GeForce Titan: 2500K vs 4790K (66 games tested)

[Deders]

BFG10K: If HT was enabled for the 4790k what games of the ones you tested would not run? I'm curious as to what effect turning off HT affects fps.

I saw one test recently with Crysis 3 on the Welcome to the Jungle level where the hyperthreaded cores of a 4790k were helping keep the framerate at 60fps compared to a 4590k, both clocked to 4.6GHz. But as the tester noted it was an extreme example.

 

[BrightCandle]

Crysis is a particularly heavy game. I saw myself there are certain scenes where all 6 cores were maxed out and the bottleneck moved from the GPUs to the CPUs immediately. The fight in the grass with the aliens is a great example of such an area, that just seemed to really hammer on the CPU. Most of the game however seemed to be heavy on the GPU.

Its not just variance between games its variance from scene to scene. What a lot of these tests don't capture is how different the frame rate and ratio of CPU/GPU usage is throughout a game. BF4 is famously more CPU multithreaded in multiplayer than single player and I see the difference in CPU usage quite notably. Unless you test the right area you don't see the difference.

But can you imagine hunting through 50 or more games with 2 or more CPUs just to see the differences? Its a Frightening amount of work but about the only way to objectively show these differences. I am bothered with the lack of quality testing done in games with these CPUs, many people are coming to the conclusion that 6 core or 4core + HT isn't worth the investment based on benchmarks on big review sites with a small list of games. If we have seen anything in this thread its not sufficient to test 4 games and further to that I would say its not sufficient to test one scene in a game either.

 

[Barfo]

So I'm good for another 2 years of 1080p gaming. Cool.

 

[amenx]

The chart look miniscule on my end. Anybody else have this problem?

No your not the only one. Tried in FF, IE and Chrome and its miniscule. There is no "yellow line" either above it to expand like normal downsized pics. I even saved the image to my pc and tried to enlarge it and it got blurry as hell

edit: woops, all is OK on Chrome.

 

[amenx]

Not surprised at all on the results. @ 2560x1600 what did you expect?

 

[netxzero64]

This is a very nice read good job TS!

 

[tidy]

.

 

[netxzero64]

Everyone needs to be very mindful of the context of these benchmarks, especially the resolution, and the "actual settings" disclaimer. Some of these are clearly far from 'worst case' CPU scenarios. Worst case in Crysis with a stock 2500k is well below 62fps (without heavy detail trade-offs), and the same applies to UT2004, where you're not going to maintain 141fps in a large, 'worst case' match on that CPU. If you have a 1920x1080 monitor, for example, you should think twice before coming to the conclusion that the 2500k is still marvelous, and that there is little to be gained from CPU upgrades.

Even so, you can still somewhat gauge on where you stand in terms of performance that is why we continuously do researches and if possible, do some benchmarking or tests under our own rigs. This is just a simple illustration of it.

For me if you want to test CPU power, you should test it on games based on RTS. The TS obviously did not have one that is why we can't really see the performance difference as almost all the games tested were pretty much GPU bound.

 

[Phynaz]

Holy smokes that's a lot of testing. Thanks for posting.

 

[BFG10K]

BFG10K: If HT was enabled for the 4790k what games of the ones you tested would not run? I'm curious as to what effect turning off HT affects fps.

From past tests on older HT systems (not my own) I found some rare older games would exhibit odd performance issues which vanished when it was turned off.

Granted, this was on the "old" version of HT and I didn't actually try it on the 4790K, but I was in no mood to debug 66 games if I ran into issues.

Now that I know everything is fine it's possible I may revisit HT in the future, but for now it stays off.

 

[BFG10K]

The Crysis one concerns me. Did you choose one of the abundant CPU bound zones, or are you up to your usual tricks?

Like I stated in the other threads, the slowest performing areas in Crysis 1 for me are consistently GPU bound at my settings, and this is repeatedly reflected in my posted results.

 

[Phynaz]

The Crysis one concerns me. Did you choose one of the abundant CPU bound zones, or are you up to your usual tricks?

What kind of accusation is that?

 

[toyota]

Like I stated in the other threads, the slowest performing areas in Crysis 1 for me are consistently GPU bound at my settings, and this is repeatedly reflected in my posted results.

there are areas of Crysis that will drop into the upper 40s and low 50s even with a 2500k because of the cpu NOT the gpu. sounds crazy as someone said that last year and I had to test for myself to see and sure enough its true. the game only effectively uses 2 cores and even my 2500k at 4.4 could only maintain frame rates in the 50s in spots and gpu usage plummets. like many games it depends where you are testing though.

EDIT: just checked and with my 4770k at 4.4 the very lowest I hit was a one time only blip to 58 fps so for the most part it was in low to mid 60s in the exact area that put my 2500k at 4.4 into the 50s. gpu clocks will drop and usage still plummets into 60-70% range though.

had to do a print screen of the screenshot since the actual screenshot is too large of a file for free pic posting.


windows 7 screenshot

 

[ocre]

From past tests on older HT systems (not my own) I found some rare older games would exhibit odd performance issues which vanished when it was turned off.

Granted, this was on the "old" version of HT and I didn't actually try it on the 4790K, but I was in no mood to debug 66 games if I ran into issues.

Now that I know everything is fine it's possible I may revisit HT in the future, but for now it stays off.

i absolutely love your thread!

i am still using an old original i7 920. i play at 4ghz.

the take away for me is not to settle for a used 2600k as i was about to pull the finger on. thanks.

i would also love to see you add HT results when you get the chance. i think there is a lot of interest in this as well

Great job

also like to thank brightcandle for his contribution, really a great thread

 

[Spjut]

You could make it more simple for yourself and only try hyperthreading in the games reported to benefit from it (for example Crysis 3 and BF4)

Personally, I've wanted to see more benchmarks for the i7 920 in modern games

 

[RussianSensation]

i absolutely love your thread!

i am still using an old original i7 920. i play at 4ghz.

the take away for me is not to settle for a used 2600k as i was about to pull the finger on.

Ya, since you have a 760, you are better off getting a GPU 2x faster than spending money on a new mobo + 2600k. In a lot of games the performance increase over a mildly overclocked 1st gen i7 is not even there and that's with a 680. You need to play a lot of RTS, MMOs, ARMA3, BF4 multiplayer or have something much more powerful than a single Titan before a 920@4.0Ghz is a major bottleneck.

http://www.guru3d.com/articles_pages/far_cry_3_graphics_performance_review_benchmark,7.html

 

[escrow4]

I have a 4770 here @ 3.9GHz with MCE on all cores with a 780 Ti GHz and I wouldn't recommend an i5 buying new. You'd want at least a 4770 equivalent. All the big AAA titles are or will be using those threads. We are not at the hexa core tipping point yet. And I never play RTS's or MMO's or anything multiplayer so CPU bottlenecking for me is close to a non issue.

 

[Carfax83]

I have a 4770 here @ 3.9GHz with MCE on all cores with a 780 Ti GHz and I wouldn't recommend an i5 buying new. You'd want at least a 4770 equivalent. All the big AAA titles are or will be using those threads. We are not at the hexa core tipping point yet. And I never play RTS's or MMO's or anything multiplayer so CPU bottlenecking for me is close to a non issue.

I agree with this. The next gen engines are all designed to use more than four threads, so hyperthreading will be showing larger performance gains than it once did.

 

[avx81]

Thanks for the overview. I think it's fair to say that outside of special cases such as BF4 multi-player and Arma 3, strategy games and MMOs, PC gamers should spend $ on a faster GPU setup and then SSD before considering an upgrade from a modern CPU such as 2500K. Once overlcocked to 4.3Ghz+, it's going to be a wash really comparing i5 2500K vs. 4790K @ 4.8ghz. Most people upgrade now for features (M.2 / SATAe) or because they are bored

What would have been interesting to see is the comparison of minimum FPS over a period of time. Let's hope next generation PC games have more NPCs, more advanced AI and physics to give us reasons to upgrade from 1st and 2nd generation i5/i7s.

Which is what I have been saying for some time now however someone will always come along and state that you need a top of the line cpu to play computer games and try to justify it. Those charts show little to no difference across most games when comparing an older i5 to a new i7 that has significantly higher clock rates. When it comes to gaming your always better off spending the bulk of your money on the video card setup or even an ssd. Also when you get into games getting over 100fps I doubt many people here can say that they can see a serious difference between 140fps and 170fps. If you can than kudos to you because I cant and could care less as long as the game is running smooth around 60. Interesting to see games like bioshock or crysis that I play somewhat often have no fps difference or 3-4 tops.

 

[thedosbox]

No such thing on my end (Firefox and IE)

It's there on my firefox. Create a new profile and check again as it might be one of your add-ons/extensions.

 

[tidy]

.

 

[schmuckley]

Run @ stock=fail
Run@ highest stable clocks on air/water=win.
I'd bet the SB chip wins due to higher clocks.
f'rinstance this is what i run for daily:
http://valid.canardpc.com/m8tuvk
A lot of Haswell chips won't run past 4.4 for daily.
I had an above avg one and it would only do 4.7 max on water.
Thanks for doing all that testing, though.
It shows a level of dedication.

 

[YBS1]

While I can appreciate the lengthy testing performed, I find the setup rather dubious for a variety of reasons. For one it's telling us what we already know, you're likely to be GPU limited at higher resolutions/settings with a single GPU. I also don't see the point of disabling hyperthreading. I have seen reports of a small number of games allegedly performing worse with HT enabled, but I have never once been able to verify that in any such game. Therefore labeling this as i5 2500k vs. current top of the mainstream line i7 4790k seems disingenuous to me as it more resembles a mildly overclocked i5 4690.

Further, amongst users running higher than the current de facto standard of 1920x1080, I would expect multi-GPU usage to climb disproportionally as compared to those running lesser resolutions. Thus running multiple GPUs in your testing would have shifted the burden in many of those games back to the CPU. I would also imagine the type of users running such resolutions that are also buying "K" CPUs are more than likely overclocking, at least mildly, making default clock testing less than ideal for real world numbers. I can see the value in wanting to test for "real world results", but I've never liked seeing results where a CPU was clearly being hamstrung by a lack of GPU power (or vice versa). You see a lot of that in this forum when people are cherry picking benchmarks..... Link to a chart showing the top 9 CPUs all listed at "64fps". "See, it's just as fast!", they smugly proclaim like they are pulling the wool over someone's eyes.

 

[tidy]

.

 

[RussianSensation]

You see a lot of that in this forum when people are cherry picking benchmarks..... Link to a chart showing the top 9 CPUs all listed at "64fps". "See, it's just as fast!", they smugly proclaim like they are pulling the wool over someone's eyes.

That's the point - to analyze real world performance of a CPU not reduce IQ to show differences between CPUs in order to justify upgrades. We do not care about 1280x1024 CPU benches on a Titan since who will play at such resolution with such an expensive GPU. Of course for someone running 2 overclocked 780s/290s at 1080/1440, there will be a larger difference between an overclocked i5 and 4790k @ 4.8Ghz. However for the majority of single-GPU owners there is still little legitimate reason to upgrade the CPU platform from 2nd gen OC i5/7s. I would bet i5 2500k @ 4.5Ghz paired with dual 880s will destroy an i7 4790k @ 4.8Ghz or even i7 5930 @ 5ghz with 780Ti SLI in 98% of games.

As much as I would love to upgrade for a tangible performance increase, I will most likely upgrade my CPU because it's fun to play with new hardware, which is why X99 sounds like a fun upgrade for 6-core E-peen

Of course people who spend $1,000+ on GPUs are unlikely to blink for the $100 difference ($80 at MC) between and i5 and i7, which is why for most enthusiast users the i7 is an automatic choice.