4C/8T processor (with iGPU bloating die size) that was released a couple months ago.

i7 3820 is 294mm2
i7 2600k is 216mm2

What's that about IGP bloating die size?

 

[cbn]

i7 3820 is 294mm2
i7 2600k is 216mm2

What's that about IGP bloating die size?

Okay so there is other non-iGPU logic on the i7-3820 not included on the i7-2600K.

So is this "other non-iGPU logic" on LGA 2011 chips increasing at the same rate as the iGPU size on the mainstream quad cores?

If the answer is no, then I certainly think 6C/6T processor without iGPU (released in the future) should be priced less than a 4C/8T processor with iGPU (released in the past).

P.S. Regarding die size costs on 32nm, I will say that the i7-3820 was released one year (Q1, 2012) after the i7-2600K (Q1,2011). So although the die size was larger on i7-3820, the node was certainly more mature (improving yields). This same relationship (possibly to lesser extent, depending on timing) should also apply to our 22nm processors. A 22nm 6C/6T chip released in the future should have a lower per xtor cost than a 4C/8T with iGPU chip released in that past.

 

[SPBHM]

i7 3820 is 294mm2
i7 2600k is 216mm2

What's that about IGP bloating die size?

sandy bridge IGP was not as huge as the current ones compared to cores

3820 had to use die space for over 2x pcie lanes and 2x memory channels I guess.

 

[Piroko]

Mostly has been said before:

• Full instruction set across all SKUs
• Kill Celeron and Pentium branding, replace with i1
• Remove at least half of the current SKUs, drop either T or S segmentation completely.
• Maybe remove the first two digits of current naming convention and replace with model year
• Reduce chipsets to two, "economy" and "enthusiast"
• Do a proper market (not marketing) differentiation between LGA and BGA
• Make sockets more future proof

I can't believe that in 2014 companies this large still have so many problems properly communicating to their customers what they're selling.

 

[jpiniero]

Full instruction set across all SKUs

Not happening.

Kill Celeron and Pentium branding, replace with i1

As I mentioned, Intel could use a branding overhaul, but I think it wouldn't be something like that. Considering Atom is substantially cheaper for Intel to make, it would make sense to replace the Core Celeron and Pentium models with Atom but give it a new name.

Remove at least half of the current SKUs, drop either T or S segmentation completely.

You do realize there's a point to the S and T models? It's for OEMs so they know what kind of cooling they need due to the TDP.

Do a proper market (not marketing) differentiation between LGA and BGA

The chips themselves aren't any different. Sure the latter is soldered onto the motherboard.

Make sockets more future proof

Not happening.

 

[crashtech]

In some ways I think the obfuscation is intentional to help retailers sell crappy products. In fact, when looking over spec sheets for lower end desktops, the model number of the CPU is sometimes left out. Turns out these are usually the ones with those J series tablet CPUs.

 

[Shivansps]

i1 and i2 instead of Celeron-Pentium its ok...

 

[Piroko]

You do realize there's a point to the S and T models? It's for OEMs so they know what kind of cooling they need due to the TDP.

In theory yes, however nothing should stop Intel from doing the same that AMD did with its A10 7700. Also, that's not an explanation for the retail availability of those SKUs.

Not happening.

That's a sure answer to lose customers to [any competing brand] unless there is an actual technical reason, which there is not.

I didn't post this as a joke:

I can't believe that in 2014 companies this large still have so many problems properly communicating to their customers what they're selling.

Intel has lost a lot of momentum with regular customers. Just look at how Tablets are shopped ("Price? OS? Storage?" in that order) compared to Laptops ("Fast enough to run -x-? Graphics? Price? Battery life?" also in that order). There should not be a need to ask these first two questions as, well, the first two questions unless you shop for something special.

 

[cbn]

sandy bridge IGP was not as huge as the current ones compared to cores

3820 had to use die space for over 2x pcie lanes and 2x memory channels I guess.

....and I have to wonder how much larger that iGPU will get with the introduction of DDR4?

 

[Maximilian]

Yeah. They need something new. Like "Quantium". Might be too close to "QuantiSpeed", which I think is one of AMD's trademarks.

Thats a blast from the past! AMD Athlon XP 3200+ with Quantispeed architecture!

Always wondered wtf "Quantispeed architecture" meant. The AXP was K7 architecture as far as I know.