Jump to content
3dmark master

Atomic 3DMark & Game Benchmark Rankings

Recommended Posts

So, I sold the RX 56 only because of the crazy prices they are going for used. I bought a Asus Strix GTX 1080Ti OC from that which was nearly a wash, I think I paid $10 bucks after fees on ebay. Few results below, I dont have my cpu clocked up but at my daily 4ghz. There is one bad, well kinda strange thing about this card... It clocks without changing anything up to 1924-1949mhz at stock. If I up the power limit to 120% it stays closer to the 1949mhz and if I up the voltage all the way up to 1.093v it usually is around 1962 to 1949mhz. Great right, should OC like a boss?? Nope! I can only get about +20 from there and will not hit 2000mhz. Now, I really dont mind since it is a high OC right from the box without having to do anything but its strange that I cant really push it anymore and really it is not even worth running it at +20... The cooler is unreal though. It runs up to 40% at the highest I have seen, cant hear it at all and has not hit over 69c yet even when gaming or benching. Still playing with the ram but it hit 12000 no problem and wanted to run some benches. I dont think Ill be able to catch Mark84 :( even if I push the cpu and ram more but still happy with the results over the RX 56.. Still looking at Dasa scores its hard to believe he pushed his to those great scores too considering its a GTX 1070. Ill probably try to up the cpu some just because Id like to hit 20k/10k in FS/TimeSpy.

 

FireStrike: 19606 - https://www.3dmark.com/fs/14708608

TimeSpy: 9837 - https://www.3dmark.com/spy/3146034

 

For those curious on Superposition, I ran this with the RX 56 and RX 64 bios for the below:

 

Performance - 1080p Extreme preset - DirectX

RX56: 3299

RX56(RX64 bios): 3809

1080ti Ram OC: 6052

1080ti: 5840

Edited by gamble

Share this post


Link to post
Share on other sites

wonder why ryzen does so badly in the combined score my 6700k is 60% faster yet your 30% faster in both gpu and physic tests

sure its probably single threaded but still the difference shouldn't be that much

Share this post


Link to post
Share on other sites

Maybe because yours is running 800mhz higher... Still piss poor combined score for sure. There's a bios update for my board but they took out OC option for some reason and supposed another bios coming to fix vulnerability.

 

Edit: Found this similar setup, combined score looks much better than mine but looks like he/she ran near the same clocks but higher ram frequency... Maybe I'll try to flash the bios on mb again, possibly it didn't take well. https://www.3dmark.com/fs/13060141

Edited by gamble

Share this post


Link to post
Share on other sites

what speed is your ram running? sora3 managed .11fps more than you in the combined test with 2666 ram

Share this post


Link to post
Share on other sites

So, I sold the RX 56 only because of the crazy prices they are going for used. I bought a Asus Strix GTX 1080Ti OC

You need to update your bio now ;)

Share this post


Link to post
Share on other sites

Currently, at 2933... Looking at previous scores I remember saying "I dont know how I got this score" with the RX 56 as it was much higher than most of my other scores even rerunning over and over trying to reproduce. But, my combined score was high and other runs it was low... strange stuff, Ill dig in more soon.

--

Yea will update sig now thanks for the reminder!

Share this post


Link to post
Share on other sites

Thx Dasa, it has to be something with that. I installed Ryzen power plan last year and when I looked it wasn't even there anymore but there has been some major updates lately. When I re-download it, it wouldn't downclock the cores on either Ryzen balance or performance modes. Only win 10 balance would drop the clocks down when not being used. Looking at resource management I do not have any parked cores. I ran out of time but it just seems like poor optimization on Futuremark side but to be fair it's been out a long time now. Looking at all the charts if I put parked at 50% it should yield better results... I need to read more since I'm still confused on how it work and why.

Share this post


Link to post
Share on other sites

like you say there has been a lot of updates that were supposed to fix a lot of these issues so that link i provided may not be relevant anymore but his scores do seem to be in the ballpark of where you are and where you should be

  • Like 1

Share this post


Link to post
Share on other sites

To be honest, I have no idea why now I can get nearly 21k with the exact same setting.... I changed nothing in the bios. The card is running at the clocks as before. My combined score jumped to 6767... I do not see an option for parked cores and how to set this as the screenshots showed in those posts. I can set min/max on the processors though. Using Ryzen balance would fail when I get to the physic test, the same is true using win10 High performance. Then I went back to win10 balanced and bam, 21k score. Possibly, it was just luck of the core threads and where they landed. Still think for a benching program they would try to make this better for all cpus.

 

20929 - https://www.3dmark.com/3dm/24679858?

 

compare - https://www.3dmark.com/compare/fs/14708608/fs/14723706#

 

Ill give it a few more runs to see if this is repeatable.

 

Edit: No, it is not repeatable meaning 50% of the time it will run great and other times I will get my 1st score with the same combined.... so annoying but glad you mentioned.

Edited by gamble

Share this post


Link to post
Share on other sites

Possibly, it was just luck of the core threads and where they landed.

guess that could be it with the two groups of cores and the high latency interface between them

Share this post


Link to post
Share on other sites

I opened a ticket with support, just to ask, below is their official rep reply. Basically, what we thought on the core threads and how win10 assigned them can impact performance:

 

Subject: Ryzen and Combined Score Inconsistent

 

Me:

"FireStrike with Ryzen is inconsistent on combined score. Same setting but will yield different results each time I run this. Several others in forums have noted the same issue. It seems like FS is not optimized for Ryzen. Please let me know when the patch will come out to address this. Ryzen2 coming soon so everyone will be running these benchmarks."

 

Rep:

"More like Ryzen is not optimized for all applications. This is how Ryzen works on any number of games etc. We have no plans to change the test in any way as changing it would invalidate all existing benchmark scores. It is quite accurate test for older DX11 games that do not have specific code paths for Ryzen - basically the result depends on which core Windows happens to put the benchmark vs. where video driver process is. Core-to-Core communication can be fast or slow, depending on that.

Beyond that, Time Spy and Time Spy Extreme are better tests for modern hardware."

 

===

 

Anyone using Superposition benchmark? Ive been using Performance - 1080p Extreme preset - DirectX.

Edited by gamble

Share this post


Link to post
Share on other sites

2.9 Gig download - but they do have a torrent which is handy. Might queue it up overnight.

Share this post


Link to post
Share on other sites

Basemark GPU

Vulkan 5712

OpenGL bc7 5781

OpenGL astc 5681

OpenGL etc2 5558

 

Basemark Web

desktop using Windows 10.0

 

Firefox 61.0 score: 829.87
CSS Capabilities
56.79%
HTML5 Capabilities
92.07%
Page Load and Responsiveness Capabilities
91.13%
Resize Capabilities
75.97%

 

system fairly much same as tag except cpu at 4.5ghz ram at 3733c16

Edited by Dasa

Share this post


Link to post
Share on other sites

running at stock on cpu, using Official Test:

 

Vulkan: 9132

OpenGL bc7: 4388

OpenGL astc: 4404

OpenGL etc2: 4369

 

Basemark Web:

 

desktop using Windows 10.0 with Chrome 67.0.3396.87 score:848.04

CSS Capabilities57.75%
HTML5 Capabilities96.58%
Page Load and Responsiveness Capabilities89.39%
Resize Capabilities75.97%

 

 

kinda strange results looking at Dasa's. Ill set my cpu to 4ghz and try again.

 

 

Edit:

@4ghz

 

Vulkan: 8921

OpenGL bc7: 5513

OpenGL astc: 5637

OpenGL etc2: 5522

Edited by gamble

Share this post


Link to post
Share on other sites

I feel ripped off considering the 2.8 Gig download. The actual test is just some slideshow, and the experience mode is short and lame.

 

For some reason the compare on powerboard function just opens up a webpage that's blank. Maybe it's an OS thing.

 

i7-4790 24 Gig, Win7 Ultimate, GTX 960

Overall score 1821

Share this post


Link to post
Share on other sites

yeah i was a bit disappointed to

 

i gave this photshop benchmark a few runs

https://www.pugetsystems.com/labs/articles/Puget-Systems-Adobe-Photoshop-CC-Benchmark-1132/

the gains from ram are not as big as some other software but 3200 ram still looks to be decent value for the performance boost it provides

 

2133c15 to 3866c16 percentage increase
Overall Score 9.5%
General Score 14.5%
Filter Score 12.2%
Photomerge Score 4.3%
GPU Score 19%

eb106f9873cf97cffa0485df9a52c910285bd53c

Share this post


Link to post
Share on other sites

Yea that bench sucked.... If you go to I think it was custom it will do full screen.

Share this post


Link to post
Share on other sites

Sold my GTX 1080 ti Strix for a good amount that covered most of buying a Gigabyte RTX 2080 Windforce.  Just did it to play with something new, wasn't sure if I would see much of anything since my Ti OC'ed well and I was moving to a lower end 2080.   Plus, I figured that I would have a fresh warranty and the latest gen for nearly break even which should sell easier in the future.  So far I've been happy with this card, I have as much fun benching and tweaking as I do gaming.  I have been messing with it for 3 days non stop tweaking and testing.  Not quite done as Im still OC the memory but wanted to research that since I have such a high OC on ram that I want to confirm others are getting before moving up anymore 🙂

This is the Non A chip with lower power limit.  I tried to flash the card, since it is based of reference design, with other bios for A chips but would not take as it would give device mismatch errors with nvflash even if I did -6 or - -protectoff.  The reason I wanted to flash is because the Power limit is set really low and can slide it up to only 111%.  If I was able to raise this, the chip would OC much more I would guess since certain benchmarks it would throttle the core due to the power limits.  

Using the OC Scanner built in tool for MSI AB, it tested to +143 on the core.  I was able to move it up manually from there a bit more to +170. 

-- --

Core Does not Throttle:  (Does not hit Power Limit, some get close but not enough to throttle)

Core Clocks:  +170 =  2070~2040 (Usually around 2055)

Firestrike

Valley

Heaven

Battlefield V

Civ6

-- --

Core Throttles: (Hits Power Limit)

Core Clocks:  +170 =  1995~2025 (Usually around 2010)

TimeSpy

Superposition

Ashes (AoS) Crazy preset 

-- --

Benchmarks Compare against my best runs on the 1080ti with this current CPU 2600x, not the 2700x results.  Some of the below might have had a slightly less OC on the cards Ram as I am tweaking still.  Also, these result might have my older Ram that I was able to run faster with this MB (3200 vs 2666), not quite a straight compare but close:

Core:  +160 / 1995~2070mhz

Ram:  +1250mhz?! (Crazy high and no glitches yet, can go higher it seems but want to research what other got.  Slider tops out at +1500)

Fan:  Curve for 70c is 80%, it stays under 70c with this curve

Power Limit:  111%

Vcore:  +100

Reference 1080 ti setting:  1960c / +600m

TimeSpy: https://www.3dmark.com/compare/spy/5963049/spy/4001157

1080ti:  10321

2080:  11420

FireStrike: https://www.3dmark.com/compare/fs/18054656/fs/15817358

1080ti: 29124

2080:  27563

Superposition:  (1080p Extreme)

1080ti:  6052 (Note, I think this was with the 1700x though)

2080:  7187

AoS:  1440p (Crazy Preset) GPU Focused

1080ti:  74 FPS, 13.5ms

2080:  75.5 FPS, 13.2ms

 

Noise:  2080 is louder slightly since I am trying to keep it under 70c.  At 65c, it goes down one tic of 15mhz.   If I run the fan at the same % as the Asus at 60%, I can't hear it over the other fans.

Ray Tracing:  Tried it out on BFV... what a joke, sure it looks nice but I want frames in FPS games.  Dx11 seems to work the best anyway for that game.  With Ultra setting core is flat at 2055mhz.  

 

Im hoping to find a bios reference design with a higher Power limit so I can flash it.  Im curious on Dlss too.  If you have a 1080ti, it probably would not be worth upgrading... I just did it to play around.  I was thinking of waiting for the Vega 7 but this card was cheap.  Maybe Ill sell this and buy the Vega 7 later , see how many side upgrades I can do 🙂

 

 

Share this post


Link to post
Share on other sites

Good luck getting a Rad7 in Australia, I suspect the usual price rorting will kick in given we get bugger supply of anything scarce here.

Share this post


Link to post
Share on other sites

Nah I was just joking about Vega 7 🙂  

So, I have been playing around some more, I have noticed that when I change the power management under control panel from Adaptive to Optimal, I can run higher clock yet still on the core.  Kinda sucks since I should have checked that first but this really only improved benchmarks a bit.  The memory has topped out at around 1300 as 1350 would bench only halfway thru.  I can game with Civ6 or BFV at 2100mhz without or very little throttle, it is nearly flat lined.   Im going to push it higher today or tomorrow to see what it's very top limit is. Of course, benchmarks, especially newer version will hit the power limit and still drop down at time to 1995 for certain frames in Timespy.  Even with these high OC, it is still only on par with 1080 ti on most as they trade blows.  And after all this Ill settle for probably 2055mhz/16000mhz on everyday use.   I really want to upgrade my CPU to Ryzen 2 when it is released, I had $ set aside for this in case they launched after CES...  Bring on the 12core 5+ ghz cpus!  Ill take some screenshots as to what I see on the throttling for the core vs power limit if someone would like to see them on certain game/benches. 

 

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

×