PC Benchmark thread (Part Deux) - Firestrike

[Brainling]

Firestrike results for my 1080 Ti

1080 Ti
Gigabyte Founder's Edition, no OC

• Fire Strike (1080p)
  18,889
  http://www.3dmark.com/fs/12007373
• Fire Strike Extreme (1440p)
  12,085
  http://www.3dmark.com/fs/12007431
• Fire Strike Ultra (2160p)
  6,662
  http://www.3dmark.com/fs/12007559
• Time Spy (1440p, DX12)
  8,134
  http://www.3dmark.com/spy/1381426

 

[little P]

Nice numbers! Thanks for doing that, it's a huge jump from a 1080 comparing your numbers to Kit's earlier post with his 1080. Wow... That 1080Ti is some piece of kit!

 

[Brainling]

It's 65+% faster than my 980 Ti across the board. 80% faster at 1080p (where this card would be total overkill).

 

[little P]

Nice one. Thanks for the info that is very interesting to see real world results. I'll be waiting on Asus to release price and release date details of their Strix 1080Ti! I'm defecting from Gigabyte for the first time in 5 years.

EDIT: First review is out and it looks like I've made the right choice

https://www.techpowerup.com/reviews/ASUS/GTX_1080_Ti_Strix_OC/

 

[Kitlope]

Just picked up a Asus Strix OC edition 1080 Ti and ran Firestrike. The numbers are lower than I figured, I'm thinking the older CPU is starting to play a roll but not sure. Any idea's guys? edit: Looking at my previous post from summer with and without CPU OC, definitely plays a roll of about 500 more Firestrike points.

Back down to 3.33 Ghz from the mild OC I usually have, 18 GB of DDR3. I'll do another one with the CPU at around 3.75 Ghz and tweak the ram to 1600 (edit: I think it went 1805).

 

[Kitlope]

Well holy fuck. Just did the CPU/ram OC and what a difference. Is this true? I can only imagine what a modern day 6 core will improve it.

 

[little P]

Nice Kit. I think also with the Firestrike test running @1080p (and any games for that matter) the results will be CPU bound from what I understand. The 1080Ti really comes in to it's own @1440p and 4K. Anything before that and it's going to be twiddling it's thumbs waiting for the CPU. Congratulations on your purchase though you mad bastard lol

 

[Kitlope]

I read that too Lil' P, 1440 is where it shines. Hence why I bought it.

No more video cards for me.

 

[Brainling]

I was wondering why your Firestrike score was nearly 1200 below mine with an OC'ed card, until you posted about needing to OC your other components. I guess my 4790k isn't so bad after all given that it did an 18,889 with the CPU being at stock speeds.

 

[Kitlope]

Yeah Brainling, the CPU I'm using is older than the hills now, although still practical the upgrade bug is biting me. I'm seriously considering a Ryzen 1600x.

Your Haswell 4790k is a great chip with more years of life in it.

 

[little P]

Well I've been playing around with overclocking my CPU to try and give it a little longer before I upgrade.

I managed to get the plucky little i7 2700K 3.5Ghz to a stable 4.5Ghz (5 hours running prime with no BSOD) and here are the results of running firestrike compared to my last run a year ago.

A substantial improvement to the physics score and a decent bump to the overall score. I'm already noticing the difference in games that are cpu heavy, so all in all pleased with the result.

Also pleasantly surprised nothing blew up... yet.

This 2700K is such a great chip! Thanks Duke

 

[Brainling]

Some score updates since I did the full system overhaul to Z370 with an i7-8700K and 3200mhz DDR4. Video card is still my 1080 Ti FE.

3DMark
Time Spy: old 8134, new 9332
Fire Strike Ultra: old 6662, new 6833
Fire Strike Extreme: old 12085, new 12783
Fire Strike: old 18889, new 22351

Cinebench R15
Multi
Old (4790K): 992
New (8700K) 1413

Single
Old (4790K): 176
New (8700K): 192

With the 3DMark scores we clearly see that CPU matters less and less as the resolution goes up and your GPU is the bottleneck. At 1080p the 8700's extra oomph makes a huge difference. At 4K, not so much. The Time Spy score is a nice bump though so the 8700K has something to say about DX12.

The real interesting score to me was the single threaded score in Cinebench. My 4790K score was on a OC'ed 4790K with all cores at 4.7ghz. The 8700K score is stock (which for this test is still 4.7ghz because that's the 8700K's single core boost) and still wins. So there is clearly IPC gains with Coffee Lake.

Overall the 8700K is a monster chip.

 

[little P]

Nice scores Brainling! That's a monster rig you've built yourself there

 

[Kitlope]

Late to the party but very nice Brainling. I've said it for a couple years now but since Coffee lake came out and 6 core becoming mainstream (The 8400/8500 is bang for your buck incredible), this is the year I need to upgrade. I like to see your scores, gives me inspiration.

Zero support for socket 1366 now by mb manufacturers and Spectre/Meltdown no "fix" by Microsoft cause too old. Time to designate this old chip to HTPC status but I'm not in a big rush. Would almost prefer to wait for the new 10nm but delayed until sometime in 2019... Two+ years of delays.

 

[Kitlope]

Ok with my latest upgrade to CPU I ran Firestrike version 1.1. Updated to latest NVidia GPU drivers. No overclocking CPU at this point but I'm thinking in the future it's a distinct possibility, even going to liquid cooling for the first time.

Intel 6 core 8086K, 16 GB DDR4 Ram @3200, Asus Z370-E MB and same Asus 1080ti OC edition that I've had since April 2017. It went from 17 618 with no OC using old 6 core up to 23 105 with new 6 core and no OC. I kinda figured it would be boosted about 30%.

 

[Kenadian]

So I was watching Linus Tech Tips on YouTube today and he's got a video about building the fastest Gaming PC money can buy right now.

It's very similar in build to mine except he's using higher frequency Ram and has 2 Asus Rog Strix 2080TI's.

I was expecting to be blown away when scoring against his machine but actually managed a higher score than his first try out with 3D Mark Time Spy 1.0.

Once he started tweaking the system though his was faster than mine but I'm satisfied to be even as close as I am.

Time Spy has me at a score of just over 11,100 with Fire Strike 1.1 scoring 23,954; currently in the top 2%.

Also ran Fire Strike Extreme Stress Test 1.1 and scored a Pass with 99.3% FPS stability.

Technically I'm running bone stock with the only exception being XMP enabled for my 3200 MHz Ram.

Results are all linked.

 

[little P]

Nice man! You must be well pleased with that

 

[Kenadian]

Nice man! You must be well pleased with that

Ya very much actually. This is by far the highest tier I've gone in a new build.

 

[Noodleguitar]

Ya very much actually. This is by far the highest tier I've gone in a new build.

It has to be that massive case.

 

[Brainling]

So I got my 2080 Ti plugged in. It's an EVGA Black Edition, so it came out of the box completely stock. It's basically the TU102 lottery. You get a GPU with a back plate and a decent cooler and it's on you to find what the boost OC for that chip is. When you buy a video card with an OC you are effectively paying for two services: a) that your chip WILL run at some level of offset, and that b) it may have been binned for much higher but NVidia has maximums that AIB partners can ship cards at, even if the chip will go much beyond that. If you're willing to play the lottery on the silicon you can save 300-400 bucks at the very top end. Given that most TU102's will overlock to something reasonable, I spun the wheel.

So far it hasn't been bad. I got +280mhz on the core and +800mhz on the memory stable. Oddly enough I was able to get +325/+850 in Time Spy Extreme (DX12), but in Fire Strike Extreme (DX11) it failed at that core and in Fire Strike Ultra it failed at that memory frequency. So I had to back it down to the eventual stable 280/800. I could probably get more out of it by playing with power stuff but I have no desire.

By the numbers I was able to get ~10% more out of the card with a very light OC session. Core clocks in the 1900mhz range under load with temps at 74C is decent silicon. It's not the best TU102 I've seen, but it's not the worst.

So taking my stock numbers from my 1080 Ti above, the 2080 Ti with this OC is almost exactly 30% faster at the top end. It also scales better at higher resolution. All as advertised.

2080 Ti (OC +280/800):
Time Spy Extreme: 6105
https://www.3dmark.com/3dm/32410427?

Time Spy: 13083
https://www.3dmark.com/3dm/32410515?

Fire Strike Ultra: 8626
https://www.3dmark.com/3dm/32410603?

Fire Strike Extreme: 16603
https://www.3dmark.com/3dm/32410715?

Fire Strike: 27407
https://www.3dmark.com/3dm/32410808?

I didn't have an FE Time Spy Extreme value but I was able to look at 3DMark's card list and see most FE's doing something around 4225, so we'll use that.

Time Spy Extreme: old 4225, new 6105 = 30.8%
Time Spy: old 9332, new 13083 = 28.6%
Fire Strike Ultra: old 6833, new 8626 = 20.8%
Fire Strike Extreme: old 12783, new 16603 = 23%
Fire Strike: old 22351, new 27407 = 18.5%

So we can extrapolate a few things here. First off the RTX cards scale best at higher resolutions in DX12. Not a huge shocker there. Interestingly Fire Strike Ultra and Extreme are weird outliers. I would expect those to be reversed.

For funsies I ran the RTX benchmark as well, stock and OC:

Port Royal (RTX - Stock): 7477
https://www.3dmark.com/3dm/32411065?

Port Royal (RTX - OC): 8348
https://www.3dmark.com/3dm/32411148?

Nothing to really compare them to, but OC'ing does effect RTX performance!

So overall, worth 1000 bucks? For most people, absolutely not. If you were already the type to buy 700+ dollar video cards though and you're right on that razor edge of performance (e.g. 4K and VR) than it all comes down to how easily you can part with 1000+USD. I happen to be in the former extremely thin category (willing to spend, doing 4K and VR) and so for me it's a nice boost to 4K which I play quite a few games at on my TV. We'll see how much of a boost I get in VR later tonight.