Killer 16 core/32 thread Xeon processor server board build for $420

[zoomzoom]

Not sure if anyone is aware, but this article was one of the TechSpot notifications I got today.

http://www.techspot.com/review/1155-affordable-dual-xeon-pc/

 

[Jailer]

Just got done reading that one a little while ago. Pretty decent for the money if you need a lot of cores.

 

[zoomzoom]

$70 per processor is insane, especially considering what they originally retailed for. I did a double take as a I was reading it and had to go back and reread to make sure I read it right lol

 

[Jailer]

Someone recently was asking about building a system with a single one IIRC.

 

[zoomzoom]

I definitely understand the appeal, however once I got further down and saw they're 115w processors, I personally couldn't offset that con with what I'd use it for. Depending on where one lives, with all components hooked up, it'd be close to $1/day to run with just one processor, which doesn't take into account the amount of heat it would put off. I live in an apt, so this would probably not be as much of a concern with someone in a house, as a perfect place for it would be in a finished basement.

I'd be interested in hearing from anyone that does choose to buy one (or two) just to see what they build and use it for =]

 

[Kevin Horton]

I definitely understand the appeal, however once I got further down and saw they're 115w processors, I personally couldn't offset that con with what I'd use it for. Depending on where one lives, with all components hooked up, it'd be close to $1/day to run with just one processor, which doesn't take into account the amount of heat it would put off.

It only uses 115W when it is running flat out. It'll use a lot less that the vast majority of the time.

 

[zoomzoom]

It only uses 115W when it is running flat out. It'll use a lot less that the vast majority of the time.

Is that how most processors run? (I only ask because both my C2750 and C2758 utilize their TDP of 24w at all times, so I'm curious if that's [Xeons] the norm and certain newer processors are the exception, or vice versa.)

Sent from my Nexus 6 using Tapatalk

 

[Kevin Horton]

Is that how most processors run? (I only ask because both my C2750 and C2758 utilize their TDP of 24w at all times, so I'm curious if that's [Xeons] the norm and certain newer processors are the exception, or vice versa.)

As near as I can tell, personal computer CPUs from the Intel 80386 (released in 1986) onwards have had some sort of power management features to allow the CPU to use less power when it was not needed. Serve the Home testing clearly showed that the C2550, C2750 and C2758 used much less power at idle than when under load.

How did you conclude that your CPUs behave differently?

 

[zoomzoom]

The wattage usage display on my UPS; the 2750 is on the AsRock C2750D4I and the the 2758 is on the SuperMicro A1SRi-2758F. I've never see either dip above 24w, nor has either ever consumed less than 24w, granted the 2758F is running Sophos UTM, which isn't resource intensive. I know during large file transfers FreeNAS will shoot up from it's normal ~82w to 96w, which I've always assumed was due to disk usage (an increase of 1.75w per disk).

Once I have the chance to read the full article and not just skim it, I'll run some benchmarks as now I'm curious =]

 

[Kevin Horton]

The wattage usage display on my UPS; the 2750 is on the AsRock C2750D4I and the the 2758 is on the SuperMicro A1SRi-2758F. I've never see either dip above 24w, nor has either ever consumed less than 24w, granted the 2758F is running Sophos UTM, which isn't resource intensive.]

The Serve the Home testing found the C2750 and C2758 platforms (i.e. the whole system, including motherboard, RAM, network adapters, attached SSD, etc) using roughly 17W at idle and 33W when under heavy load. Your values are in between those, which suggests your usage isn't fully loading up the CPU to force it to use maximum power. It is impossible to know how much of your power is consumed by the CPU, and how much by all the other stuff, such as chassis fans, hard drives, SSDs, etc.

Note that their testing (if I am interpreting it correctly) found that the RAM was a very significant percentage of the total power consumed (28W of the 33W total platform power under load was due to RAM - this seems crazy high, so perhaps I am misinterpreting their results).

If your displayed power consumption never changes, I bet it is close to idle power. What is the C2750 running, and are there any hard drives or chassis fans on that system?

 

[zoomzoom]

Both specs and all hardware are in my signature:

• C2750D4I:
  • only 12 of the 16 HDDs are directly connected
    • 1TB 7200rpm 2.5" HGST (x3)
    • Samsung 120GB 850 Evo SSD (x1)
    • 4TB Seagate NAS 5400rpm (x8)]
    • Remaining 4 being an offline backup array that is disconnected except during the backups every month.
  • 32GB U-ECC RAM 1.35v
  • Vantec 4 core USB 3 PCIe

• A1SRi-2758F
  • only running a single 128GB Samsung 850 Pro SSD and 16GB of ECC RAM 1.35v
  • I'll eventually convert it from a Sophos only SA into a Sophos VM install since Sophos only allows for a max usage of 6 cores and 16GB RAM. I can't recall if this applies to UTM 9 or if it's only applicable going forward with XG (which won't be finalized until next year, as a substantial amount UTM functionality is still in the process of being ported over).

 

[Kevin Horton]

Both specs and all hardware are in my signature:

• C2750D4I:
  • only 12 of the 16 HDDs are directly connected
    • 1TB 7200rpm 2.5" HGST (x3)
    • Samsung 120GB 850 Evo SSD (x1)
    • 4TB Seagate NAS 5400rpm (x8)]
    • Remaining 4 being an offline backup array that is disconnected except during the backups every month.

I surprised the power consumption on the C2750 system is as low as you see. The specs for my 4TB WD Red drives claim they consume 3.3W each at idle. Your Seagates should have similar power usage. I wonder how accurate the wattage display on your UPS is.

 

[zoomzoom]

I surprised the power consumption on the C2750 system is as low as you see. The specs for my 4TB WD Red drives claim they consume 3.3W each at idle. Your Seagates should have similar power usage. I wonder how accurate the wattage display on your UPS is.

4TB Seagate NAS consumes:

• 4.8w on average (38.4w)
• 3.95w @ idle (31.6w)
• 0.5w @ standby (4w)

1TB HGST 7K1000 consumes:

• 1.8w on average (5.4w)
• 2.1w during seek (6.3w)
• 0.8w @ idle (2.4w)

120GB Samsung 850 Evo consumes:

• 3.5w - 4.7w on average

I'm not sure how much the the 4 core Vantec PCIe card uses, however with the 24w from the processor, it puts the wattage at 71.9w. Take into consideration the 3 fans and PCIe card and that should make up the ~10.1w difference.

Although I am baffled why you think the UPS may not display wattage correctly.

 

[jgreco]

Although I am baffled why you think the UPS may not display wattage correctly.

I am baffled why you would think the UPS would be a high accuracy watt meter, rather than merely a tool to help users make sure that they weren't overloading the UPS.

 

[Kevin Horton]

I'm not sure how much the the 4 core Vantec PCIe card uses, however with the 24w from the processor, it puts the wattage at 71.9w. Take into consideration the 3 fans and PCIe card and that should make up the ~10.1w difference.

Although I am baffled why you think the UPS may not display wattage correctly.

I was confused by one of your earlier posts, where you seemed to say that you never saw more than 24W on either the C2750 or C2758 boxes. Then you mentioned FreeNAS which would consume much more, but I interpreted that as referring to a third system. Now that I have deciphered your signature, I see that the C2750 box is running FreeNAS.

I use a Kill A Watt power meter to monitor power usage of my devices. It has a claimed accuracy of 0.2% of the full scale range.

 

[jgreco]

I use a Kill A Watt power meter to monitor power usage of my devices. It has a claimed accuracy of 0.2% of the full scale range.

They wish. I mean, not a bad device, great for the price, but not that accurate. 2%? I might believe that.

http://forums.anandtech.com/showthread.php?t=137169

etc.

 

[Bidule0hm]

It has a claimed accuracy of 0.2% of the full scale range.

RMS or TRMS?

 

[Kevin Horton]

RMS or TRMS?

They claim TRMS for at least some of their models. I'm not sure if that is true for all models or not.

 

[Kevin Horton]

They wish. I mean, not a bad device, great for the price, but not that accurate. 2%? I might believe that.

http://forums.anandtech.com/showthread.php?t=137169

etc.

I agree. Hence the inclusion of the word "claim". But, Consumer Reports says it is "accurate", whatever that means. It is almost certainly better than the indication on a UPS which is where this all started.

 

[zoomzoom]

I am baffled why you would think the UPS would be a high accuracy watt meter, rather than merely a tool to help users make sure that they weren't overloading the UPS.

If one is only looking to verify the wattage at a specific point in time, all wattage meters should be able to determine accuracy to within +/- 1w. I deal with Ohms law mostly in working on my car, but it applies regardless of dc/ac power.

To solve for P (wattage), V(olts) x I (amperage), V2/R(esistance), or I2 x R.

I was confused by one of your earlier posts, where you seemed to say that you never saw more than 24W on either the C2750 or C2758 boxes. Then you mentioned FreeNAS which would consume much more, but I interpreted that as referring to a third system. Now that I have deciphered your signature, I see that the C2750 box is running FreeNAS.

I should have articulated that better, as I do see how it could come across the way it did.