Jump to content
3dmark master

Atomic 3DMark & Game Benchmark Rankings

Recommended Posts

Radeon VII

2000 core
1126mv
+20 Power
1000 mem

100% fan
J max 97c
ambient ~25c

Graphics Score
30 827
http://www.3dmark.com/fs/18622210

2010MHz even up to 1200mv crashes
Auto OC = instant blue screen

Graphics score was ~200 lower with 1168mv
Actual clock speed was ~1900-1980

Share this post


Link to post
Share on other sites
Posted (edited)

22,378
Graphics Score    32,319
http://www.3dmark.com/fs/18631627

2045 core
1166mv (Fixed it was crashing instantly over 2000MHz with 2d v at default as it didn't ramp up quick enough)
+20 Power (peak power use 489w at wall)
1200 mem
J max 108c in Graphics test one thermal throttled to 1950-2000MHz while graphics test 2 ran 2000-2045MHz
Ambient ~10c

 

Superposition:  (1080p Extreme)

6173

2000 core 1138mv
1200 mem
100c J max

So just ahead of gambles 1080ti but it seems the 2080 does really well in this test.

Edited by Dasa
  • Like 1

Share this post


Link to post
Share on other sites

Great results!  Looks like that card has lots of potential.   Look forward to seeing more 🙂

After the new board I was able to run my ram at it's rated speed of 3466 but it requires 1.58v so running it at 3266 16-16-16-36 1T. 

Did a compare of the last few cards below.  I know your 1070 was a better OC'er but threw it up mine as well.  This is with my 2080 running as has much as it will go +165 (1995 to 2085c) / +1300 8300m.  Power limit is what holds it back, the temps with fan at 70% never hit over 67c so I dont think water would help out much for gaming or benching.  Especially since it runs sub 60c for much of the bench runs as they are short.  I can game at this setting but run around +70 2000c/+1000 8000m for daily as there just isn't much gain past that it seems. 

FireStrike:  https://www.3dmark.com/compare/fs/18631627/fs/18632700/fs/14716114/fs/13625478/fs/12182486

Graphics Score:
2600x/Rtx2080:  29187
1700x/Gtx1080 Ti:  29677
1700x/RX56(flashed to 64):  22098
1700x/Gtx1070:  19865

--

TimeSpy: https://www.3dmark.com/compare/spy/6510932/spy/3321791/spy/2719676/spy/1480735

Graphics Score:
2600x/Rtx2080:  11975
1700x/Gtx1080 Ti:  10418
1700x/RX56(flashed to 64):  7203
1700x/Gtx1070:  6340

Superposition:  7555

Share this post


Link to post
Share on other sites
Posted (edited)

I think I am close to having my 24\7 stable OC dialed in using MorePowerVII_77+ and quiet fan\pump curves that keep the GPU under 70C hot spot.

2100 Core 1202mV 1230 Mem

 

The increased SOC from this powerplay profile is what let's me hit 2100MHz the next step up (EvenMorePowerVII_1293++) only brought with it another 25MHz.

MorePowerVII_77+ https://www.overclock.net/forum/67-amd/1633446-preliminary-view-amd-vega-bios-131.html

+77%(+20%) max power limit, 400A TDC Core (330A), 66A TDC SoC (50A)
1193mV SoC (1168mV), max HBM 1300MHz (1200MHz)

 

Heaven Extreme: 4237

Valley HD Extreme: 5770

Superposition 1080P Extreme: 6528

Superposition 4K Optimized: 9013

FireStrike: 22,836 - Graphics Score 33,409 - Physics Score 14,455 - Combined Score 9121 http://www.3dmark.com/fs/19456389

FireStrike Extreme: 13,686 - Graphics Score 16,027 - Physics Score 14,445 - Combined Score 6296 http://www.3dmark.com/fs/19456438

FireStrike Ultra: 7697 - Graphics Score 8039 - Physics Score 14,424 - Combined Score 3602 http://www.3dmark.com/fs/19456461

TimeSpy: 9390 Graphics Score 10,479  CPU Score 5911 http://www.3dmark.com/spy/7309743

 

 

24-7.jpg

 

Edited by Dasa

Share this post


Link to post
Share on other sites
Posted (edited)

I didn't appreciate just how much of it is finding the right v for the set clock speed and just how much slower more v can be even with a higher set clock.
2175MHz core set for both the below
179FPS graphics test 1 1238mV <Crashed test 2 This would have mostly been running 2100MHz+
172FPS graphics test 1 1243mV <This would have been around 2050-2125MHz
FireStrike 23,333 Graphics Score 34,436 http://www.3dmark.com/fs/19496909

 

Edit:

Tried again this morning with the window open at 3c ambient so junction maxed out at 60c but didn't do any better
Got in a run set at 2221MHz 1271mV 1287MHz mem but only scored 34,353
Peak power use from the wall was 559w

 

Swapped back to 2100MHz 1212mV and it doesn't matter if power limit is at 0% or 99% the max power use from the wall remains at 509w in firestrike.
While I have seen up to 30% power limit affect clocks in other programs power use never went much over 500w.

 

For comparison
Firestrike grapics test 1
Max power use from wall
Stock 1143mV 367w
Undervolted 1001mV 315w
I did get sick of 100% fan speed though so these two will be ~20w lower due to that.

Edited by Dasa

Share this post


Link to post
Share on other sites

Chernobylite is a new early access game built on the Unreal engine with a benchmark.

This is at Ultra 3440x1440

c%20bench.jpg

Game has some promises and it will be interesting to see how it develops as they add extra content.

  • Yes Sir! Very atomic! 1

Share this post


Link to post
Share on other sites

Do you have to buy it?  Any demo?  Didnt find one in a quick search. 

Share this post


Link to post
Share on other sites

Looks like you need to buy it and even then you don't get much more than a demo for $40 at this stage.

 

As for the game it is fairly short just a few maps they will add more but it will never be a large game with just a few quick missions.

A fair bit of the outdoors FPS drops lower than it does in the benchmark.

Edited by Dasa

Share this post


Link to post
Share on other sites

New score with 7700K@5.1GHz -1AVX 1.375v, 4x8GB 3600c14, RadeonVII 2150-core 1275-mem.

http://www.3dmark.com/fs/20952799

24 408

  • Graphics Score 33732
  • Physics Score 16205
  • Combined Score 10 549

GPU score is a little lower with newer drivers the clock speed is more stable with less dips but cant hit the same highs although I am sure I could do better if I found the sweat spot for v vs clock again.

 

Here is a userbenchmark run at the same settings

UserBenchmarks: Game 119%, Desk 105%, Work 84%
CPU: Intel Core i7-7700K - 109.4%
GPU: AMD Radeon-VII - 138.5%
SSD: Samsung 970 Evo NVMe PCIe M.2 1TB - 284.8%
SSD: SanDisk Ultra II 960GB - 108.1%
HDD: WD Green 3TB (2011) - 47.4%
RAM: G.SKILL F4-3200C14-8GVR F4-3200C14-8GFX F4-3200C14-8GVR F4-3200C14-8GFX 32GB - 127.5%
MBD: Asus MAXIMUS VIII GENE
[IMG]

Share this post


Link to post
Share on other sites

Nice score, AMD seems to like this bench.  Core clock is up there! How's temps at high clocks on gpu/cpu? 

Edited by gamble

Share this post


Link to post
Share on other sites

With pump at 100% the GPU tops out around 70c Junction thanks to the Storm G5.

CPU runs ~60c at 5GHz 1.35v Prime 95 thanks to delid with liquid metal under the IHS and I was only feeding it 1.375v for 5.1GHz so there is probably room to push higher assuming it hasn't just hit a wall which from memory it may have.

Although due to AVX -1 it is running 100MHz less in Prime95.

 

Also had a crack at superposition which favors Nvidia using the settings there running over at OCAU

https://forums.overclockers.com.au/threads/unigine-superposition.1219642/page-19

superposition%205.1ghz%203600c14%20power

Edited by Dasa
  • Yes Sir! Very atomic! 1

Share this post


Link to post
Share on other sites

Not sure how good it is as everyone on OCAU seems to be using 2080Ti and getting 30k in Superposition but there is a lot of old 1080Ti tests where they are a few thousand points higher so it seems to favor Nvidia a fair bit unlike Firestrike.

The memory test on Userbenchmark is fairly good.

 

Mate upgraded to a 9700K when I was helping him build a custom loop so I swapped him for my 6700K for his 7700K. Same IPC but can run 5GHz 24\7 with less v than the 6700K needed to do 4.6GHz.

However the 6700K was able to hit higher RAM frequency but at least the 7700K can manage very tight RAM timings ok.

 

Still have yet to go back to OC it and take a decent picture but after a lot of teething problems his system is running well.

IMG_20190810_173554%201080.jpg

Edited by Dasa
  • Yes Sir! Very atomic! 1

Share this post


Link to post
Share on other sites

I'm think of going hard tubes next...   ^ It looks great.  I do like the ease of use on flex tubes so not sure yet.  I like the open case but I think it might be somewhat loud, no?  

Share this post


Link to post
Share on other sites

It is only as loud as the parts you put in it so as long as you have no HDD or coil whine... a D5 pump is nice and quiet at low RPM and with enough rads you can run the fans around 0-1000RPM depending on load.

 

Airflow over parts may not be so good so when pushing high overclocks for benchmarking you may want to aim a large fan at the case.

 

Hard tubing isn't great for someone that changes gear a lot as you need to redo the tubes every time you change a part since GPU cores and CPU locations move around and they usually still hold water when draining making it more difficult to get apart without making a mess.

Edited by Dasa
  • Yes Sir! Very atomic! 1

Share this post


Link to post
Share on other sites

Downloading Cryengine Neon Noir Ray Tracing Benchmark which doesn't require the Nvidia RT cores.

 

Ultra detail 1080p

NEON%20NOIR.jpg

 

Very high detail 1080p

NEON%20NOIR%20very%20high.jpg

 

The benchmark does seem to favor Nvidia with 1080 TI getting 10k at ultra and 2080 Ti getting 15k

Edited by Dasa

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×