Intel Xeon E5-2670 Deal and Price Tracking

[jeffmcjunkin]

I run Prime95 using all dual CPU 2670 cores at 100% load and 128GB of memory and I couldn't hear fans from the other side of my room. Got two cheap Coolermaster Hyper 212x coolers on a Gigabyte board.

Peter99 - what Gigabyte board, case, case fans, etc are you using? I'd like a quiet setup for home as well, and your Prime95 test sounds like it's pretty ideal.

 

[StammesOpfer]

Has anyone noticed hardware related to these CPUs are going up? A couple weeks ago, you could get dual 2670s and 192GB RAM for around $1,000 USD. Now, everything on ebay is $1,200 minimum. I'm wondering if supply is dwindling or sellers are marking up prices.

I don't know.
$270 (best offer) + $30 s/h Quanta S210-X22RQ 2U Bare Bones Server w/DAS2RSMBACO LGA2011 Sys Bd/HDD Trays
$255 + ~$20 s/h for 3x (192GB) 64Gb. Kit NT8C72C4NG0NL-CG : Nanya 8Gb. 2Rx4 PC3L-10600R
$110 for 2x INTEL SR0KX E5-2670 20M Cache, 2.60 GHz, 8.00 GT/s PROCESSOR CHIP- IN STOCK

$685 Looks pretty good. With dual 10Gbe SFP+ and an LSI 2008 SAS controller built in.

 

[Joseph Nunn]

PassMark Software - CPU Benchmarks - Multiple CPU Systems

So 2670s and the dual-socket boards and RAM to go with them have gotten so cheap that it has me wondering: starting fresh, is the savings up front worth going refurb 2x2670 vs. a single E5-2620v4? Obviously the dual-2670 setup will have more raw power, but it's also noisier, sucks more power, and considerably more outdated, not to mention lacking the warranty of new. These boxen I'm considering would be an ESXi 6.0 node and a NAS.

I debated recently going v4 vs 2670 v1 and I went v1 basically because its cheaper. $60 vs $417 per CPU is a big difference. Also the newer motherboard was more expensive, and the ddr4 ram was more expensive. So the price difference is significant.

If you are looking for the mostest for the leastest, the dual 2670 v1 is very hard to beat. The following links were very helpful in comparing the various SKUs, although not every v4 is represented yet.

PassMark CPU Benchmarks - Single Thread Performance
PassMark Intel vs AMD CPU Benchmarks - High End
PassMark Software - CPU Benchmarks - Multiple CPU Systems

Joseph

 

[jwegman]

If you are looking for the mostest for the leastest, the dual 2670 v1 is very hard to beat.

There wouldn't be over 2000 posts in this thread if it weren't true for a lot of people

 

[Joseph Nunn]

The Motherboards are not cheap but the rest of the value proposition is awfully compelling for a home server or workstation.

It all depends on your requirements and your budget. For me it is the most compute bang for the bucks that I can scrape together to run xeon phis on.

Next gen I may have to go with gpgpus or lots of cpu cores as the price to play with the new xeon phi is likely more than I will be able to justify.

I think the next big thing is going to be greater data bandwidth and speed while cpu's and gpu's continue to stall. 14nm is exciting this week but physics is a bitch. Big data and pixels scale fairly well. Most other problems knock up against amdahl very quickly.

I would also like to play around with an Intel Phi, but sadly missed out on the special $195 deal from Colfax Intel Xeon Phi 31S1P Special Promotion at the end of last year. I'm looking around for prices similar to the deal, but the cheapest I can find is ~$400 a board, where are you sourcing your Phis?

Before anyone says anything, if you click through the Colfax deal has ended.

Joseph

 

[jjoyceiv]

The following links were very helpful in comparing the various SKUs, although not every v4 is represented yet.

I've already been all over Passmark's lists lately, and while they're good, they're sadly missing the 2620v4. Obviously 2x2670 will be more powerful, but it's a number I've been wanting anyway.

 

[Joseph Nunn]

I've already been all over Passmark's lists lately, and while they're good, they're sadly missing the 2620v4. Obviously 2x2670 will be more powerful, but it's a number I've been wanting anyway.

The 2630 v4 is on the lists, its .1 GHz faster and has 2 more cores. With a bit of math you can easily figure out what a 2620 v4 should look like.

Joseph

 

[MountainDew]

I don't know.
$270 (best offer) + $30 s/h Quanta S210-X22RQ 2U Bare Bones Server w/DAS2RSMBACO LGA2011 Sys Bd/HDD Trays
$255 + ~$20 s/h for 3x (192GB) 64Gb. Kit NT8C72C4NG0NL-CG : Nanya 8Gb. 2Rx4 PC3L-10600R
$110 for 2x INTEL SR0KX E5-2670 20M Cache, 2.60 GHz, 8.00 GT/s PROCESSOR CHIP- IN STOCK

$685 Looks pretty good. With dual 10Gbe SFP+ and an LSI 2008 SAS controller built in.

Well you cheated! I've just done searches on complete builds. But good information nonetheless. Thank you!

 

[jjoyceiv]

The 2630 v4 is on the lists, its .1 GHz faster and has 2 more cores. With a bit of math you can easily figure out what a 2620 v4 should look like.

Joseph

I have my estimates (I went for the average performance boost from the v3 series to the v4 series), so yeah, good enough.

 

[Peter99]

Peter99 - what Gigabyte board, case, case fans, etc are you using? I'd like a quiet setup for home as well, and your Prime95 test sounds like it's pretty ideal.

Gigabyte GA-7PESH2, Corsair HX850i Platinum PS (runs in silent mode all the time), while the case is an old tower which I bought in 1997 that fitted the SSI-EEB format board on its removable try like a glove (7 MoBo holes aligned, incl one Baby AT hole, so I drilled four more). The case fans are no name 80mm. With this setup the case and CPU fans are not a problem (silent, cpu temps mid 50s degrees Celsius under load), but the memory temperatures are too high when running Prime95 (got Samsung ECC low voltage memory). The airflow is practically non-existent and that is shown in the IPMI memory stats (75C). Not keen on dremelling so moving the entire setup into a newer InWin case that supports EATX OOB and has a big silent 400mm fan above memory, plus can fit massive Scythe Mugen coolers if needed. With that setup I can stand 1 metre from the case and hear nothing. Total cost NZ$1200 ($850). Now if only Grid K1 would come down in price...

 

[Joseph Nunn]

I have my estimates (I went for the average performance boost from the v3 series to the v4 series), so yeah, good enough.

Eh, not what I was thinking, as you have to treat estimating the single thread speed differently from the multi-threaded.

Specifically the former involves computing the ratio of clock speeds of the two processors (2630v4 and 2620v4) at single core max turbo and multiplying the single thread score of the 2630. The latter is similar to the previous solution but using full core max turbo for the first ratio times the multi-core score, and you additionally multiply the result by the ratio of cores of the two processors.

Or you can just blindly guess based on a v3 score of what I have no idea as it just makes no sense.

EDIT: Ah I see what you did there, you compared the 2630 v3 to the 2630 v4 using the average performance boost between generations. Thats an estimate to be sure, but you are comparing apples to oranges, I think you'd get better estimates using the ideas I sketched out above to compare intragenerationally.

Joseph

 

[Cipher]

I've just completed ordering the final parts for my 2670 build which will include the following:

-Supermicro SC846E16 chassis
-Supermicro X9DR3-LN4F+ motherboard
-2 x Intel E5-2670

Given this 4U chassis and the dual CPU configuration, is it possible to run these CPUs with only a heatsink and no fan? I've searched through this thread for terms like fanless and air cooling, but haven't found any information on this possibility.

If anyone is doing this, what heatsink model are you using?

 

[oasis475]

I've just completed ordering the final parts for my 2670 build which will include the following:

-Supermicro SC846E16 chassis
-Supermicro X9DR3-LN4F+ motherboard
-2 x Intel E5-2670

Given this 4U chassis and the dual CPU configuration, is it possible to run these CPUs with only a heatsink and no fan? I've searched through this thread for terms like fanless and air cooling, but haven't found any information on this possibility.

If anyone is doing this, what heatsink model are you using?

Hi,

I am running dual e5-2760 with passive heatsinks. Part number AUPSRCBTP.

My case has two fans but there are no fans on the heatsinks


Thank you

 

[Cipher]

Hi,

I am running dual e5-2760 with passive heatsinks. Part number AUPSRCBTP.

My case has two fans but there are no fans on the heatsinks

Thank you, Oasis475!

I've got two Supermicro heatsinks (snk-p0048ps) coming with my motherboard so I'll try those with the air shroud to see what temps are like when running some heavier tasks.

 

[RyC]

I have received the AXX3U5UPRAIL today. They look fabulous, and are of tool-less design. That means, you do not need to put the square rack nut/screw in order to install them.

The only issue I have is they are way too long. My rack is 36 inch deep. These, in their shortest state, is probably 38-39 inch long. So when they are installed with the rack, the rear end will extrude for about 4 inch outside the rack. At Intel's website, i didn't see the spec listed.

Now I probably have to turn another 90 degree so the rear end will face an open shelf and not be dangerous to anyone who may walk into them. But that will prevent I put the rear screen door on. Another rail I bought is exactly the same length: the AXXPRAIL, an upgrade for 1U/2U, even though the R2312GL4 system I got already came with basic rail. The basic rail's length is just right for my rack, although it lacks the nice, smooth operation from the AXXPRAIL. I guess I will have to live with those mistakes.

So anyone is interested in getting them -- make sure your rack is at least 40" deep, otherwise you will end with some really cool, smooth rails which are too long for the rack.

I actually a somewhat opposite problem, it seems the maximum mounting depth for the outer rails seemed to be only about 30.5 inches, while my rack was configured for about 31 inches between the front and rear posts. So I had to un-rack everything, adjust the posts, adjust the rails for everything else, and then re install everything.

Do you have pictures of how it's installed and can you confirm the mounting depth? The rails do extend a few inches in the back past the rear post mounts, which hits right against the fan wall in my cabinet, but it still fits.

 

[svt3391]

I actually a somewhat opposite problem, it seems the maximum mounting depth for the outer rails seemed to be only about 30.5 inches, while my rack was configured for about 31 inches between the front and rear posts.

Were you talking about the same AXX3U5UPRAIL, for the P4000 chassis? The front and the rear mount points can be extended further than 33 inches (my posts are 33 inches apart).

Unfortunately I don't have any picture right now. The past week had been hectic for work. This weekend I absolutely did nothing but decompress.

 

[zanechua]

Thank you, Oasis475!

I've got two Supermicro heatsinks (snk-p0048ps) coming with my motherboard so I'll try those with the air shroud to see what temps are like when running some heavier tasks.

I actually have this exact build running. Only difference is the heatsinks which are the snk-p0048p.

I am using the fan shroud that came with my CS846TQ but it doesn't seem to be fitting well with the new X9 motherboard. I had to remove a small plastic piece (probably used to redirect the airflow properly) and it fits better. Not perfect with the locks and all but it fits okay. I have all the original fans running, the 3 HDD fans and the 2 fans at the back.

My temps are around 40C and 44C, 40C being the one closer to the hard disks and 44 the one further away.

Not too bad considering the ambient temp where I'm at is around 30C. Quite decent I guess.

 

[RobertFontaine]

This is definitely the thread that keeps on giving.

I've been waiting almost 4 months for my RMA on my motherboard. Fubar.

There is a chance I might see my board this month from Supermicro but I am no longer holding my breath.

... So after you finally got your dual 2670 up and running... What did you build next?

... I haven't seen much chatter around good price performance of v3 chips or higher core/clock v1 or v2 chips.

... Are there any great values to be found I the next generation or higher performance chips yet?

... Has ddr4 ecc memory dropped low enough to be comparable to ddr3 yet?

 

[jwegman]

@RobertFontaine There's some good deals on Think Geek Canned Unicorn Meat

 

[RobertFontaine]

I will take this as a no.