Celeron Dual Core E1200 preview

[Extelleron]

Originally posted by: 21stHermit

Originally posted by: VirtualLarry
I'd still buy a $60 E2140 over this thing though.

I'd buy the $85 E2180 over the E2140. The only way these low buck CPU's make any sense is on a $50 MB, which you can't OC.

I bet Intel's OEM customers are getting a really sweet deal on these CPUs though.

Since cache takes more die space than the core, can you imagine how many E1200 dies Intel gets on a 300mm wafer. A bizillion.

I'm fairly certain this is still Allendale core, which means it has a 111mm^2 die size and 2MB of L2 cache on die. Some of these E1200s probably have 2MB of functioning L2 and are just disabled, others will be cores that failed as E4xxx's or E2xxx's.

Cache actually doesn't take up too much space on the processor die as it's much denser than logic transistors.

A single C2D core with no L2 cache is 19M transistors / 36mm^2 in size, meaning that 2 cores will be around 72mm^2 for 38M transistors. Conroe has ~290M transistors in a 143mm^2 die, meaning half of it is cache and 248M~ SRAM transistors fit into the same space as 38M logic transistors.

It just isn't worth it for Intel to fab 3-4 different CPUs with different amounts of cache - it's easier and in the long run cheaper to make 1-2 different CPUs and just disable cache to make lower SKUs.

 

[21stHermit]

Originally posted by: VirtualLarry

Originally posted by: 21stHermit

I bet Intel's OEM customers are getting a really sweet deal on these CPUs though.

Since cache takes more die space than the core, can you imagine how many E1200 dies Intel gets on a 300mm wafer. A bizillion.

Yeah. I wonder if they are making a seperate die for the E1200, or if they are all dies with failed cache, from a higher-series CPU.

While I can't speak for Intel, I very much doubt that anywhere in Intel's process would they "reuse" partially failed higher series as a lower series. The screwing around to keep track of which go where would just not be worth the trouble. The enormous volumes would argue against such a tactic in any case.

However, I do believe Intel puts all the same multipliers on a single wafer. A simple test and laser castration makes the die a E2140 or E2160 or . . .

To me this is all CPU legend, unless of course your AMD converting failed Quads to Triples.

 

[21stHermit]

Originally posted by: nonameo

Originally posted by: 21stHermit

Originally posted by: VirtualLarry
I'd still buy a $60 E2140 over this thing though.

I'd buy the $85 E2180 over the E2140. The only way these low buck CPU's make any sense is on a $50 MB, which you can't OC.

I bet Intel's OEM customers are getting a really sweet deal on these CPUs though.

Since cache takes more die space than the core, can you imagine how many E1200 dies Intel gets on a 300mm wafer. A bizillion.

Not if it's an allendale core. The CPU is built with 2mb, and the rest is later disabled to make 21x series and 1xxx series CPUS.

Can you prove that?

I just don't believe Intel would waste all that real estate on unused cache. The Allendale core simply refers to the computational core which requires extensive testing to assure it's defect free. Reusing that core with multiple cache sizes reduces R&D overhead and time to market.

 

[Tempered81]

Originally posted by: Extelleron

Originally posted by: 21stHermit

Originally posted by: VirtualLarry
I'd still buy a $60 E2140 over this thing though.

I'd buy the $85 E2180 over the E2140. The only way these low buck CPU's make any sense is on a $50 MB, which you can't OC.

I bet Intel's OEM customers are getting a really sweet deal on these CPUs though.

Since cache takes more die space than the core, can you imagine how many E1200 dies Intel gets on a 300mm wafer. A bizillion.

I'm fairly certain this is still Allendale core, which means it has a 111mm^2 die size and 2MB of L2 cache on die. Some of these E1200s probably have 2MB of functioning L2 and are just disabled, others will be cores that failed as E4xxx's or E2xxx's.

Cache actually doesn't take up too much space on the processor die as it's much denser than logic transistors.

A single C2D core with no L2 cache is 19M transistors / 36mm^2 in size, meaning that 2 cores will be around 72mm^2 for 38M transistors. Conroe has ~290M transistors in a 143mm^2 die, meaning half of it is cache and 248M~ SRAM transistors fit into the same space as 38M logic transistors.

It just isn't worth it for Intel to fab 3-4 different CPUs with different amounts of cache - it's easier and in the long run cheaper to make 1-2 different CPUs and just disable cache to make lower SKUs.

if its an allendale core, will cpu-z show allendale or conroe?

 

[Extelleron]

Originally posted by: jaredpace

Originally posted by: Extelleron

Originally posted by: 21stHermit

Originally posted by: VirtualLarry
I'd still buy a $60 E2140 over this thing though.

I'd buy the $85 E2180 over the E2140. The only way these low buck CPU's make any sense is on a $50 MB, which you can't OC.

I bet Intel's OEM customers are getting a really sweet deal on these CPUs though.

Since cache takes more die space than the core, can you imagine how many E1200 dies Intel gets on a 300mm wafer. A bizillion.

I'm fairly certain this is still Allendale core, which means it has a 111mm^2 die size and 2MB of L2 cache on die. Some of these E1200s probably have 2MB of functioning L2 and are just disabled, others will be cores that failed as E4xxx's or E2xxx's.

Cache actually doesn't take up too much space on the processor die as it's much denser than logic transistors.

A single C2D core with no L2 cache is 19M transistors / 36mm^2 in size, meaning that 2 cores will be around 72mm^2 for 38M transistors. Conroe has ~290M transistors in a 143mm^2 die, meaning half of it is cache and 248M~ SRAM transistors fit into the same space as 38M logic transistors.

It just isn't worth it for Intel to fab 3-4 different CPUs with different amounts of cache - it's easier and in the long run cheaper to make 1-2 different CPUs and just disable cache to make lower SKUs.

if its an allendale core, will cpu-z show allendale or conroe?

It should show Allendale, but if the version of CPU-Z wasn't built to recognize the processor, then it could be reading it incorrectly and display it as Conroe.

 

[Tempered81]

my 4300 is read in cpu-z as conroe cpu-z is 1.42 is. is that correct?

 

[Extelleron]

Originally posted by: 21stHermit

Originally posted by: nonameo

Originally posted by: 21stHermit

Originally posted by: VirtualLarry
I'd still buy a $60 E2140 over this thing though.

I'd buy the $85 E2180 over the E2140. The only way these low buck CPU's make any sense is on a $50 MB, which you can't OC.

I bet Intel's OEM customers are getting a really sweet deal on these CPUs though.

Since cache takes more die space than the core, can you imagine how many E1200 dies Intel gets on a 300mm wafer. A bizillion.

Not if it's an allendale core. The CPU is built with 2mb, and the rest is later disabled to make 21x series and 1xxx series CPUS.

Can you prove that?

I just don't believe Intel would waste all that real estate on unused cache. The Allendale core simply refers to the computational core which requires extensive testing to assure it's defect free. Reusing that core with multiple cache sizes reduces R&D overhead and time to market.

There are several reasons why this is done.

-It's more complex to manufacture 4-5 different cores for each series of SKUs. It's a lot easier to have only Conroe for the 4MB parts and Allendale for the <2MB parts, and then Conroe-L for the single-core CPUs. Also, it eliminates the possibility of producing too many of a certain CPU and being stuck with processors that won't sell. If Intel had seperate cores and had a bunch of 512KB CPUs sitting around that wouldn't sell, what would they do? With one core for three SKU ranges, this possibility is eliminated as the same core fits a variety of products.

-It increases yield. If you have an Allendale-2MB core for the E4xxx series, an Allendale-1MB core for the E2xxx series, and then an Allendale-512K core for the E1xxx series, then you just throw away any cores that don't have enough functioning cache. You'll be throwing away a bunch of Allendale-2MB cores that are fine except for some of the L2 cache having defects. By using these CPUs for the 1MB and 512KB parts, then you eliminate a lot of waste. It's the same thing AMD is doing with its tri-core CPUs; some quad-core CPUs have one core that is defective, so instead of throwing them away they sell them as 3-core CPUs.

-It wouldn't be much cheaper, anyway. Allendale with 4MB is 143mm^2, with 2MB it is 111mm^2. That means 2MB = 32mm^2~, then 1MB = 16mm^2, 1.5MB = 24mm^2. A fictional Allendale CPU with 512KB of L2 would be approximately 87mm^2 in size. That's not too much of a savings over 111mm^2 and not worth dealing with the problems above.

 

[Extelleron]

Originally posted by: jaredpace
my 4300 is read in cpu-z as conroe cpu-z is 1.42 is. is that correct?

It should be Allendale, all E4xxx processors are Allendale core. A way to tell this is by the stepping of the chip - your E4300 should be L2 or M0 stepping. Conroe would be B2 or G0 (there is a B3 but I think that's just Quads).

 

[Tempered81]

its "L2", but everest & Cpu-z call it a "conroe" what is the difference in conroe &* allendale

 

[NoobyDoo]

X-bit Review

First of all we decided to find out what would be the maximum frequency for our dual-core Celeron test processor at nominal voltage, which was in our case equal to 1.28V. The CPU overclocked to 2.96GHz. The result is overall quite logical, considering that top processors on M0 core stepping feature clock frequencies of 2.6GHz.

The second overclocking experiment was performed with the processor Vcore increased to 1.5V. In this case the CPU remained stable at 3.4GHz frequency. This is a typical and expected result, because the Core 2 Duo E4000 and Pentium E2000 processors based on the same core overclock to pretty much the same frequencies. As for the temperature, Celeron CPU working at this speed heated up to 64ºC under full workload.

All in all, the new Celeron E1200 processor turns out much slower than CPUs from other families built on Core micro-architecture. Even Pentium E2160 priced only $11 more (according to the official price-list), outperforms dual-core Celeron E1200 by about 17-18% on average ... Celeron 440 is about 12% slower than the new Celeron E1200.

However, despite all those things we have just said we can?t proclaim the new Celeron E1200 the best choice in its price segment. The thing is that dual-core AMD processors priced the same provide higher performance. Athlon 64 X2 4000+ outperformed Celeron E1200 in our tests in all tasks except those dealing with image and video processing that are not sensitive to L2 cache size.

 

[VirtualLarry]

Originally posted by: 21stHermit
While I can't speak for Intel, I very much doubt that anywhere in Intel's process would they "reuse" partially failed higher series as a lower series.

Sure they would. Wafers/dies cost $$$, and if they can re-use a die as a lower-spec part, they save $$$. They do it all the time, many semiconductor mfgs do that. GPU mfgs too.

 

[LostPassword]

well judging from that xbit review, it looks like it'll do good for a cheap video playback machine.
but for gaming, might as well step up to e21x0 series.

 

[coldpower27]

Originally posted by: Extelleron
It should be Allendale, all E4xxx processors are Allendale core. A way to tell this is by the stepping of the chip - your E4300 should be L2 or M0 stepping. Conroe would be B2 or G0 (there is a B3 but I think that's just Quads).

E4300 are L2 only, M0 is only used for E4400+ SKU's.

Originally posted by: jaredpace
my 4300 is read in cpu-z as conroe cpu-z is 1.42 is. is that correct?

CPU-Z sometimes can't read the processor properly, and may not get the codename right, but it should get the stepping right and as Ext said, E4xxx are Allendale only.

 

[harpoon84]

Honestly, at this price range, it doesn't make sense to save $10 or $20 to get such a castrated CPU. Both X2s and E21x0 are better options for a few bucks more.

It might start getting interesting when the Celeron models start at 2GHz+... but at 1.6GHz there are better buys out there.

 

[aigomorla]

Originally posted by: Zap

Originally posted by: Yoxxy

Originally posted by: Zap

Originally posted by: Yoxxy
I hope they build these into an integrated motherboard.

You mean like the Intel D201GLY2 board? It is a mini ITX board with integrated everything and a 1.2GHz Celeron M-ish soldered on, for around $80.

exactly what I mean, only now with dualcore goodness, and C2D IPC advancements.

Actually, just saw that Intel board/CPU combo on CostCentral for $66.

actually the only mini itx boards to take c2d is the socket P meroms. I have a htpc setup with a T7300 as the core on a mini itx.

absolutely love it...

the socket p boards are kinda expensive tho.


www.logicsupply.com thats where i goto for all my mini itx projects. Now i just need to think of a way to drop this htpc in my car.