Intel Optane 905P 1.5TB on sale $299

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

nexox

Well-Known Member
May 3, 2023
1,959
975
113
is there any way to justify this for a gaming or regular workstation?
Realistically unless you're running databases or other workloads with a lot of sync writes, Optane isn't really noticeably better than a half-decent NAND SSD for general storage.

That said, I love my Optane drives and am really hoping this deal expires before I give in... running two in RAID1 would be much easier than getting around to actually implementing proper backups for the FusionIO card where I store all my CAD work. (I know RAID isn't a replacement for backups... but when you're running a log structured filesystem it kinda is.)
 

ribroc

ebay hardware hobbyist
Feb 16, 2022
30
29
18
Is that metadata-only... or metadata-and-small-files? 100GB seems like a lot for metadata-only and 15TB... unless you have millions and millions of small files?
Yes, that's with the small blocks files. The numbers are for my pool at a small block size of 128k or 256k size, sorry if that was unclear. If it was metadata only, I'd average 120GB/100TB, but then the scrub benefit would not be as dramatic.
 

Auggie

Active Member
Nov 26, 2022
112
140
43
is there any way to justify this for a gaming or regular workstation?
Wendell from Level1Techs gives a pretty decent overview of why to like Optane in desktops here. Unless you're chasing that last little bit of latency and responsiveness, a good modern drive will be close. There are even scenarios where the faster linear reads will be more beneficial than the super low latency. That being said, I use it as a boot drive for everything I can except mobile devices. I just wish P5800x drives were a bit more reasonable.
 

i386

Well-Known Member
Mar 18, 2016
4,846
1,894
113
36
Germany
is there any way to justify this for a gaming or regular workstation?
If you're using windows: open up performance monitor (the one from the task manager) and go to the disk tab. Under storage you can see the queue depth of the different devices in your system.
Run your "regular" workload and check what the highest number is that you see. The lower this number is the better optane will perform compared to a regular nand ssd. At something like >64 regular "nand" will be on par or even better than first gen optane (like these 905p)

For me and my workloads (java, dotnet and occasionally c++ solutions + large git repositories) optane is excellent.
 

unwind-protect

Well-Known Member
Mar 7, 2016
611
250
63
Boston
is there any way to justify this for a gaming or regular workstation?
Gaming - no.

Workstation - depends on the work. I can imagine that compiles of large systems speed up by a couple of percent. Putting your Chrome profile on one could be beneficial. I have my gaming machine Windows boot on Optane but never measured whether it actually does anything faster.
 

Weapon

Active Member
Oct 19, 2013
349
137
43
I caved, in for 2 for my upcoming Proxmox server. What's everyone using to connect these? Probably go with a PCI-E to U.2 adapter for cleaner cabling, but are there any good sources for M.2 adapters?
 

nexox

Well-Known Member
May 3, 2023
1,959
975
113
I caved, in for 2 for my upcoming Proxmox server. What's everyone using to connect these? Probably go with a PCI-E to U.2 adapter for cleaner cabling, but are there any good sources for M.2 adapters?
If you don't have some kind of backplane then the cabling issues happen at the drive side, either a slot or M.2 adapter on the motherboard doesn't make a huge difference (unless you go for one of the cards that directly mounts two U.2 drives, but that comes with cooling and fit issues.) One reason I'm convincing myself not to buy these is because even though I 3D printed a rack for 6x U.2 drives for my workstation case there's almost no way I can fit another two cables in there.
 
  • Like
Reactions: l.code

Thomas H

Member
Dec 2, 2017
95
77
18
55
Posting my fio benchmark results all set at QD1 comparing Optane 905P vs Optane P1600X vs Samsung 960 EVO vs HDD. My test setup is Supermicro 6029U-E1CR4T and Proxmox 8.2.

Test Drives:
Code:
Toshiba MG07SCA14TE, HDD, SAS
Intel Optane 905P 1.5TB (SSDPE21D015TA), SSD, NVME, U.2
Samsung 960 EVO 250GB, SSD, NVME, M.2
Intel Optane P1600X 110GB (SSDPEK1A1118GA), SSD, NVME, M.2
Benchmark Summary:
Code:
sda        (HDD):  read: IOPS=81, BW=326KiB/s (334kB/s)(95.6MiB/300005msec)         lat (usec): min=1847, max=214536, avg=12251.68, stdev=4798.69
nvme0n1   (905P):  read: IOPS=75.3k, BW=294MiB/s (309MB/s)(86.2GiB/300001msec)      lat (usec): min=10, max=4040, avg=12.70, stdev= 3.19
nvme1n1 (960EVO):  read: IOPS=13.4k, BW=52.3MiB/s (54.9MB/s)(15.3GiB/300001msec)    lat (usec): min=15, max=9589, avg=74.29, stdev=34.00
nvme3n1 (P1600X):  read: IOPS=93.5k, BW=365MiB/s (383MB/s)(107GiB/300001msec)       lat (usec): min=8, max=3999, avg=10.29, stdev= 1.82
Full results for Optane 905P:
Code:
# fio fio0n1
rd_rnd_qd_1_4k_1w: (g=0): rw=randread, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=io_uring, iodepth=1
fio-3.33
Starting 1 process
Jobs: 1 (f=1): [r(1)][100.0%][r=272MiB/s][r=69.6k IOPS][eta 00m:00s]
rd_rnd_qd_1_4k_1w: (groupid=0, jobs=1): err= 0: pid=9100: Sun Jun 23 09:39:49 2024
  read: IOPS=75.3k, BW=294MiB/s (309MB/s)(86.2GiB/300001msec)
    slat (usec): min=3, max=4034, avg= 4.05, stdev= 1.50
    clat (nsec): min=147, max=4030.6k, avg=8655.94, stdev=2751.48
     lat (usec): min=10, max=4040, avg=12.70, stdev= 3.19
    clat percentiles (usec):
     |  1.000000th=[    8], 25.000000th=[    9], 50.000000th=[    9],
     | 75.000000th=[    9], 90.000000th=[   10], 99.000000th=[   22],
     | 99.900000th=[   38], 99.990000th=[   43], 99.999000th=[   70],
     | 99.999900th=[  334], 99.999990th=[  392], 99.999999th=[ 4047],
     | 100.000000th=[ 4047]
   bw (  KiB/s): min=228840, max=359504, per=100.00%, avg=301559.66, stdev=25534.39, samples=299
   iops        : min=57210, max=89876, avg=75389.74, stdev=6383.62, samples=299
  lat (nsec)   : 250=0.01%, 500=0.01%, 750=0.01%, 1000=0.01%
  lat (usec)   : 2=0.04%, 4=0.05%, 10=95.86%, 20=2.94%, 50=1.10%
  lat (usec)   : 100=0.01%, 250=0.01%, 500=0.01%
  lat (msec)   : 4=0.01%, 10=0.01%
  cpu          : usr=16.30%, sys=83.66%, ctx=4638, majf=0, minf=81
  IO depths    : 1=100.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%, >=64=0.0%
     submit    : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
     complete  : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
     issued rwts: total=22598761,0,0,0 short=0,0,0,0 dropped=0,0,0,0
     latency   : target=0, window=0, percentile=100.00%, depth=1

Run status group 0 (all jobs):
   READ: bw=294MiB/s (309MB/s), 294MiB/s-294MiB/s (309MB/s-309MB/s), io=86.2GiB (92.6GB), run=300001-300001msec

Disk stats (read/write):
  nvme0n1: ios=22592062/0, merge=0/0, ticks=215251/0, in_queue=215251, util=100.00%
fio configuration used on all drives (queue depth QD1):
Code:
# from intel recommendations for optane
[global]
name= OptaneFirstTest
ioengine=io_uring
hipri
direct=1
size=100%
randrepeat=0
time_based
ramp_time=0
norandommap
refill_buffers
log_avg_msec=1000
log_max_value=1
group_reporting
percentile_list=1.0:25.0:50.0:75.0:90.0:99.0:99.9:99.99:99.999:99.9999:99.99999:99.999999:100.0
filename=/dev/nvme0n1


[rd_rnd_qd_1_4k_1w]
bs=4k
iodepth=1
numjobs=1
rw=randread
runtime=300
write_bw_log=bw_rd_rnd_qd_1_4k_1w
write_iops_log=iops_rd_rnd_qd_1_4k_1w
write_lat_log=lat_rd_rnd_qd_1_4k_1w
 
Last edited:

T_Minus

Build. Break. Fix. Repeat
Feb 15, 2015
7,828
2,196
113
I've had optane in my desktop since they came out, getting ready to try a 2TB WD NVME SN850X instead... we'll see if I go back to Optane :D :D

For VMs there's no comparison still... even a pool of high-end NVME vs a pool of Optane the optane wins, but the margin is very-close.

I have some business-use "real world" tests on my "to do" for months comparing optane to other nvme for databases, web server, specific software, etc... problem is no time to build the real-world test setup :D :D but someday I'll post it and share.

With that all said... I still buy these 1.5TB when they go on sale LOL!!! 2x mirrored in server is such a great VM setup, and smaller ones 1x in lower-power systems makes them much much faster, and extends their life so much longer.
 

Auggie

Active Member
Nov 26, 2022
112
140
43
Mine came in blister pack only. No box, no m.2 adapter cable.

Not a big deal, but a box is pictured, and these came retail with an adapter that I was planning on using.
 

Aquatechie

Member
Oct 29, 2015
37
10
8
Mine came in blister pack only. No box, no m.2 adapter cable.

Not a big deal, but a box is pictured, and these came retail with an adapter that I was planning on using.
I can also confirm that you'll only receive the drive blister packaging. I, too, was hoping to get the adapter and cable included when I bought some during the sale from late last year.
 

Snorgle

New Member
Apr 18, 2020
12
13
3
it's hard to see on the page, but if you add it to your cart it should show up there
I'm guessing the gift card promo may be done...or I'm just not eligible because I wasn't signed up for the newsletter or something.