Discussion Performance per litre - Round 4

[hereforthefeast]

@dannycpham wow that z39 cooler is quite impressive, did you use the normal version or the "black" version that's from sff noir?

 

[dannycpham]

@dannycpham wow that z39 cooler is quite impressive, did you use the normal version or the "black" version that's from sff noir?

I used the SFF noir one. I wanted to buy the normal version, but they shipped me one without screws, AM4 bracket, and a used fan. . SFF noir shipped it out and it arrived within 3 days.

 

[Aux]

That PC is long gone so no unfortunately. I can submit a different entry

please do . .

 

[Aux]

That PC is long gone so no unfortunately. I can submit a different entry

Ok, I have inserted your result into round 3 shown below.

 

[guryhwa]

i have a project coming with EPYC 7742 and RTX3070 in 3L watercooled case with internal psu to break the PPL record

21.5cm * 19cm*7cm~2.86 L;with 7742and RTX3070:
estimated:

 

[rfarmer]

i have a project coming with EPYC 7742 and RTX3070 in 3L watercooled case with internal psu to break the PPL record

Looking forward to the Cinebench score.

 

[DrHudacris]

i have a project coming with EPYC 7742 and RTX3070 in 3L watercooled case with internal psu to break the PPL record

21.5cm * 19cm*7cm~2.86 L;with 7742and RTX3070:
estimated:

Please have mercy...

 

[guryhwa]

Looking forward to the Cinebench score.

and after a light OC

https://smallformfactor.net/forum/threads/g-unique-diy-stuffs-rome-wasn‘t-built-in-1-day-but-can-be-built-in-3litres-a-3l-64c-rtx3070-rome-itx-watercooled-internal-psu-ppl-beast.16678/#post-250554
updated case size is 21.8cm * 19cm * 6.5cm，2.69L

estimated Round 4 PPL:
65K(CPU CB23)*13K(RTX3070)/2.69L^1.5 /100

 

[Aux]

i have a project coming with EPYC 7742 and RTX3070 in 3L watercooled case with internal psu to break the PPL record


21.5cm * 19cm*7cm~2.86 L;with 7742and RTX3070:
estimated:

ooooooo

 

[codinghorror]

I was gonna enter the Zotac Magnus One but.. there's no way lol

Zotac Magnus One mods

I'm a huge fan of the Magnus One.. until I get my hands on Beast Canyon, I think it's the NUC to beat in terms of power to size ratio. I love Phantom Canyon but the RTX 2060 is a tad too underpowered for 4k these days, and that power brick.. sheesh. Excellent review at AnandTech...

smallformfactor.net

Case: Velka 5 Rev 2.0 177 x 99 x 285 mm = 4.99L ... vs ... 265.5 x 249 x 126 mm = 8.3L

And any Zen 3 is gonna clobber Intel on top of that!

 

[GuilleAcoustic]

I may need to bring up the bounding box issue in regards to the AC adapter, which has a cable strain relief notch on one side. Has a decision been made on whether or not this should be included in the dimensions, or should I post a photo so you can see what it looks like?

hmmm ... if you do, then you'll have to add wifi antennas to the bounding box or any "stiff" cable. AC cables have a quite big connector on the PSU side too ... it looks more like external brick hate to mee than real "scientific" issue.

 

[GuilleAcoustic]

LOLOL ... suavid was a spam (with suspicious links in his/her second post) ... everyone say goodbye to him/her

 

[timginter]

Great competition! I'll join with my Ryzen 5 3500X (brand spanking used from ebay) and Zotac RTX 2070 Mini, MadMax-Frankenstein build Time to take you on, @guryhwa




Really glad with above - my build is a few points higher than someone else's with same CPU - means my open Frankenstein doesn't impact performance or temps





Case maker: myself, name: Lapdesk

Spoiler: dimensions photos

Bounding box volume (rounded to 2 decimal places):

• 400.00 * 280.00 * 76.00mm (including the sticking-out M4 screw on top): 8.512 litres

I could have replaced the legs with M4 posix screws and pretend it's a vertical build... but let's not take ourselves too seriously.
I'm not counting the hinges and the screen, though - a PC wouldn't have an integrated screen anyway... give me a chance here.

1st calculation: ((6797 * 8679) / (8.512^1.5)) / 100 = 23,754.152
2nd calculation: ((6797 * 206.12) / (8.512^1.5)) / 100 = 564.144

If I re-arranged hardware more efficiently I could get volume down by almost 2 litres, but wouldn't be useable as a laptop. Still huge looking at other examples here, made me realise how small your SSF builds really are!

 

[timginter]

Do I get extra points for "personal heater" function? Temperature of the top plate, just above the GPU after Octane benchmark (GPU's backplate is touching the aluminium sheet) - 40 degrees Celsius:

 

[Goatee]

Never done this before, so please scrutinise and make sure I have this right:

• Picture of actual PC build = 1.989 L

• Picture of Cinebench R23 (30 mins) CPU benchmark result with CPU - Z

• Picture of Octane benchmark with GPU - Z

• Picture of Timespy benchmark result with GPU - Z
  • Time Spy on default preset and score is Graphics score, not total score.

• Case maker, name and exterior dimensions (including feet) to be shown to corroborate volume

REVOCCASES RCC-TIPIO1 - (256 x 185 x 42mm) = 1.989 L​

• Volumes calculated as bounding box
  • 1.989 + 0.188 = 2.177 (2.18 with rounding)

Brick - 101 x 69 x 27mm = 0.188L

• Round 4 PPL calculations
  • ((CineBench R23 score * Time Spy graphics score)/(Build Volume^1.5))/100
• ((6068 * 6063) / (3.2120) ) / 100 = 114,540
  • ((CineBench R23 score * Octanebench 4.0 graphics score)/(Build Volume^1.5))/100
• (6068 * 257.18) / (3.2120) / 100 = 4858

 

[Boil]

@Aux

I know Round Four has stagnated a bit, just a handful or so of entries...?

Thinking towards cross-platform benchmarks, looking at Blender Benchmark...

Thinking a full run of all scenes, once for CPU & once for GPU; or maybe run all scenes on CPU twice (back-to-back, no cool down) calculate results of second run (after everything is "warmed up" from the first run), same for GPU; using the most recent version of Blender from the drop down menu...

Now I need to see if there is a way to load the CPU & GPU at the same time, maybe there is a way to run two instances of Blender Benchmark concurrently...?

As for a formula to calculate a "score"...?

Maybe Round Four can "reset" & switch over to some sort of cross-platform (Windows, Linux, macOS) benchmark, and if the CPU & GPU can be stressed at the same time, all the better; get some heat soak on a tightly packed SFF chassis...

Thoughts...?

 

[CubanLegend]

and after a light OC

https://smallformfactor.net/forum/threads/g-unique-diy-stuffs-rome-wasn‘t-built-in-1-day-but-can-be-built-in-3litres-a-3l-64c-rtx3070-rome-itx-watercooled-internal-psu-ppl-beast.16678/#post-250554
updated case size is 21.8cm * 19cm * 6.5cm，2.69L

estimated Round 4 PPL:
65K(CPU CB23)*13K(RTX3070)/2.69L^1.5 /100

I look forward to seeing the final build pictures, and the actual timespy score. so you can submit it to the scoreboard.

I've been away from the forums and i didn't know there was a round 4, i thought Round 3 was it...?

 

[Damascus]

I'll hop in now that my S4T build has coalesced i7 6700 + MXM GTX 1070 on a b250m m-stx motherboard. Case is an S4T from NFC systems, 3.3L volume with an internal hdplex300w


Scores are:

• ((5053 * 5246)/(3.3^1.5))/100 = 55,551
• ((5053 * 133.64)/(3.3^1.5))/100 = 1503

 

[Damascus]

Double post

 

[CubanLegend]

Note this is my Brickless S4M-C from 2017, updated it in 2020 to brickless for PPL3, and now resubmitting for PPL4 as a curiosity. Time to see it age and be destroyed by newer performance-per-liter submissions, they'll maybe even inspire me for my NEXT upcoming gaming build...!?

PC itself:

CinebenchR23:

Octanebench:

Timespy:


Scores:

• ((6098 * 7438)/(4.2^1.5))/100 = 52695
• ((6098 * 158)/(4.2^1.5))/100 = 1119