BE-2350, Merom, or Core 2 Duo in a home server?

[ninethirty]

Questions much like this have been asked several times, so I've been reading furiously the last few days trying to do my homework. But there's enough out there that I'm bound to have missed something, so I'll spell out what I've been able to gather or assume, and hopefully someone will point out what I've missed.

The challenge:
I just got married, and now I'm trying to consolidate several machines into one smaller, lower power, neatly-cabled [cat-friendly] machine. It will act as a MythTV frontend and backend crunching HDTV (720p, not full-on Bluray movies), web server, firewall, router, and slimserver. I'd like it to sit on a bookcase in the living room, so smaller & cooler is certainly better. And it's on 24/7 and usually just serving the occasional webpage and routing, so lower idle wattage is better.

The options:
Coming in I'd assumed that I'd be looking at C2D since it was the most efficient (power per watt) -- for instance, Anandtech's review of the 45w AMD chips showed the e4300 was still a lower-power system. But it's frequently quoted around here that the AMD x2 can idle lower, especially if undervolted. So, here's my options:

• A BE-2350 chip running in a NForce 6150 motherboard -- the ASUS M2NPV-VM has TV out, which would save me from needing a separate card. Lower idle wattage, undervoltable, but would it be fast enough for 720p with MythTV?

• A E4500 running in a low-power motherboard, with a separate graphics card. Anandtech seems to think they idle lower than the AMD, but most other reports around here seem to indicate that they'll idle higher, and won't undervolt significantly. So, I'm gathering that their higher performance-per-watt would be a waste in my case?

• A Merom chip running in a 945GT or 945GM motherboard -- there are a couple of compelling motherboards for this chip, including a Tyan that is FlexATX. But 945GT is an old chipset, and I've heard the occasional bad report of GMA950 graphics and MythTV (can't adjust the overscan or somesuch?) Graphics issues aside, does the Merom really buy anything over the AMD? Higher peak performance, I would assume, but the power numbers look similar (at least in terms of TDP). If the Merom board needed a discrete graphics card (FX5200 or 7300LE or whatever), would that negate the difference?

So, the BE-2350 looks like it may be the best option -- thoughts? Did I miss something obvious?

Thanks!

[drees]

The idle power draw numbers listed on that Anandtech review seem very high to me, but their choice of graphics card and motherboard probably inflate those numbers a lot. For a better comparison where the motherboard choice is similar to what you're planning on doing, have a look at Tom's Hardware BE-2350 review. The BE-2350 system uses less power at idle and load than an E2160 system according to their tests. I don't know how the E4500 compares to the E2160 processor.

In most cases the AMD system should idle lower than the Intel counterpart.

For MythTV use, just make sure you use a Nvidia graphics card. Integrated cards should be better for low power draw.

The BE-2350 will be plenty fast for HDTV viewing. Make sure you enable CnQ in the bios and that CnQ is functioning properly (setup and run the cpuspeed daemon). If you install Fedora 7, also install the powertop tool and you can use that to help reduce power consumption.

[vg30et]

If you do go for the BE2350 chip, I'd recommend a different/newer chipset such as the AMD690 or Nvidia 7050. The M2NPV-VM chipset is a bit more power hungry than the newer ones. I believe reviews showed a 5 watt or so difference at idle.

[Max Slowik]

Seconded. The video out of the MCP61, I think, actually makes video playback worse. A lot worse...I would go as far as to say that the video playback is actually replaced with an animated .gif of the goatse guy.

[ninethirty]

[Max Slowik] Seconded. The video out of the MCP61, I think, actually makes video playback worse. A lot worse...I would go as far as to say that the video playback is actually replaced with an animated .gif of the goatse guy.

Ha -- good to know. Ok, I won't drop down to the 6100. It looks like the 6150 is still separate from MCP61, though (at least based on some preview articles).

[vg30et] If you do go for the BE2350 chip, I'd recommend a different/newer chipset such as the AMD690 or Nvidia 7050. The M2NPV-VM chipset is a bit more power hungry than the newer ones. I believe reviews showed a 5 watt or so difference at idle.

Interesting. I wish I could go with the AMD690, as it gets good reviews in the SPCR and linux worlds, but I haven't heard any good reports of the on-board graphics and MythTV. A shame that ATI/AMD hasn't stepped up their acts more in the Linux graphics driver department -- they've been promising for years. I'll look into it a bit more, though...perhaps something has changed. Meanwhile, the 7050 is certainly an option. I've heard good reports of the Biostar 7050 board and MythTV.

Thanks for the responses, guys. Any Merom supporter want to weigh in? I think that's the main competitor here.

[ninethirty]

With a little further investigation, it looks like the Biostar TF7050-M2 may be the way to go. I've read a few newsgroup postings that don't make the AMD690 look good for MythTV.

I was skeptical about the BE-2350 from a performance/watt perspective, but a lower idle wattage may make up for it. I've heard varying things about undervolting the Merom (undervolting via BIOS disables EIST?) while undervolting the BE-2350 seems more encouraging. So the situation from the low-power perspective is actually the reverse of the high-performance perspective: the AMD chips are priced competitively with the Core 2 Duos performance-wise, but if you include overclocking it's no competition; the C2D will overclock by 100% or more. Conversely, Merom and the low-power AMDs have similar TDP specs (Merom with the edge), but it appears that including undervolting gives the AMDs can pull out ahead.
Is that too simplistic?

[drees]

I think you've summarized the situation well.

[ronrem]

The X2 3600 is VERY low price now and a lot of people have undervolted as low as .08V. At that level it's temps,power draw may pretty much match the BE chips.

Folks were using XP and P4 chips for Tv,media,Home server tasks,for years Any X2 or Core 2 has a LOT more capability than the GAMER heavyweights of just a few years ago.

[ninethirty]

Folks were using XP and P4 chips for Tv,media,Home server tasks,for years Any X2 or Core 2 has a LOT more capability than the GAMER heavyweights of just a few years ago.

It's true...for anything server-related, any XP-class chip would do fine. Shoot, a C7 would do fine. It's the MythTV part that I'm a bit fuzzy on -- folks have been playing media on boxes since P133's, but media en/decoding requirements have kept up with hardware, and nowadays decoding 1080i is out of the reach of most modern chips, without help from hardware decoders like the NVidia 8x00 series. Now, I'm just dealing with 720p, so hopefully my hardware will suit...but I'm running a test Myth install now with an existing X2 3600+ system, just in case.

[klankymen]

FWIW, the e4300 draws less than the BE-2300 (ad idle AND load), although it costs more.

Beyond money, there are simply better cheaper boards available for AM2 than C2D.

So I would go for amd.

[ninethirty]

FWIW, the e4300 draws less than the BE-2300 (ad idle AND load), although it costs more.

Really? Where did you read that? I had heard lower draw at load, but all the graphs I've seen showed the AMD drawing less at idle -- with the exception of the one Anandtech review I linked above.
I agree about the motherboards, though -- the G33 boards haven't impressed me as much as the Nforce 7050 and AMD790 boards for AM2.

[jojo4u]

FWIW, the e4300 draws less than the BE-2300 (ad idle AND load), although it costs more.

Intel refined the manufacturing, true. Both should be equal by now at idle on the CPU level, but I've yet to see a intel desktop platform with equal power draw to a 690G or 7050 platform.

[klankymen]

FWIW, the e4300 draws less than the BE-2300 (ad idle AND load), although it costs more.

Really? Where did you read that? I had heard lower draw at load, but all the graphs I've seen showed the AMD drawing less at idle -- with the exception of the one Anandtech review I linked above.
I agree about the motherboards, though -- the G33 boards haven't impressed me as much as the Nforce 7050 and AMD790 boards for AM2.

http://www.hardtecs4u.com/reviews/2007/ ... index5.php
and
http://www.hardtecs4u.com/reviews/2007/ ... ndex28.php

"Last" (red) means load...

[ninethirty]

http://www.hardtecs4u.com/reviews/2007/ ... index5.php
and
http://www.hardtecs4u.com/reviews/2007/ ... ndex28.php

"Last" (red) means load...

Interesting. Thanks for those links.
I had heard that the most recent steppings of C2Ds were more efficient -- anyone know if the undervolting situation has changed? As I understood, they were hard-limited in software undervolting, any undervolting via BIOS disabled SpeedStep/EIST. If that's true, an AMD is still the way to go for low power if undervolting is an option...

Thanks for the feedback, guys.

[accord1999]

Asus MBs (at least the P5B deluxe I used) won't automatically step down the voltage if you change the voltage from default or overclock. It'll still drop the multiplier. Of course, if you undervolt, it doesn't really matter.

[rpsgc]

Another comparative review.
http://www.digit-life.com/articles2/cpu ... -2160.html

It seems E2xxx CPUs are really lower powered than BEs.

[ninethirty]

Another comparative review.
http://www.digit-life.com/articles2/cpu ... -2160.html

It seems E2xxx CPUs are really lower powered than BEs.

Thanks a lot for that link. That was an interesting article. Especially interesting that the E4400 and E6x20 series were both lower voltage than the BEs at stock. It really does come down to the potential for undervolting.

Asus MBs (at least the P5B deluxe I used) won't automatically step down the voltage if you change the voltage from default or overclock. It'll still drop the multiplier. Of course, if you undervolt, it doesn't really matter.

That's what I'd heard; thanks for the confirmation. Any change from the default voltage disables speedstep. For the record, my AMD desktop motherboards do the same for overclocking, but I'm not sure if they do it for undervolting. Something to try.

It makes it something of a gamble -- the default Speedstep voltage settings are pretty good, but an AMD could potentially undervolt quite a bit better (down to 0.7 or 0.8 volts).

[slimeballzz]

Another comparative review.
http://www.digit-life.com/articles2/cpu ... -2160.html

It seems E2xxx CPUs are really lower powered than BEs.

Any idea how they measured power consumption?

I did not read the article but I looked at the test bed configuration and power consumption charts.

Is there any chance the motherboard and boxed cooler could have skewed the results?

[jaganath]

http://www.digit-life.com/articles2/cpu ... final.html

our power consumption measurement is based on Ohm's law, namely I=dU/R (1). In our case, strength of current (I), necessary to calculate power consumption, is obtained by measuring a voltage drop (dU) at the instrument shunt with the (R) 0.025 Ohm resistance. Then, the obtained value is inserted into formula (1) to get the strength of current in amperes. After that, we use one of the oldest formulas of electrical engineering, namely P=UI, to get power consumption at ATX12V. In our case, a voltage drop at the shunt (dU) is measured with a multimeter on the right of the photo, and the effective voltage in the circuit (U) – with a multimeter on the left. Voltages are measured with digital TRMS multimeters UT60E.

An evident shortcoming of this procedure is that it measures not CPU power consumption, but power consumption of VRM that powers CPU. Initially, we proceed from the reasonable, but not proved assumption that VRM is powered by ATX12V only. In its turn, ATX12V is used to power only VRM of a processor (it can be proved only by the schematic diagram of a motherboard under review, which is impossible in most cases). We decided to give this method a try and to accumulate statistics (which may become an empiric proof whether our assumptions are right or wrong), even though we are not sure it's 100% correct in all cases. But our readers should not forget what was mentioned above: we can call results obtained by this test procedure "approximate power consumption of a processor" only with a lot of reservations.