3DMark Time Spy Benchmark with Ray Tracing to Launch in September

[Carclis]

This news went largely unheard whilst all attention was given to the recent launch of Nvidia's 20 series products but 3DMark developer has announced that it has a version of its popular TimeSpy benchmark with ray tracing elements added in. The launch of this iteration of Time Spy is aimed at late September which will coincide with the release date of the 20 series Nvidia products.
 

Quote

UL Benchmarks could not yet make public the next iteration of the suite, but they did confirm they are currently developing a ray tracing benchmark (without disclosing exact dates). The release would, then, "align with the launch of Redstone 5", said a UL representative. Redstone 5, or “Windows 10 1809” as it's now known is the Microsoft Fall 2018 OS update and is scheduled to launch in late September. The release is said to include a new version of the 3DMark TimeSpy benchmark.

 

Personally I'm a little suspect of the intentions of UL Benchmarks (developer) since the timing of this launch is too coincidental for a third party developer whose purpose is to accurately represent the performance of all graphics cards in relation to each other.

 

Source

[williamcll]

It would be interesting to see how other cards try to do ray tracing

[ScratchCat]

It is not their job accurately represent the performance of each card, only to provide a measure of performance which can be used to estimate the relative perfomance of cards. There is no such thing as a benchmark for all use cases.

 

This benchmark still will represent the real life perfomance in ray tracing tasks if the ray tracing is not a NVIDIA GameWorks type package with bias for NVIDIA cards and even then if studios use NVIDIA's package this would represent the perfomance many ray tracing games.

[mr moose]

2 hours ago, Carclis said:

whilst all attention was given to the recent launch of Nvidia's 20 series products

 

Nvidia released a new card?   I'm still waiting for the 1180.

[Carclis]

17 minutes ago, ScratchCat said:

It is not their job accurately represent the performance of each card, only to provide a measure of performance which can be used to estimate the relative perfomance of cards. There is no such thing as a benchmark for all use cases.

 

This benchmark still will represent the real life perfomance in ray tracing tasks if the ray tracing is not a NVIDIA GameWorks type package with bias for NVIDIA cards and even then if studios use NVIDIA's package this would represent the perfomance many ray tracing games.

My concern is that standard Time Spy may be overlooked and replaced with the new ray traced version. Whilst there isn't anything wrong with having a benchmark like this it screams some of the age old phrases of "take benchmarks with a grain of salt" or "wait for non-synthetic benchmarks". The timing just makes it feel like Nvidia marketing to me.

[GoldenLag]

2 hours ago, williamcll said:

It would be interesting to see how other cards try to do ray tracing

intel 4000 series graphics vs raytracing

[mr moose]

46 minutes ago, Carclis said:

My concern is that standard Time Spy may be overlooked and replaced with the new ray traced version. Whilst there isn't anything wrong with having a benchmark like this it screams some of the age old phrases of "take benchmarks with a grain of salt" or "wait for non-synthetic benchmarks". The timing just makes it feel like Nvidia marketing to me.

So long as people don't try to compare different test configurations and endeavor to keep apples to apples it shouldn't be a problem. 

[yian88]

*with fake ray tracing, vendor locked* Fixed news article

[The Benjamins]

3 hours ago, mr moose said:

Nvidia released a new card?   I'm still waiting for the 1180.

They are skipping 11xx and going to 20xx.

[Tech Enthusiast]

4 hours ago, Carclis said:

My concern is that standard Time Spy may be overlooked and replaced with the new ray traced version. Whilst there isn't anything wrong with having a benchmark like this it screams some of the age old phrases of "take benchmarks with a grain of salt" or "wait for non-synthetic benchmarks". The timing just makes it feel like Nvidia marketing to me.

Or, you know... they want to update their benchmark to reflect the new technology?

People may want to see the new stuff benchmarked as well.

 

No idea where you get that tinfoil idea from.

This is pretty normal and expected. Also, expect the same treatment from 3dmark and others.

[Carclis]

13 minutes ago, Rattenmann said:

Or, you know... they want to update their benchmark to reflect the new technology?

People may want to see the new stuff benchmarked as well.

 

No idea where you get that tinfoil idea from.

This is pretty normal and expected. Also, expect the same treatment from 3dmark and others.

There is no single game with ray tracing available right now that I know of. So it is surprising to me that a benchmark like this would have relevance. In order for a benchmark to have meaning it would need to reflect real world applications and we don't know to what extent software will utilise ray tracing. So there is a potential for ray tracing to be overused. All I'm saying is that coming out of the gate so soon seems premature and naive.

[pas008]

20 minutes ago, Carclis said:

There is no single game with ray tracing available right now that I know of. So it is surprising to me that a benchmark like this would have relevance. In order for a benchmark to have meaning it would need to reflect real world applications and we don't know to what extent software will utilise ray tracing. So there is a potential for ray tracing to be overused. All I'm saying is that coming out of the gate so soon seems premature and naive.

dx12 same thing

async same thing on aots

 

benchmarks exist for certain workloads hence this update

 

no big deal here

[Carclis]

9 minutes ago, pas008 said:

dx12 same thing

async same thing on aots

 

benchmarks exist for certain workloads hence this update

 

no big deal here

Ashes was an actual game with a benchmark, similar to the built-in one that the Tomb Raider games feature so it's a little different. The last three 3DMark iterations have launched at times coinciding with Nvidia cards; Time Spy with the GTX 10 series, Fire Strike with the GTX 700 series and 3DMark 11 with the GTX 500 series. I'm just calling it as I see it because it's a trend that seems to be continuing. Speaking of asynchronous compute, didn't it only get supported in 3DMark after Pascal launched?

[ScratchCat]

57 minutes ago, Carclis said:

There is no single game with ray tracing available right now that I know of. So it is surprising to me that a benchmark like this would have relevance. In order for a benchmark to have meaning it would need to reflect real world applications and we don't know to what extent software will utilise ray tracing. So there is a potential for ray tracing to be overused. All I'm saying is that coming out of the gate so soon seems premature and naive.

Ray tracing games have been announced, we have seen playable demos of the technology. I agree that this benchmark will not become the go-to benchmark for unbiased testing but only for ray tracing technology, just like the VR benchmark.

 

The timing does suggest NVIDIA support in the creation to make the RTX series look good in this benchmark.

[pas008]

58 minutes ago, Carclis said:

Ashes was an actual game with a benchmark, similar to the built-in one that the Tomb Raider games feature so it's a little different. The last three 3DMark iterations have launched at times coinciding with Nvidia cards; Time Spy with the GTX 10 series, Fire Strike with the GTX 700 series and 3DMark 11 with the GTX 500 series. I'm just calling it as I see it because it's a trend that seems to be continuing. Speaking of asynchronous compute, didn't it only get supported in 3DMark after Pascal launched?

new features from cards to benchmark

tessellation, compute shaders, multithreading

firestrike for multigpu and then 4k

and timespy for more dx12 features

 

nvidia is pushing forward features so why not have benchmarks for those features

[Jito463]

2 hours ago, Rattenmann said:

My concern is that standard Time Spy may be overlooked and replaced with the new ray traced version.

2 hours ago, Rattenmann said:

Also, expect the same treatment from 3dmark

Time Spy is a module within 3DMark, not a separate test.

[Tech Enthusiast]

3 hours ago, Carclis said:

There is no single game with ray tracing available right now that I know of. So it is surprising to me that a benchmark like this would have relevance. In order for a benchmark to have meaning it would need to reflect real world applications and we don't know to what extent software will utilise ray tracing. So there is a potential for ray tracing to be overused. All I'm saying is that coming out of the gate so soon seems premature and naive.

There is new technology. 

This new technology will be tested.

 

The companies updating their benchmarks for new tech will be used, others won't. 

It is as simple as that.

[Tech Enthusiast]

1 hour ago, Jito463 said:

Time Spy is a module within 3DMark, not a separate test.

Well, there you go then.

 

Unlikely to see any major benchmarks not updating for new tech.

[mr moose]

8 hours ago, The Benjamins said:

They are skipping 11xx and going to 20xx.

No they're not, WCCFtech explicitly said there were leaks and rumors absolutely confirming the possibility that the next Nvidia card was an 1180. I'll not hear about this supposed 2080.

 

 

 

 

Sorry should have put the /s on the last post.

[Sauron]

14 hours ago, Carclis said:

Personally I'm a little suspect of the intentions of UL Benchmarks (developer) since the timing of this launch is too coincidental for a third party developer whose purpose is to accurately represent the performance of all graphics cards in relation to each other.

Why? With this launch from nvidia we have cards that can run raytracing with acceptable performance - it's only natural to want to measure it. Even if AMD doesn't have a direct competitor it's at least useful to know how the 20xx cards compare against each other...

5 hours ago, Carclis said:

Ashes was an actual game with a benchmark, similar to the built-in one that the Tomb Raider games feature so it's a little different. The last three 3DMark iterations have launched at times coinciding with Nvidia cards; Time Spy with the GTX 10 series, Fire Strike with the GTX 700 series and 3DMark 11 with the GTX 500 series. I'm just calling it as I see it because it's a trend that seems to be continuing. Speaking of asynchronous compute, didn't it only get supported in 3DMark after Pascal launched?

That's probably because for the past few years nvidia has been leading the performance race, hence the need for new benchmarks to take advantage of the changes. The last thing you want is to be bottlenecked by the cpu when running a gpu benchmark.

[porina]

What was Futuremark has been forward looking. They want to provide something to test with. As with any synthetic benchmark, you can question its relevance, but given Fire Strike and Time Spy have been long used, and have variations like extreme and ultra to push hardware harder. A raytracing variation is a logical extension to that.

 

They will make a best guess as to how software might use those new features which may or may not represent future games. In a similar way, Time Spy right now wont tell you how all games will behave.

 

What might be more interesting is just how do pre-Turing cards do with it, or will they not run at all? I don't know if there will be a software emulation layer to provide the functionality.

[Carclis]

9 hours ago, Sauron said:

Why? With this launch from nvidia we have cards that can run raytracing with acceptable performance - it's only natural to want to measure it. Even if AMD doesn't have a direct competitor it's at least useful to know how the 20xx cards compare against each other...

That's probably because for the past few years nvidia has been leading the performance race, hence the need for new benchmarks to take advantage of the changes. The last thing you want is to be bottlenecked by the cpu when running a gpu benchmark.

Allow me to disagree with the bolded part, at least as far as the performance shown would demonstrate and the developers comments would lead me to believe. For me benchmarking ray tracing feels like the comparisons made between the Fury X and GTX 980 Ti at 4k in order to show that the Fury X was faster. Sure, it had a lead at 4k but neither card was able to yield playable performance at that resolution, making the comparison meaningless. At present benchmarking ray tracing is in exactly the same boat since playable performance is not currently feasible, and if it is it will require a 1080p configuration that nobody buying a $1200 GPU will be running, or will wish to resort to. This is the main reason I'm questioning the relevance of the benchmark, but I understand the desire of those wishing to show off their shiny new GPU as well.

[Sauron]

4 hours ago, Carclis said:

Allow me to disagree with the bolded part, at least as far as the performance shown would demonstrate and the developers comments would lead me to believe. For me benchmarking ray tracing feels like the comparisons made between the Fury X and GTX 980 Ti at 4k in order to show that the Fury X was faster. Sure, it had a lead at 4k but neither card was able to yield playable performance at that resolution, making the comparison meaningless. At present benchmarking ray tracing is in exactly the same boat since playable performance is not currently feasible, and if it is it will require a 1080p configuration that nobody buying a $1200 GPU will be running, or will wish to resort to. This is the main reason I'm questioning the relevance of the benchmark, but I understand the desire of those wishing to show off their shiny new GPU as well.

The difference ought to be much more marked than that between the fury x and the 980ti at 4k. Even if it can only do 1080p 30, it should be at least an order of magnitude faster than non-rt-native cards. Yes, it's not exactly practical for games, but then again that's not the only thing gpus are used for. For an animation studio, being able to render raytracing that quickly would be great.

[Carclis]

2 hours ago, Sauron said:

The difference ought to be much more marked than that between the fury x and the 980ti at 4k. Even if it can only do 1080p 30, it should be at least an order of magnitude faster than non-rt-native cards. Yes, it's not exactly practical for games, but then again that's not the only thing gpus are used for. For an animation studio, being able to render raytracing that quickly would be great.

I was just pointing out that the scenario was somewhat ridiculous which you seem to agree with. Ray tracing seems to be a professional feature that is included in the cards because the architecture is not a gaming focused one. In my opinion it's just marketed as a gaming feature.

[Mira Yurizaki]

I don't get what ray tracing has to do with NVIDIA only, when ray tracing is now a feature of D3D12.

 

It'd be like claiming back in 2003-2004 that all DX9 benchmarks were ATI biased even though NVIDIA didn't deliver competent DX9 hardware to begin with until GeForce 6. Especially considering the Radeon 9800 was released before DX9 officially launched.

Edited August 28, 2018 by M.Yurizaki