Intel Celeron E3x00

[Matija]

X-bit labs benchmarked the new Celerons.

Display: http://www.xbitlabs.com/articles/cpu/di ... e3300.html
Print: http://www.xbitlabs.com/articles/cpu/pr ... e3300.html

The new CPUs use 20W at full load. Overclocking at stock voltage pushed the E3300 from 2.5 to 3.3 GHz.

Overall very impressive.

Edit: I'll insert the power consumption graph here

By the way, we carried out a separate experiment to measure the power consumption of the CPU alone, independent of the rest of the system hardware components. In this case we measured the consumption along the 12 V power line connected directly to the processor voltage regulator on the mainboard. In other words, this measurement method took into account the efficiency of the voltage regulator circuitry.

[mczak]

X-bit labs benchmarked the new Celerons.

Display: http://www.xbitlabs.com/articles/cpu/di ... e3300.html
Print: http://www.xbitlabs.com/articles/cpu/pr ... e3300.html

The new CPUs use 20W at full load. Overclocking at stock voltage pushed the E3300 from 2.5 to 3.3 GHz.

Overall very impressive.

Edit: I'll insert the power consumption graph here

Those are nice cpus indeed, the lower power draw compared to E5200 is nice though I'd suspect that's mostly just because E5200 uses older core stepping. And I question the validity of the power draw measurements, they seem a tad low and inconsistent (for instance the e1600 draws 6W more than e5200 at the wall, but 2W less measured at the 12V line). Also, the X2 5600+ result is really really high - not that I'd expect it to be competitive but either they got the very old 89W version instead of the 65W version or the board doesn't actually use that 12V line only for the cpu (it would also be possible the intel board used other voltage lines for, say, FSB voltage, which will also add to cpu power draw).
Overvolting to 1.6V doesn't look healthy no matter the temperature but it sure is a very good overclocker.
If retail pricing of current E1600 celerons is any indication (expecting E3300 to sell for the same) it should however indeed end up with the same price as the Athlon II 240, I'd really wished xbitlabs would have used that for comparison (though, the E3200 will be the better deal than the E3300 - can just as well save a few bucks more if you're so low-budget since the performance difference is tiny, otherwise you should probably move up to the E6300 anyway).

[ilovejedd]

And I question the validity of the power draw measurements, they seem a tad low and inconsistent (for instance the e1600 draws 6W more than e5200 at the wall, but 2W less measured at the 12V line). Also, the X2 5600+ result is really really high - not that I'd expect it to be competitive but either they got the very old 89W version instead of the 65W version or the board doesn't actually use that 12V line only for the cpu (it would also be possible the intel board used other voltage lines for, say, FSB voltage, which will also add to cpu power draw).

Yeah, the derived processor power consumption doesn't seem to be all that accurate. The article specifically mentioned Brisbane so the 5600+ should be a 65W TDP part.

If retail pricing of current E1600 celerons is any indication (expecting E3300 to sell for the same) it should however indeed end up with the same price as the Athlon II 240, I'd really wished xbitlabs would have used that for comparison (though, the E3200 will be the better deal than the E3300 - can just as well save a few bucks more if you're so low-budget since the performance difference is tiny, otherwise you should probably move up to the E6300 anyway).

Yep, the E3200 looks like it's going to be a winner assuming you've already got a compatible LGA-775 board and an older processor. If just building new, an AMD-based platform still seems to be better bang for the buck (at least based on street prices). I'll probably upgrade my parents' Pentium DC E2160 to the E3200. Should make for a nice low-cost upgrade. Remind me again why we still need the Atom on the desktop?

[thejamppa]

Remind me again why we still need the Atom on the desktop?

Basicly we don't, that is just what Intel wants us to think...

[jessekopelman]

Remind me again why we still need the Atom on the desktop?

Basicly we don't, that is just what Intel wants us to think...

I'm sure Intel would rather you buy a Celeron, actually. I think it is the ASUS and MSI of the world that think you should buy there extreme-SFF Atom solutions.

[ilovejedd]

I'm sure Intel would rather you buy a Celeron, actually.

We can only speculate. Last I checked, 1Ku pricing for the Atom 330 was $43 - exactly the same as the E1400. Rumor is the Celeron E3200 will also be clocking in at $43. So the question is which one costs less to manufacture for Intel? The Atom 330 or the Celeron E3200?

I think it is the ASUS and MSI of the world that think you should buy there extreme-SFF Atom solutions.

I'm actually guilty of favoring extreme-SFF computers. Scratch that, I'm fascinated by ultra-tiny computers. I have a couple of ION and D945GCLF2-based builds and am planning a new build based on the D945GSEJT. Compared to how much VIA Mini-ITX boards used to cost, the Atom solution is damned cheap. However, I just don't see the Atom as a value solution like Intel has on their roadmap - at least not on the desktop. With current pricing, I see the Atom as being for a more niche market. Those who are willing to pay a (slight) premium and forego performance to get the absolute smallest builds.

For majority of users, though, Micro-ATX is small enough and given the choice between a similarly-priced Wolfdale and Atom, I think it's pretty much a no-brainer. If you add AMD which has both low-cost CPU and chipset into the picture, then the Atom makes even less sense. Cost of the new Atom would need to be lower than that of the Celeron- or Athlon-based platform for it to be practical.

[mczak]

Yeah, the derived processor power consumption doesn't seem to be all that accurate. The article specifically mentioned Brisbane so the 5600+ should be a 65W TDP part.

No not necessarily. I'm only talking about Brisbane, where the 5600+ existed both as 89W and 65W version. The 5600+ Windsor was indeed always 89W IIRC.

[Joe Public]

When Xbitlabs overclock their samples, they seem to have to increase their voltage so much. I remember when they reviewed a Celeron 440... To get it to 3 GHz stable, they had to increase Vcore to 1.5V. When I got mine, it did 3 GHz just fine at stock voltage.

[ilovejedd]

When Xbitlabs overclock their samples, they seem to have to increase their voltage so much. I remember when they reviewed a Celeron 440... To get it to 3 GHz stable, they had to increase Vcore to 1.5V. When I got mine, it did 3 GHz just fine at stock voltage.

Keep in mind that chips aren't homogeneous. While yours may have been a good overclocker, the sample they tested may not. The only way to get 3.0 GHz guaranteed is to buy a chip already at that frequency (e.g. E6850, E8400, etc). Otherwise, it's all luck of the draw.

Unfortunately, I don't think there are any processors with 4.0~5.0GHz stock clock available.

[micksh]

Finally, mini-ITX cases with crippled PSU like Antec ISK-300-65 will make sense when used with Zotac 9300 MB - this CPU would not have to push them to the limit. Just a little performance drop comparing to e5200, should be good for small HTPC.

Considering paired with the said motheroard it will be cheaper than ION Atom 330 MBs, so why would anyone want ION for desktop? For low-power mini-ITX servers use Intel 945 atom, for HTPC atom flash video performance is not good enough.

Intel is doing excellent job.

[jessekopelman]

I'm sure Intel would rather you buy a Celeron, actually.

We can only speculate. Last I checked, 1Ku pricing for the Atom 330 was $43 - exactly the same as the E1400. Rumor is the Celeron E3200 will also be clocking in at $43. So the question is which one costs less to manufacture for Intel? The Atom 330 or the Celeron E3200?

Cost to produce isn't the only factor. You also have to consider production capacity and the very high cost to retool or expand it. My understanding is that demand for Atom is exceeding capacity. That is good for Intel in that they can charge a greater premium for it. But even though it is cheaper to produce and thus higher margin, that does not mean Intel should automatically switch production from Celeron to Atom. The cost to expand Atom production likely exceeds the increased profitability of the chips. This type of thing is Intel's forte, even more so than the actual development of the chips. That they are still continuing with Core2-based chips is indicative that the cross-over has not yet been reached where it would be more effective to convert all low end products to Atom. The X-Factor here is actually i5/i7. Once that platform matures, all high-end products will migrate from Core2 and at that point it would probably be a no-brainer to switch the low-end to Atom (Atom2?).

[jessekopelman]

Finally, mini-ITX cases with crippled PSU like Antec ISK-300-65 will make sense when used with Zotac 9300 MB - this CPU would not have to push them to the limit. Just a little performance drop comparing to e5200, should be good for small HTPC.

Considering paired with the said motheroard it will be cheaper than ION Atom 330 MBs, so why would anyone want ION for desktop?

You want a case much smaller than ISK-300 and thus can't fit the kind of heatsink you'd need to cool a Celeron. At load, you are saving about 20W of heat with the Atom and in a very small case that is significant. Mobile of Desktop would also be a solution here, but if you don't need the processing power, Atom is much much cheaper.

[ilovejedd]

Finally, mini-ITX cases with crippled PSU like Antec ISK-300-65 will make sense when used with Zotac 9300 MB - this CPU would not have to push them to the limit. Just a little performance drop comparing to e5200, should be good for small HTPC.

Sorry, but aside from the heat issue as jessekopelman mentioned, it's not a guarantee that your 65W power supply will be able to support a build based on desktop parts even if total system power consumption maxes out at, say, 50W.

I requested specifications on an 80W DC-DC power supply board included with my Apex MW-100 case. Most of the current is on the 3.3V and 5V rails. The 12V rail could only handle 2.5A and that's where you'll get problems if using regular desktop parts. PicoPSU's work fine since, iirc, they're able to deliver their full wattage rating over the 12V rail.

@jessekopelman
Hmm, true. By the way, didn't Intel contract out part of the Atom production to another company?

[micksh]

Sorry, but aside from the heat issue as jessekopelman mentioned, it's not a guarantee that your 65W power supply will be able to support a build based on desktop parts even if total system power consumption maxes out at, say, 50W.

There has been a review of ISK300 posted on this site. It was done with e7200. It worked. New Celerons are not going to be more power hungry than e7200, much less, in fact, and that's the point. If you are questioning the review, you are posting to the wrong thread.

[micksh]

You want a case much smaller than ISK-300 and thus can't fit the kind of heatsink you'd need to cool a Celeron. At load, you are saving about 20W of heat with the Atom and in a very small case that is significant. Mobile of Desktop would also be a solution here, but if you don't need the processing power, Atom is much much cheaper.

Zotac IONITX-D-E costs $180 on Newegg. And Zotac 9300 costs $105 minus $20 rebate. http://www.newegg.com/Product/Product.a ... 6813500019. This Cleleron CPU will probably cost $45.
Now how exactly in your math the Atom is much cheaper? And how is it much much cheaper? I was talking about ION.

What do you mean by "you want a case much smaller than... "? Wrong assumption. Yes, there are some pre-built ION desktops and they are smaller than most of mini-ITX cases. But not that much smaller. And, most important, will I be willing to answer a question from my children - "why can't we see hulu show that my friends were watching"? No, I don't want to go into that conversation and I will rather chose a system that is a little bigger but it does everything.

"You want a case much smaller than ISK-300 "

NO! I dont! I want Mini-ITX that does the job which ION does not!.

[ilovejedd]

This post:

You want a case much smaller than ISK-300 and thus can't fit the kind of heatsink you'd need to cool a Celeron. At load, you are saving about 20W of heat with the Atom and in a very small case that is significant. Mobile of Desktop would also be a solution here, but if you don't need the processing power, Atom is much much cheaper.

was in answer to the question:

Considering paired with the said motheroard it will be cheaper than ION Atom 330 MBs, so why would anyone want ION for desktop?

Zotac IONITX-D-E costs $180 on Newegg. And Zotac 9300 costs $105 minus $20 rebate. http://www.newegg.com/Product/Product.a ... 6813500019. This Cleleron CPU will probably cost $45.
Now how exactly in your math the Atom is much cheaper? And how is it much much cheaper? I was talking about ION.

What do you mean by "you want a case much smaller than... "? Wrong assumption. Yes, there are some pre-built ION desktops and they are smaller than most of mini-ITX cases. But not that much smaller. And, most important, will I be willing to answer a question from my children - "why can't we see hulu show that my friends were watching"? No, I don't want to go into that conversation and I will rather chose a system that is a little bigger but it does everything.

"You want a case much smaller than ISK-300 "

NO! I dont! I want Mini-ITX that does the job which ION does not!.

You asked the question so he was just showing you some scenarios where an ION is preferable to a Mini-ITX mobo+Celeron E3000 solution - and that is in very, very small form factor builds. He's not saying that you want a tiny case. Maybe he should have added "If" at the beginning of the statement to be clearer.

jesse also mentioned mobile on desktop as an alternative to the ION, but if you thought the ION was expensive, wait until you see how much you'll have to pay if you go the MoDT route. The expense comparison was between the ION and MoDT solutions. Not ION vs GF9300+Celeron E3000.

I'm not questioning the ISK 300 review as I haven't actually read it until now. The case was too big for me so I just never bothered. I'd like to point out, though, that the DC-DC board seems to be better made than others found in a lot of Mini-ITX cases.

Antec ISK-300 DC-DC board
5V / 7A Max.
12V / 5A Max.
3V / 6A Max.

Note that the Antec power supply can handle up to 60W on the 12V rail. In comparison, the power supply that came with my case can only put out 30W max on the 12V rail.

Apec MW-100 DC-DC board
5V / 5A Max.
12V / 2.5A Max.
3V / 6A Max.

So while the power supply on ISK-300 is well-built and suitable for low-power desktop builds (props to Antec for that), there's a large number of Mini-ITX cases (mostly designed to work with VIA and Atom boards) whose included power supply is not suitable for builds using desktop parts. You made a blanket statement that "mini-ITX cases with crippled PSU like Antec ISK-300-65 will make sense", but as far I'm aware, the Antec case is the exception rather than the rule.

Please note that neither jessekopelman nor myself are advocating the use of the ION/Atom for normal desktop users. Just look at the earlier posts in this thread. The ION/Atom is just too slow for what a lot of people are using their computer nowadays (e.g. Hulu, etc). I would never recommend the ION for someone who's building their first HTPC. However, we do recognize that a market for it exists.

[micksh]

ilovejedd, thank you for explanation, I understand now. Sorry, I made my post look like it was rather emotional which was unnecessary.

[jessekopelman]

Yes, when I wrote "you want," I meant some generic person not micksh. I thought it was obvious that the "you" was not micksh, but we all know what happens when one assumes things. Perhaps what I should have written is the more formal, "one wants." In the US, at least, that sort of writing has fallen out of favor, but it does still have a purpose.