3dmark Time Spy DX12 test

only single 770 result i could find, since you asked, was some random on YT with an 1800/1280MHz OC that got 2252. so pretty good I guess.

edit comments say 1800 boost and 1280 VRAM

 
only single 770 result i could find, since you asked, was some random on YT with an 1800/1280MHz OC that got 2252. so pretty good I guess.

edit comments say 1800 boost and 1280 VRAM



1800mhz on a gtx 680 seems a little extreme :p


I think that's the memory clock, nonetheless, good result!
 
only single 770 result i could find, since you asked, was some random on YT with an 1800/1280MHz OC that got 2252. so pretty good I guess.

edit comments say 1800 boost and 1280 VRAM



What a slide show.
 
Last edited:
1800mhz on a gtx 680 seems a little extreme :p
I think that's the memory clock, nonetheless, good result!
that was a gtx770, not a 680, like you asked for. and re the 1800/1280 I have no idea what those cards run at normally, never used one. this was the only result I found and I only looked because you asked.

What was a slide show.
huh?
 
that was a gtx770, not a 680, like you asked for. and re the 1800/1280 I have no idea what those cards run at normally, never used one. this was the only result I found and I only looked because you asked.

huh?
A gtx 770 is a gtx 680, which is why i asked about it

They run at less than 1ghz stock
 
oh ok its a rebadge. I was confused for a min. and yeah if they are <1GHz stock he prob got the numbers backwards.
 
oh I see. you mean the "video" slide show. yeah mine isn't very fast either. first test gets 15-24FPS second get 12-16FPS and the cpu test starts around 40 and drops to 12FPS. not exactly cutting edge hardware...
 
oh I see. you mean the "video" slide show. yeah mine isn't very fast either. first test gets 15-24FPS second get 12-16FPS and the cpu test starts around 40 and drops to 12FPS. not exactly cutting edge hardware...

Every new 3DMark does this though, and I feel the old ones were far more pleasing to look at. 3dm06 was great, the one after with the fireflies at night as one of the tests was pretty too. Then came vantage which was butt ugly :(

3dm06 with the steampunk zeppelin battalion fighting the giant gyrados was epic.

Now o think about it I remember a space one as well.
 
Every new 3DMark does this though, and I feel the old ones were far more pleasing to look at. 3dm06 was great, the one after with the fireflies at night as one of the tests was pretty too. Then came vantage which was butt ugly :(

3dm06 with the steampunk zeppelin battalion fighting the giant gyrados was epic.

Now o think about it I remember a space one as well.
yeah that is true! the have gotten less cool ever year. 06 was really cool and the fireflies in the forest was one of my favs!
if you haven't noticed the time spy is looking at display cases of all the old 3dmark scenes.
unfortunately I have never had the horsepower to run new 3dmarks at a decent speed as I'm usually running something mid-range since I game at 1080p on my 50" plasma.
 
yeah that is true! the have gotten less cool ever year. 06 was really cool and the fireflies in the forest was one of my favs!
if you haven't noticed the time spy is looking at display cases of all the old 3dmark scenes.
unfortunately I have never had the horsepower to run new 3dmarks at a decent speed as I'm usually running something mid-range since I game at 1080p on my 50" plasma.
The last 3dm I ran at full speed was whatever was available when the 570 launched I guess.

Yeah I noticed the old 3dm scenes, but the time spy bench overall sucks. The second test with the cool astrarium thing was better, but still they could have made it less boring
 
Yeah I could have gone for more fantasy and explosions and less story telling. The lights over the toadstools was probably my favorite of all time.
 
yeah 3dmark needs to go back to actiony, explosion, smoke filled fantasy and space fights. screw the slow paced sub diving, museums and killer instinct knockoffs! moor splosions!!!

edit: or some really cool mech scenes with bullets and missiles flying everywhere, smoke trails, cool lighting, particle effects in the snow/sand and heat distortion!
 
Yeah I could have gone for more fantasy and explosions and less story telling. The lights over the toadstools was probably my favorite of all time.

Unigine Heaven got it right Imo, it's beautiful, but dated
 
yeah 3dmark needs to go back to actiony, explosion, smoke filled fantasy and space fights. screw the slow paced sub diving, museums and killer instinct knockoffs! moor splosions!!!
I'd rather less action and more scenic. Dammit. I want 3d mark Nagrand.

Nagrand in the burning crusade was one of the most beautiful game settings ever
 
I really wish Nvidia would make another demo with a hot chick. They had a good thing going for a while.
 
I really wish Nvidia would make another demo with a hot chick. They had a good thing going for a while.

What about a dx12 demo with Ruby? Man I used to love downloading the new demos when I purchased new cards whether its nvidia or amd. It used to be so much fun. I miss those days. Put in the card download the demo and drools. lol
 
1. You have cited AMD-colloborated port of DOOM utilizing AMD's exclusive extensions before. That makes a person calling Ghostbusters reboot a good movie look not biased, that's how hypocritical you are right now.

2. You didn't just link to Mahigan's fantasizing about what Pascal and Polaris have, did you? I mean, after Mahigan's predictions on Pascal and Polaris his hypotheticals should honestly only be looked at drunk. And yes, he may have an EE degree with major in GPU design for all i care, his predictions on Polaris have discredited him as objective entity.

3. And finally, i've just read the post and boy, he is wrong
 
Last edited:
Quoting 3DM developer

Yes it is. But Engine cannot dictate what the hardware has available or not.

Async compute is about utilizing "idle" shader units. Slower the card, less idle ones you have. Less capable hardware may also be hard pressed to utilize all of them even if the engine asks nicely. Also there may be limitations as to what workloads in the engine *can* run in parallel. Yes, Time Spy is very graphics-heavy, since, well, its a graphics benchmark. But even there many of the rendering passes have compute tasks that can use this.

Ultimately some AMD cards gain quite a bit (ie. they have a lot of shader units idling while rendering and they are very good at using them for the available paraller loads). Some AMD cards gain less or not at all (either less capable at paralleriziing, less idle shader units or no idle shader units at all - for example a HD 7970 is hard pressed to have any to "spare")

Some NVIDIA cards cannot do this at all. The driver simply says "hold your horses, we'll do this nicely in order". Some NVIDIA cards can do some of it. They might use another way than AMD (more driver/software based), but the end result is the same - the card hardware is capable of doing more through some intelligent juggling of the work.

Async disabled on maxwell btw
 
I'm getting a very low CPU score. IDK why.

3889 pts on the CPU in my sig. And it actually runs at 4625 not 4.5 as it says there. I saw 4790K-s getting 1000 points more, that are not even clocked that high. How can this be?

This seems to be an ongoing issue for me, as I remember getting a very low CPU score even in the Cloud Gate test.
 
Yeah, I thought I did too and I just looked up the XL-R8R demo video, and it wasn't nearly as cool as I recall.

What do you mean? It was cool when it was new. They started getting less cool with 3Dmark 05, and become completely uncool with vantage.
 
With specs in sig... upload_2016-7-18_6-16-8.png
 
Hmmm looks like people at overclock.net have found FM making a "lite" version of async compute that benefits only pascal video cards. Plenty of evidence showing this.

Is this the PhysX disaster all over again?

Kyle was right about 3dmark! Can't trust a benchmark that isn't a game.
 
Hmmm looks like people at overclock.net have found FM making a "lite" version of async compute that benefits only pascal video cards. Plenty of evidence showing this.

Is this the PhysX disaster all over again?

Kyle was right about 3dmark! Can't trust a benchmark that isn't a game.

Ok some of the talk over there is just stupid. If Time Spy is coded to only benefit Nvidia, then why do AMD cards get a bigger boost from Async Compute?!? People get so wrapped up in the circle jerking each other they forget all about logic.

If 3D Mark was trying to shill for Nvidia Async Support they did a shitty job.
 
Last edited:
Hmmm looks like people at overclock.net have found FM making a "lite" version of async compute that benefits only pascal video cards. Plenty of evidence showing this.

Is this the PhysX disaster all over again?

Kyle was right about 3dmark! Can't trust a benchmark that isn't a game.

Ok some of the talk over there is just stupid. If the Time Spy is coded to only benefit Nvidia, then why do AMD cards get a bigger boost from Async Compute?!? People get so wrapped up in the circle jerking each other they forget all about logic.

If 3D Mark was trying to shill for Nvidia Async Support they did a shitty job.

Not to mention a permanently damaged reputation and pot eulogy sales if such a vast conspiracy where true. Companies rarely try their hand at such an incredible level off deception.
 
Hmmm looks like people at overclock.net have found FM making a "lite" version of async compute that benefits only pascal video cards. Plenty of evidence showing this.

What evidence? That thread is just a few people making baseless accusations and a bunch of idiots lapping it up.
 
Hmmm looks like people at overclock.net have found FM making a "lite" version of async compute that benefits only pascal video cards. Plenty of evidence showing this.

Is this the PhysX disaster all over again?

Kyle was right about 3dmark! Can't trust a benchmark that isn't a game.
o-PITCHFORK-facebook.jpg
Ok some of the talk over there is just stupid. If the Time Spy is coded to only benefit Nvidia, then why do AMD cards get a bigger boost from Async Compute?!? People get so wrapped up in the circle jerking each other they forget all about logic.

If 3D Mark was trying to shill for Nvidia Async Support they did a shitty job.

i love how Mahigan has become "some people on overclock.net" and his totally conjectural contentions became " have found FM making a version of async compute that benefits only pascal video cards. Plenty of evidence showing this."

What evidence? That thread is just a few people making baseless accusations and a bunch of idiots lapping it up.

Someone should track down all of Mahigan's theories on async compute and Maxwell and put them in a blog post. This is the fifth or sixth, I'm not sure. Worth noting though, if he's immortal then he's guaranteed to get it right at some point
 
[Various] Futuremark's Time Spy DirectX 12 "Benchmark" Compromised. Less Compute/Parallelism than Doom/Aots. Also...


What evidence? That thread is just a few people making baseless accusations and a bunch of idiots lapping it up.

Originally Posted by Doothe

Time Spy is basically only just doing that new feature that Pascal has - it preempts some 3D work, quickly switches context to the compute work, then switches back to the 3D.

So it seems to me that Time Spy has a very minimal amount of async compute work compared to Doom and AotS, and the manner in which it does its async is friendly to Pascal hardware. I don't think it's necessarily "optimized" for nvidia, as GCN seems to have no issue with context switching either. It's just not being allowed to take full advantage of GCN hardware.

s51q4IX.jpg
 
  • Like
Reactions: N4CR
like this
[Various] Futuremark's Time Spy DirectX 12 "Benchmark" Compromised. Less Compute/Parallelism than Doom/Aots. Also...




Originally Posted by Doothe

Time Spy is basically only just doing that new feature that Pascal has - it preempts some 3D work, quickly switches context to the compute work, then switches back to the 3D.

So it seems to me that Time Spy has a very minimal amount of async compute work compared to Doom and AotS, and the manner in which it does its async is friendly to Pascal hardware. I don't think it's necessarily "optimized" for nvidia, as GCN seems to have no issue with context switching either. It's just not being allowed to take full advantage of GCN hardware.

Use spoilers for large images please.

That's not what Pascal does at all, preemption has little to do with async compute. Context switching an SM is just as slow as on Maxwell. There is no context switching.
The picture you posted is of a 390X anyway, I don't know what you thought it would prove.

Async Shaders relies on context switching, it's not "GCN doesn't have a problem with context switching", it's "GCN makes use of the independent queues by context switching the CU and running compute shaders while waiting on dependencies".

GCN does context switching. Pascal and Maxwell do not.
 
Don't mean to get in the way of all the charts/discussions...I'll just leave these graphics scores here:

MSI 390 @ 1100/1650 --> 4131
HIS 7870 @ 1175/1200 --> 1639 (now *this* was a slideshow, averaged 11 fps and 9 fps respectively on the two tests)
 
Back
Top