Intel Core i7-9700K 9th Gen CPU Review: Eight Cores And No Hyper-Threading

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

mlee 2500

Honorable
Oct 20, 2014
298
6
10,785
I'm not sure that your CPU is the place to try and save $60....especially if you're a gamer spending $600+ on a GPU and over $100 on a cooler. If you're price conscious then there are plenty of other places you can save a few bucks and not entail a performance delta, or have to upgrade a year or two sooner.
 

BorgOvermind

Distinguished
Oct 4, 2011
46
0
18,540
CPUs should not be tested in FPS games. CPUs should be tested in strategy games, where you can spawn thousands of units and really test the computer power of the CPU. That way the video card won't matter that much in the benchmarks.
 

BorgOvermind

Distinguished
Oct 4, 2011
46
0
18,540
CPUs should not be tested in FPS games. CPUs should be tested in strategy games, where you can spawn thousands of units and really test CPU power. That way the video card will not influence the results that much and significant differences could be seen between CPUs.
 

InvalidError

Titan
Moderator

The only ways to avoid cache-related side-channel attacks is to either bypass the caches at the expense of a huge performance impact or security-harden the algorithm by interleaving decoys in the code to scramble the side-channel, which also significantly degrades performance and efficiency.

If you ask hardware to make these sorts of decisions by itself, you are going to end up with random severe performance degradation as the CPU mistakenly flags threads as possibly malicious.
 

mlee 2500

Honorable
Oct 20, 2014
298
6
10,785


I second that recommendation! For instance, Total War: Warhammer is a benchmark you only occasionally find in online CPU reviews, but which I always pay the most pay attention to.
 

mlee 2500

Honorable
Oct 20, 2014
298
6
10,785


Yep. If it was easy or incurred no overhead they would have fixed it by now.

 

paul horn

Distinguished
Sep 4, 2009
13
1
18,515
I understand the Hyper thread issue well knowing that AMD and INTEL both have issues with sucurity. It is the fact the complete RYZEN family does not support LINUX and might fail.
 

paul horn

Distinguished
Sep 4, 2009
13
1
18,515
That is only one of the processors. Note all RYZENS will run LINUX. And not all LINUX versions run RYZEN.
AMD claims it has fixed it but people are still having issues with the systems locking up.
 

logainofhades

Titan
Moderator
They used 2 different Ryzen's. They used a Ryzen 7 1800x and a Threadripper 1950x. Sounds like the distros need fixed to run AMD's Ryzen arch. That's why I never use Linux. I remember days of it wouldn't work, if you used a particular motherboard, or didn't run a certain bios revision for boards that did work. Linux, on a consumer level, will never be a viable alternative.
 

paul horn

Distinguished
Sep 4, 2009
13
1
18,515
Windows is junk and intrusive. AMD should not be allowed calling their processors compatible if it can't run existing OS's.
 

kanewolf

Titan
Moderator


Ryzen IS compatible with Linux. You just need a distro with a 4.18 or new kernel. I just built a Fedora 29 box with a Ryzen 2600 and B450 motherboard.
 
I was just looking over the numbers and wondering how they had their R7 2700X configured in its 4.2Ghz overclock. I've done a lot of rendering and workstation use with my 2700X which is only oveclocked in that it has PBO enabled and I have the RAM set at 3600Mhz. My 2700X will boost to 4.175 - 4.225 all core during rendering and heavy loads. When only 3 cores are utilized it will boost to 4.3 - 4.350Ghz. I have a negative offset and max out at 1.4V under load. My benchmark numbers are better to much better than the 4.2Ghz overclock 2700X used in this article. I think its very telling that in several benchmarks their stock 2700X outperformed their overclocked 2700X. That just shouldn't happen very often, and definitely shouldn't happen in over half their benchmark results. The overclocked R7 2700X in this article was far from stable.

Speaking more of overclocks and stock performance how it is fair to pair the i7 9700K and i9 9900K with a Godlike motherboard that costs $600 and is like the #1 offender of default overclocking those processors right out of the box. The default specification for the 9900K for example is 95W yet the Godlike will run it at 155W with a 4.7Ghz all core overclock strait out of the box. At its actual 95W limit the 9900K will only run at 4.0Ghz all core. By pure definition even the "stock" benchmark scores for the Intel 9th Gen 8 cores are in actuality overclocked scores. What we have in these graphs is 9th gen overclock and 9th gen extreme overclock, but we have no actual 9th gen stock benchmarks.
 


A review like this would doubtless show Intel with a slight IPC advantage leading to slightly better single core performance which would lead to slightly better gaming performance when locked at the same speed. It won't be a huge advantage, but gaming advantage Intel.

Now with regards to productivity it would be a different story. With a lot of Adobe apps or anything that utilizes a lot of single core execution Intel will edge the 2700X slightly. However in most rendering and heavily multithreaded apps the i7-9700K would be left in the dust and with its better multicore scaling ability the 2700X would pull away from the i9 9900K at locked clocks. It would be interesting to see because I too would like to know how far apart the 9900K and 2700X would be with locked clocks but the 2700X from all tests I've seen would be with winner. Productivity advantage AMD.
 

mlee 2500

Honorable
Oct 20, 2014
298
6
10,785


Yeah, more cores, less threads. Turns out for most games that's better...some of which actually incur a penalty with hyperthreading. On the other hand, massively parallel or concurrent applications and workloads typical of the HEDT space will suffer a bit. Basically it depends on what you do with your computer, but even in the latter case, the extra cores and higher frequencies compensate in all but long-running, non-sequential compute scenarios.
 
I would like to see a locked at same clocks review, for these CPU's. It seems you really have to push Intel, to handily surpass AMD. It reminds me of the FX vs Sandy bridge days.

Good luck getting Ryzen to 5.0GHz or higher on air or water. To reduce the speed to match Ryzen is not a fare comparison. It is like taking a 400 horse power engine [Intel] in a car and restrict the fuel flow so only 200 horse power is available so a 1970 Chevy Biscane with a small block V8 will have a chance against a Corvette.