Intel Core i9-7900X 3.3 GHz

[W1zzard]

The Core i9-7900X is where Intel's new HEDT processor lineup truly begins, delivering on all the platform's three main goals - more cores, more memory channels, and more PCIe lanes. We pit it against the Ryzen Threadripper 1950X for what will be the pitched battle of the $1000 giants.

Show full review

 

[TheGuruStud]

Man, those apps are brutally crappy. How many didn't scale at all? 1700 and 1800 winning or coming very close even though clocks should be the nearly the same as TR on low threads? CB score is woefully low for fast ram. It scores that with stock 2666.

Devs are bums.

Thanks for the list of exactly what software to avoid at all costs LOL

 

[W1zzard]

Man, those apps are brutally crappy. How many didn't scale at all? 1700 and 1800 winning or coming very close even though clocks should be the nearly the same as TR on low threads? CB score is woefully low for fast ram. It scores that with stock 2666.

Devs are bums.

Thanks for the list of exactly what software to avoid at all costs LOL

Lol, what do you propose we use for testing?

 

[TheGuruStud]

Lol, what do you propose we use for testing?

I wasn't busting your balls this time LOL. A lot of those are popular. It's just an observation. You can see how the high turbo clock of a couple cores on the 7900x shines in those apps (plus nice intel bump alone haha). I surely don't think we're seeing any XFR action, though. Did you look at clocks by chance?

So much for productivity when they're so limited.

 

[The Von Matrices]

Is that a bug in CPU-Z or did you disable two of the cores when overclocking?

 

[bug]

Man, those apps are brutally crappy. How many didn't scale at all?

Unfortunately, that is the real world, whether you like it or not.
But even so, there seem to be few reasons to pick this over a cheaper Threadripper.

 

[tomi832]

Lol, what do you propose we use for testing?

hey nice job!
I know that the question is not for me but I do have something in mind.
I think that a lot of the CPU tests are not that good...
like, 5 of the tests there are measured by milliseconds...not even seconds.
because of the fact that they are really fast, their devs only use a single thread...so all of the Ryzen products are basically killed there and the intel ones gain an unfair advantage at the summary of the CPU tests because of 4-5 tests that wont have any real difference for the user between the R3 1200 and the i9 7980XE...
I don't know enough about the web ones but they also feel like they are unnecessary...but IDK too much about them and maybe they do matter.
I'm saying it because sometimes people come and say for example "the 7900X is 10% faster than the 1950X look at it here" and show me the graphs from here, while its not true at all. a summary is great for comparison but its flawed if some of the benchmarks are unnecessary and make one side have an unfair disadvantage...

 

[bug]

hey nice job!
I know that the question is not for me but I do have something in mind.
I think that a lot of the CPU tests are not that good...
like, 5 of the tests there are measured by milliseconds...not even seconds.
because of the fact that they are really fast, their devs only use a single thread...so all of the Ryzen products are basically killed there and the intel ones gain an unfair advantage at the summary of the CPU tests because of 4-5 tests that wont have any real difference for the user between the R3 1200 and the i9 7980XE...
I don't know enough about the web ones but they also feel like they are unnecessary...but IDK too much about them and maybe they do matter.
I'm saying it because sometimes people come and say for example "the 7900X is 10% faster than the 1950X look at it here" and show me the graphs from here, while its not true at all. a summary is great for comparison but its flawed if some of the benchmarks are unnecessary and make one side have an unfair disadvantage...

Think about CPU as a classic ICE engine: you can add more cylinders to get more performance out of it, but individual cylinder performance still matters

 

[tomi832]

Think about CPU as a classic ICE engine: you can add more cylinders to get more performance out of it, but individual cylinder performance still matters

but I'm talking about checking things with ms...will you feel it if your car can run on 20 KM/h or 22 KM/h on the first gear? yet it will be 10% more powerful...maybe even the one with 20 KM/h will have a better acceleration and higher top speed...but people will say that the one with 22 KM/h is better and faster because of that...that's non-sense.
its better if you have a 22 KM/h first gear rather than 20, and in some cases the one with the 22 KM/h will perform better because you needed to be on the first gear for the power...but it wont change the fact that at full use and at most cases the 20 KM/h car will be better.
and I didn't talk about all the single-thread tests...I talked about the tests that are so fast that they measure them with ms (thus testing them wont matter because the average user wont feel any difference at all unless he tries...and maybe even than he wont) and that the devs made it single-threaded because it didn't need any more power rather than a single core (since it's so fast).

 

[bug]

but I'm talking about checking things with ms...will you feel it if your car can run on 20 KM/h or 22 KM/h on the first gear? yet it will be 10% more powerful...maybe even the one with 20 KM/h will have a better acceleration and higher top speed...but people will say that the one with 22 KM/h is better and faster because of that...that's non-sense.
its better if you have a 22 KM/h first gear rather than 20, and in some cases the one with the 22 KM/h will perform better because you needed to be on the first gear for the power...but it wont change the fact that at full use and at most cases the 20 KM/h car will be better.
and I didn't talk about all the single-thread tests...I talked about the tests that are so fast that they measure them with ms (thus testing them wont matter because the average user wont feel any difference at all unless he tries...and maybe even than he wont) and that the devs made it single-threaded because it didn't need any more power rather than a single core (since it's so fast).

They're still an indication of raw power. As @W1zzard above noted, they're the only ones available. So until something else emerges, we're stuck with what we have.
I see your point, but I feel a test suite without those tests would be even more misleading than a suite including those tests despite them being (arguably) not the best indicators.

 

[Vya Domus]

Lol, what do you propose we use for testing?

It would certainly be interesting doing something like running several , different instances of video encoding at the same time or running a game at the same time. This sort of testing would make much more sense for these sort of CPUs, after all they have insane amounts of cores/threads and it's pretty obvious that past a certain point the scaling stops when you run just one piece of software. Multitasking tests seem a much better fit for these kinds of products.

 

[Upgrayedd]

What is up with WD2 in 720p? Why is there such a difference between the 7700k and 6700k?

 

[EarthDog]

Thank for the testing.

8c/16t in overclocking?

I still am not a fan of game testing in 720p as it blatently exaggerates any gains found at higher resolutions. So people see the inflated numbers make the decision based off that when at 1080p, where most people play anyway, the difference is a lot less and higher resolutions its gone. Why intentionally create an unrealistic environment whose datapoints do not scale??

Im also incredibly surprised your aio held up at a whopping (for aio) 1.35v. We max out at 4.5ghz and 1.23V for our aio (h115i) as it breaks 90c in testing (povr and occt).

Multitasking tests seem a much better fit for these kinds of products.

doesnt rog realbench run three things at once?

 

[jigar2speed]

Wait no power consumption numbers of Intel 7900X @ 4.5GHZ Why ?

 

[bug]

I still am not a fan of game testing in 720p as it blatently exaggerates any gains found at higher resolutions.

That is just an indication about how well the CPU is doing when pushing game engines. It's not meant to be translated into overall gaming performance, but you're right, noobs won't read it that way. At the same time, I wouldn't get rid of those benchmarks, but I don't know how to better convey what they represent either. Maybe a disclaimer on the respective page(s)?

To add to what you just said, whoever runs these CPUs, is not even gaming at FHD, they're more likely to be in 4k territory. Yet, a more powerful CPU, even if it's not the bottleneck in those scenarios, will still have more spare HP to do other things in the background.

Finally, I think real world usage is not even the target of benchmarks. Benchmarks simply highlight the basics, the "building blocks" of real world usage. We're simply supposed to pick the "building blocks" that make up our own usage and decide on what to buy based on that.

 

[EarthDog]

Wait no power consumption numbers of Intel 7900X @ 4.5GHZ Why ?

At 4.5ghz using 1.23V (and all cores/threads) running default test in occt, i am hitting 320W at the wall, so a bit under 290W for the entire system.

That is just an indication about how well the CPU is doing when pushing game engines. It's not meant to be translated into overall gaming performance, but you're right, noobs won't read it that way. At the same time, I wouldn't get rid of those benchmarks, but I don't know how to better convey what they represent either. Maybe a disclaimer on the respective page(s)?

To add to what you just said, whoever runs these CPUs, is not even gaming at FHD, they're more likely to be in 4k territory. Yet, a more powerful CPU, even if it's not the bottleneck in those scenarios, will still have more spare HP to do other things in the background.

Finally, I think real world usage is not even the target of benchmarks. Benchmarks simply highlight the basics, the "building blocks" of real world usage. We're simply supposed to pick the "building blocks" that make up our own usage and decide on what to buy based on that.

Test at 1080p as well.

But yeah, what it does or can do at 720p is no indication of anything outside of 720p gaming...which nobody in their right mind has this cpu and whatever high-end gpu was used and runs at 720p. Its not building on anything when the foundation doesnt support the upper floors...or the sheets dont match the curtains...

Edit: Spare horsepower on a decacore while gaming is a joke. Even with a game that uses 6 cores (not many) there is still an entire '7700k' worth of untapped performance on hand.

 

[natr0n]

What were the temps ?

 

[Vya Domus]

doesnt rog realbench run three things at once?

Don't know about that benchmark , but it would still be just a static benchmark. What I mean is combinations of real world software.

Edit: Spare horsepower on a decacore while gaming is a joke. Even with a game that uses 6 cores (not many) there is still an entire '7700k' worth of untapped performance on hand.

Exactly , encode a video with that spare 7700K.

 

[bug]

Don't know about that benchmark , but it would still be just a static benchmark. What I mean is combinations of real world software.

You do realize you're simply opening up the issue of what combinations are legit and which aren't. It's really just a can of worms.

Exactly , encode a video with that spare 7700K.

Well, at least it tells you you can leave that AV do its thing while you're gaming, or a piece of backup software or something.

 

[EarthDog]

Testing at 720p wouldnt remotely tell you that

Rog realbench runs 3 things at once... all real world testing... at least 2 of three iirc.

 

[FR@NK]

Overclocking all cores to 4.5GHz is not the best way to clock these processors as you lose the added performance from turbo boost 3.0.

Mine is setup for one core to run at 5GHz and another to do 4.9GHz while the rest are at 4.3GHz. This allows the turbo boost driver to assign demanding threads to the fastest cores while keeping the parallel workloads and less demanding threads on the slower cores. I could push the other cores higher but I might have to lower the faster cores and also I like to keep the voltage and the temps down. Since my x299 board lets you set VID and multiplier on a per core basis; you can really fine tune the best cores for max performance.

 

[bug]

Testing at 720p wouldnt remotely tell you that

Not directly, but it tells you which CPU has a better chance of being able to do that, that's all.

 

[EarthDog]

Does boost 3.0 work when manually overclocked? As long as turbo is enabled i guess..?

Not directly, but it tells you which CPU has a better chance of being able to do that, that's all.

we'll agree to disagree.

 

[FR@NK]

Does boost 3.0 work when manually overclocked? As long as turbo is enabled i guess..?

Yea it does. Look at the core usage.

See if you can push some of your cores abit higher. You might as well just get a 8700k if you want to lock all your cores at the same speed.

 

[raptori]

Why we don't have thermal tests for CPUs at TPU ? , for idle , stock and OC for say two coolers one popular air-cooler and one water-cooler .

At least to see when intel will decide to change their toothpaste to a real thermal one .