Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Why is E3-1275v5 more expensive then E3-1245v5 - when 1245 has better CPU score?
New on LowEndTalk? Please Register and read our Community Rules.

Why is E3-1275v5 more expensive then E3-1245v5 - when 1245 has better CPU score?

I just got a Xeon E3-1245v5 for my server at home. Then on another forum one comment that I did do a bad choice, not going for 1275v5 that cost around €70 more then the 1245v5.

I can see that the normal speed is 3.60 Ghz on the 1275v5 and 3.50 Ghz on my 1245v5.
The max speed is 4 Ghz on the 1275v5 and 3.9 Ghz on the 1245v5.

But still, when I look at the PassMark score, my 1245v5 has a score of 10295 and the 1275v5 has a score of 10277. So my 1245v5 is actually faster then the 1275v5.

Why do people think the 1275v5 is better then the 1245v5 then? The single thread score is also better on the 1245v5. Of course, it's not much, but still, it's faster.

I'm sure a 1275v5 server is more expensive per month if you rent it from a host also. So again, why?

Kenneth Myhre - free Windows templates for OVH/Hetzner/Kimsufi/

Powered by, backed up by OVH, Kimsufi and


  • rds100rds100 Member
    edited January 2017

    Probably not enough samples at for a reliable average. Look at the individual scores that they used to calculate the average for each CPU - see if there are any outliers.
    The 1275v5 should be the (slightly) better CPU.

    Thanked by 1myhken


  • am guessing you are using passmark as a guide? don't it's total nonsense never go by them. use something independant such as geekbench

    Thanked by 1myhken

    I AM BACK :)
    Working Windows Server 2012 R2 on 6GB! Beat that!

  •,88177 The higher clock is not the price worth.

    Thanked by 1myhken

    Hostbasic (aff)

  • Yes, I used PassMark as a guide for this check.

    Kenneth Myhre - free Windows templates for OVH/Hetzner/Kimsufi/

    Powered by, backed up by OVH, Kimsufi and

  • WilliamWilliam Member, Provider
    edited January 2017

    Realistically? Because AWS or someone else needed a specific CPU in large QTY the price dropped; you can see this on the flood of X56 (Chinese super computers) and E5-26 v1 CPUs (AWS, also Quanta switches) on ebay.

    The Intel prices are only relevant for a few weeks after release and far from what the largest customers pay, 3770S CPUs had 250$+ initial price, sold for 200$ nearly immediately and now you can buy them in lots from China for less than 100$ each. Excellent CPUs btw.

    The 75's selling point is that the turbo frequency is 4Ghz while it keeps the same 80W TDP/85W power budget - this means the DIE is more efficient than a 45.

    The DIE itself is nearly the same (and if you buy an i3 you get an E3 with 2 cores disabled, either by testing failure or stock availability; only this enables to offer a large range of CPUs at reasonable prices, else the DIE waste would drive up costs).

    Same applies to GPUs - a 1070 DIE is largely identical to a 1080 and the Tesla (enterprise) and Titan (prosumer) cards are very related as well. The only one here as outlier seems to be the 1050 Ti. The 980 Ti was pretty much a Titan card with less memory.

    A 390 is a 290, a 280X is a 7950 (nearly 100% identical) and a Fury Nano is the best-of-the-best chips of the Fury X/Fury while the Radeon Pro Duo (2 Fury essentially, but not PLX and rather multi-GPU DIE) gets the medium line as TDP is not that low set.

    Lastly, there ARE some operations that need extreme high single core clock - the 1275 is now not a good example but the X5698 would be: 2x 4.4Ghz at the same TDP as a 6x 3.7Ghz X5690 (130W), used for stock trading where overclocking (which yes, can easily reach this areas) is not usable (Intel guarantees correct working at 4.4Ghz - OC does not guarantee shit).

  • Passmark predicts the running speed of my own workloads very very accurately across cpus I've tried. Haven't really compared geekbench partly because it's hard to find a list of all the scores in one place. For passmark I use this:

    I'd love to see something like that for geekbench, if anyone knows of a place.

    Thanked by 1Junkless

    #lexit spread the word.

Sign In or Register to comment.