E5-2673 V3 OEM EBAY

Fleat

New Member
Feb 6, 2016
12
2
3
First time poster here, but there isn't a lot of information about this processor out there yet so I figured I would chime in.

My friend and I were both building new servers, and decided to gamble on this. We bought from a different eBay seller, but everything turned out perfectly. We both have the processors up and running on an ASRock X99 Extreme 4.

Passmark seems to be around 24-25k. Mine has been running for a couple of weeks now, and serving a bunch of VM's without any issues to speak of. Before buying, I sent the seller (out of Japan) a question asking about board compatibility, and he assured me it would work on the ASRock.
 

Fleat

New Member
Feb 6, 2016
12
2
3
Yes, this is not ideal if you strictly need high single threaded performance. It is ideal for highly parallelized applications though and not a bad workhorse for other uses as well.

I decided to go this route because power here is expensive, and a dual 2670 V1 box just wasn't practical power consumption wise for the same number of threads. This processor sips power in comparison in a single cpu format and I plan to run this for at least 4-5 years. I should come out ahead running 24/7 in just 2.5 years even considering the higher up front cost on the v3 setup.

Extrapolating a bit further, this cpu PassMarked around 24,000. According to the PassMark website, the 2670 V1 gets around 12,500.
PassMark - Intel Xeon E5-2670 @ 2.60GHz - Price performance comparison
Now we all know PassMark isn't a perfect representation of the actual performance you will see, but it does give you some sort of baseline to work with.

With that in mind, it seems like you are only losing a small amount in highly parallelized workloads to the much thirstier dual 2670 v1 setups and you are getting much newer technology across the board. Just my 2 cents anyways.

If anyone wants more detail or benchmarks with the processor, I would be happy to run them if the software is free.
 

chinesestunna

Active Member
Jan 23, 2015
569
152
43
54
How expensive is your power? Feels like unless you're running full load most of the time it would be hard to recoup $300+ price difference in chip cost alone, not even counting DDR4 vs DDR3 pricing difference of the platform
 
  • Like
Reactions: Fleat

Fleat

New Member
Feb 6, 2016
12
2
3
Ah, so this is where the numbers get interesting (for me anyways). My current power cost per kWh is 0.13111 so lets run some calculations with that. I used a kill-a-watt to get these numbers for the E5-2675 V3 so the accuracy is probably not perfect, but close enough for our uses.
E5-2675 V3 Idle Wattage: 52 watts
E5-2675 V3 Load Wattage: 160 watts

E5-2670 (x2) Review Idle Wattage: 200 watts
E5-2670 (x2) Review Load Wattage : 375 watts

Current Midwest kWh cost: $0.13111
E5-2675 V3 Annual Costs
Idle: $59.72
Load: $183.76
50/50: $121.74

E5-2670 (x2) Annual Costs
Review Idle: $229.70
Review Load: $430.70
50/50: $330.20

Additional Cost to run 2x E5-2670 V1 (E5-2675 V3 as 0)
Idle: +$169.98 ------------------ 3 Year Cost = +$509.94
Load: +$246.94 ---------------- 3 Year Cost = +$740.82
50/50: $208.46 ---------------- 3 Year Cost = +$625.38

Now, this doesn't seem that bad but unfortunately I am moving soon to somewhere with much higher energy costs. So lets just see how things look with the new 0.32111 kWh cost.

Future kWh cost: $0.32111
I won't go through the entire calculations... but things get pretty damn ugly here.
Total additional cost to run the dual 2670 V1 box at just idle year around is $416.31. Over three years, that adds up to an additional $1248.93.

With that said, I have heard some rumblings from various forum posts that a dual 2670 box can actually be around 100 watts at idle with all power saving features turned on. With that in mind, you can just halve the cost savings.

In either scenario, I would personally rather go with the newer platform. Also, just a quick disclaimer that this is all some napkin math so I could be off on some stuff.
 

chinesestunna

Active Member
Jan 23, 2015
569
152
43
54
@Fleat great analysis and for napkin math it's very reasonable. The E5-2670 idle power did seem high but that's based on Bit-tech's setup for that review and config. A few users here at STH got about 75-100W idle in the price tracker thread: 76W idle, 2x E5-2670 SR0H8, win 2k12r2
Anyways, the math would certainly swing more in v3's favor as power cost increases and at $.32 per kWh I would agree it's a good move.

Thanks for sharing again
 
  • Like
Reactions: Fleat

Weapon

Active Member
Oct 19, 2013
136
31
28
Interesting. Where are you moving that charges $0.32111 per kWh? Guessing outside of the U.S.?
 

Fleat

New Member
Feb 6, 2016
12
2
3
Yes, I will hopefully be moving out of the US within the next year. I also believe that your electricity can be that expensive within the U.S. as well (in Hawaii primarily).
 

Weapon

Active Member
Oct 19, 2013
136
31
28
Yes, I will hopefully be moving out of the US within the next year. I also believe that your electricity can be that expensive within the U.S. as well (in Hawaii primarily).
Even Hawaii is only $0.27 according to eia.gov

Thanks for the analysis. The v3 is tempting but electric is only $0.08 here so no way to justify for me.
 
  • Like
Reactions: Fleat

Fleat

New Member
Feb 6, 2016
12
2
3
I totally understand that, and it seems like you made the right call for your situation. If I had and would continue to utilize cheaper power, I probably would build a handful of those dual 2670 setups (if I could find cheap boards).

I also feel like it hits another niche that my friend who also purchased one falls in. He was okay paying a little extra to get the newer technology, and a more future proof socket. He also plans to use it for 5+ years so the annualized return on power savings increases the longer you own it.

I am sure this forum reaches a fairly wide audience, so hopefully it helps some people out there who have to watch the electrical meter like a hawk or have other reasons to consider additional options :)
 

T_Minus

Build. Break. Fix. Repeat
Feb 15, 2015
7,280
1,737
113
CA
"E5-2670 (x2) Review Idle Wattage: 200 watts"

That's a VERY HIGH idle.

And, fwiw, you can easily get into $00.3x rates in CA, I think it's Tier 3 now, and I imagine most people who have electric heat and AC will get that with the new way of calculating rates/tiers/times.
 

Fleat

New Member
Feb 6, 2016
12
2
3
Yes I agree, that seems much higher than most of the reports I see on the forums. Like I stated in my initial post, I think you can cut the power consumption costs in half (around 95-100 watts) if you are mirroring setups that other people are building and reporting values for. This can still lead to a significant savings if you are one of the unfortunate souls living with tiered pricing or in a high kWh state / country.
 

PigLover

Moderator
Jan 26, 2011
3,021
1,323
113
The EIA site is complete BS. It says CA rates are $0.1738/kwh. Maybe in a few small localities with quasi-gov electric utilities like Silicon Valley Power. But 90%+ of the state gets their power from PG&E, SCE or SDG&E and it's impossible to get more than a pittance of kwh from them for anything close to that.
 

T_Minus

Build. Break. Fix. Repeat
Feb 15, 2015
7,280
1,737
113
CA
Lol yeah @PigLover I think Tier 1 is .13 (or it was) but even us, who have no central air or any heat avg ~.24 :(
 

Evan

Well-Known Member
Jan 6, 2016
3,314
578
113
"E5-2670 (x2) Review Idle Wattage: 200 watts"

That's a VERY HIGH idle.

And, fwiw, you can easily get into $00.3x rates in CA, I think it's Tier 3 now, and I imagine most people who have electric heat and AC will get that with the new way of calculating rates/tiers/times.
That is not that high, I am sure the DL380's idle 180-200 running ESX (which is not very power saving). May log into work later to check I am not going crazy but sure that's what I remember.

Using a SM board and single PSU those idle figures in the other thread really surprised me how low. Since the 2670's are so cheap for a lot of people it makes sense.

Power in Singapore and Australia is ~20-30 cents per kWh (depending on tarrif etc and location or service provider)
 

T_Minus

Build. Break. Fix. Repeat
Feb 15, 2015
7,280
1,737
113
CA
We should all keep in mind that "idle" measurements can vary a good deal due to PSU(s) (# and efficiency), Motherboard(IPMI or not?), Chassis(# of Fans & Type), RAM Quantity & Type -- are the largest things to take into account for a 'barebones' idle.

When I check idle on a 2P board it's 4 Sticks of RAM per-CPU and likely with a desktop PSU as that's what I have on my bench. I know SM Fans + SM PSU will add another 15-30w depending on the chassis and # of psus, etc...

And, just for fun since I have this on my bench right now:
X10SRH-CF + 10 Core 2011-r3 (all performance mode) 3ghz w/4x16GB DDR4 & 1 SSD, and 1 2500 RPM Noctua Industrial Fan & 1 Case Fan & 1 Heat-Sink FAN I'm idling at 70 Watts *EVil Grin*

And just because I had this from testing P3700 800GB :)
ScreenClip.png
 
Last edited: