9900K not boosting past 4.7ghz

[Zando_]

4 hours ago, maizenblue said:

Basically the boost speed is 4.7 at stock for all intents and purposes. So i basically paid 325 dollars over a 3700x set up for an extra 300mhz clock speed. Yay intel.  

4.7 is the stock all-core boost. It's what you'll see them hit in any benches using more than a few cores. I think 5Ghz is a single core boost though, so if you're not getting over 4.7 in single core stuff then that is being funky. 5.0 all core is MCE or manual overclocking though, the only 9900-anything to do 5Ghz all core stock is the KS. 

Is there a reason you bought a K SKU and don't plan to overclock? What's the main use case for your rig? If it's gaming alone then an OC would be highly advised, CPU overclocks aren't too hard to get stable. 

Also pretty sure the funkiness you've observed is from using a bunch of bloatware software. Intel's Turbo Boost isn't like AMD's PBO. It hits what it's rated to, it doesn't magically boost higher or lower based on bin from what I've seen, unless temps are high enough it throttles. All my Intel CPUs behave normally with their Turbo Boosts, though I do rarely run them stock. 

[Guest]

10 minutes ago, maizenblue said:

Thanks. As much as I have buyers remorse for getting a 9900k, I did have sound reasons why I went with it over AMD initially. AMD still hasnt solved the memory latencey issues completely with Ryzen and this is reflected in some aspects of its performance, and seems like it would always be a small handicap vs intel with this generation of ryzen. The 3700x-3900x also still has lower IPC compared to Intel. I've seen tests at 4ghz with Ryzen 3000 vs Intel that still show intel beating them handily, so its not JUST clock speed. 

 

The buyers remorse comes from the fact I spent WAY more on a mobo than I planned when I already commited to intel and bought the 9900k. The lower mid range mobos just all had their fatal flaws so I was kind of stuck getting an expensive higher end one. Ryzen on the other hand has a couple good lower mid range mobos that would have been perfectly fine with a 3700x. In hindsight its just not remotely worth what I paid. Lesson learned.

 

The second reason is I suspect I really just did get a bad batch of silicon with my particular 9900k. Boost speeds aside, It just benchmarks a lot lower than others Ive seen in things like cinebench and timespy at stock settings. I've done so much messing with windows, bios and other things short of OCing that I think thats just the straw I drew. My Cinebench scores are particularly crappy, despite some fairly high speed memory, a good CPU cooler and a high quality motherboard.

Shit all this worry over a cinebench score? 

Well, set the priority for the benchmark to high or real time, that should get you some more points.

Your memory speed is pretty much the normal for both AMD and Intel platforms, we can call it average speed. 

Not that memory speed and latency plays a very large role, I gain probably 30-50 points but well above 3600mhz.

 

If you're kind of benching to see performance, I'd like to suggest you to set up a Windows 7 drive and use that for strictly benchmarking. W10 is way heavy of an operating system for adventures of pushing clocks and getting scores. Granted, it's old and not secure, but strictly for benchmarking your hardware, W7 would be a better route in this aspect. 

 

You didn't really waste moneys here. That setup will retain it's value for quite a while. Look at 7700K. They still go used (US) for around 250 on fleabay, new 400-450 Newegg (US). The 8700K goes for more now than it did at release time. Funny, break even or maybe a profit?? (probably not a profit, but get most moneys back after a year or two even).

 

Anyhow, you are only stating that these arbitrary numbers bother you. You shouldn't let that happen. Just arbitrary numbers you're chasing after. Maybe worry less about that and focus on how the system handles the tasks you built it for.

 

My 2 cents.

 

[maizenblue]

32 minutes ago, Zando Bob said:

4.7 is the stock all-core boost. It's what you'll see them hit in any benches using more than a few cores. I think 5Ghz is a single core boost though, so if you're not getting over 4.7 in single core stuff then that is being funky. 5.0 all core is MCE or manual overclocking though, the only 9900-anything to do 5Ghz all core stock is the KS. 

Is there a reason you bought a K SKU and don't plan to overclock? What's the main use case for your rig? If it's gaming alone then an OC would be highly advised, CPU overclocks aren't too hard to get stable. 

Also pretty sure the funkiness you've observed is from using a bunch of bloatware software. Intel's Turbo Boost isn't like AMD's PBO. It hits what it's rated to, it doesn't magically boost higher or lower based on bin from what I've seen, unless temps are high enough it throttles. All my Intel CPUs behave normally with their Turbo Boosts, though I do rarely run them stock. 

Its not that I plan on never overclocking, its that I feel like I shouldnt need to right now. This CPU is a relatively hot running power hungry beast as it is. What kind of annoys me is the fact it seems to be performing below expectations in games like battlefield 1. I wish to hell I would have paid close attention to how my 4790k performed before I took it apart. Id love to do a comparison, but it doesnt FEEL any better and possibly even worse. I could be wrong though because I never really paid close attention to FPS before now. 

 

I know, first world problems right lol.

 

EDIT: I also had a strange experience putting together this rig. The CPU retention arm was insanely hard to pull down. Like 4 times harder than the previous 3 cpus ive installed. It was so bad that I actually stopped a couple times and made sure my processor was properly seated and the arm hinge was working fine. Then after Iatched the arm down the tangs that actually push the CPU into the socket seemed really off center. With one side BARELY on the edge of the CPU and the other one towards the middle. It was unsettling, like there could be a little bit of sidewards pressure on it. I just said fuck it thats just how it is and its not going anywhere so I should be fine. I still almost feel like maybe the retention arm was slightly out of specs. 

 

Doubt that has anything to do with its performance, but Its still in the back of my mind. 

 

 

[Zando_]

3 minutes ago, maizenblue said:

Its not that I plan on never overclocking, its that I feel like I shouldnt need to right now. This CPU is a relatively hot running power hungry beast as it is. What kind of annoys me is the fact it seems to be performing below expectations in games like battlefield 1. I wish to hell I would have paid close attention to how my 4790k performed before I took it apart. Id love to do a comparison, but it doesnt FEEL any better and possibly even worse. I could be wrong though because I never really paid close attention to FPS before now. 

 

It should be a good bit better, BF1 and BFV take advantage of anything with more than 4 cores. I don't see how it'd feel worse unless something is funky with your setup. 

AFAIK they aren't hot running stock, what cooler are you using? People manage to stuff these in tiny tiny SFF rigs on teeny coolers and then run them stock with okay temps, or undervolt for decent ones. I haven't ever used a 9900K myself tho, so your personal experience is more valuable here. Still, my various Xeons, 4930K, 5820K, 5960X, 8600K, all were basically cool as ice until I started pumping more voltage in for overclocking. The 8700K I had for a short while was hot but that was also in a 15.6" laptop chassis... so yeah. Did have another 8700K for a very, very short while and I don't recall that ever hitting bad temps in my testing or in the previous owner's use. That was on a Cryorig H5 Ultimate, which is a midrange tower, pretty beefy but not anything like an NH-D15 or Dark Rock Pro 4 (or 280mm+ AIO, which are comparable to the big air towers). 

[maizenblue]

1 minute ago, Zando Bob said:

It should be a good bit better, BF1 and BFV take advantage of anything with more than 4 cores. I don't see how it'd feel worse unless something is funky with your setup. 

AFAIK they aren't hot running stock, what cooler are you using? People manage to stuff these in tiny tiny SFF rigs on teeny coolers and then run them stock with okay temps, or undervolt for decent ones. I haven't ever used a 9900K myself tho, so your personal experience is more valuable here. Still, my various Xeons, 4930K, 5820K, 5960X, 8600K, all were basically cool as ice until I started pumping more voltage in for overclocking. The 8700K I had for a short while was hot but that was also in a 15.6" laptop chassis... so yeah. Did have another 8700K for a very, very short while and I don't recall that ever hitting bad temps in my testing or in the previous owner's use. That was on a Cryorig H5 Ultimate, which is a midrange tower, pretty beefy but not anything like an NH-D15 or Dark Rock Pro 4 (or 280mm+ AIO, which are comparable to the big air towers). 

I have a noctua d15s. I also have a case with extremely good airflow. A corsair 450D with a really well ventilated front cover.(Its a dust magnet). It has 2 140 mm fans blowing directly on the CPU and GPU. Also have a 120mm noctua exhaust fan. Its a total wind tunnel and with low ambient temps this time of year. 

 

Still I see it spike to 79C running cinebench or gaming at times. By contrast my 4790k with a crappy h60 cooler NEVER got higher than 71C that I saw in all the years I owned it, and that cpu was also known for running hot. I doubt temps have anything to do with performance in this case though. 

 

[Guest]

2 minutes ago, maizenblue said:

I have a noctua d15s. I also have a case with extremely good airflow. A corsair 450D with a really well ventilated front cover.(Its a dust magnet). It has 2 140 mm fans blowing directly on the CPU and GPU. Also have a 120mm noctua exhaust fan. Its a total wind tunnel and with low ambient temps this time of year. 

 

Still I see it spike to 79C running cinebench or gaming at times. By contrast my 4790k with a crappy h60 cooler NEVER got higher than 71C that I saw in all the years I owned it, and that cpu was also known for running hot. I doubt temps have anything to do with performance in this case though. 

 

Your 4790K was half the transistor density.............................................

[Zando_]

55 minutes ago, maizenblue said:

Still I see it spike to 79C running cinebench or gaming at times. By contrast my 4790k with a crappy h60 cooler NEVER got higher than 71C that I saw in all the years I owned it, and that cpu was also known for running hot. I doubt temps have anything to do with performance in this case though. 

Aha. Seems to be normal stock temps, review I just looked at quoted 76C max on an NH-D15S (same finstack, offset with only one fan, is like 3C hotter than the D15), but reviewers are often using an open bench so temps are slightly lower. Still well below unsafe temps. I guess I'm used to it? My 5960X hits up to 77-78C max when running the same clocks (4.7Ghz all core) and it's on water.

 

55 minutes ago, ShrimpBrime said:

Your 4790K was half the transistor density.............................................

^^^ Also this. 9900K is literally twice the CPU, running higher clocks as well, though thanks to being yet more refined it will hit those at a lower voltage.

[maizenblue]

One odd thing I noticed is that in Ai suite under CPU usage it shows 18 circles with a varying percentages. Are those cores or threads? One of them reads like 6 digits that spin up and down to say 75369 and then down to say 31675 continually. What does that mean?

[Zando_]

4 hours ago, maizenblue said:

One odd thing I noticed is that in Ai suite under CPU usage it shows 18 circles with a varying percentages. Are those cores or threads? One of them reads like 6 digits that spin up and down to say 75369 and then down to say 31675 continually. What does that mean?

Haven't the foggiest. Only time I've ever used what I think (maybe?) was AI Suite was to fuck around with fan control on my CH7. For monitoring, use HWiNFO, HWMonitor, CPU-Z (GPU-Z for GPUs), etc. HWiNFO has the most information and is one of the most accurate (also if you have a streamdeck there's plugins to spit data from it out there), CPU-Z is the best for eyeballing stuff on your CPU/RAM, GPU-Z has a bunch of GPU-specific data in an easier to read format than HWiNFO. HWMonitor is HWiNFO with less information but easier to read.

Links:

• https://www.hwinfo.com/download/
• https://www.cpuid.com/softwares/hwmonitor.html
• https://www.cpuid.com/softwares/cpu-z.html
• https://www.techpowerup.com/gpuz/
   

[maizenblue]

6 hours ago, Zando Bob said:

Haven't the foggiest. Only time I've ever used what I think (maybe?) was AI Suite was to fuck around with fan control on my CH7. For monitoring, use HWiNFO, HWMonitor, CPU-Z (GPU-Z for GPUs), etc. HWiNFO has the most information and is one of the most accurate (also if you have a streamdeck there's plugins to spit data from it out there), CPU-Z is the best for eyeballing stuff on your CPU/RAM, GPU-Z has a bunch of GPU-specific data in an easier to read format than HWiNFO. HWMonitor is HWiNFO with less information but easier to read.

Links:

• https://www.hwinfo.com/download/
• https://www.cpuid.com/softwares/hwmonitor.html
• https://www.cpuid.com/softwares/cpu-z.html
• https://www.techpowerup.com/gpuz/
   

What about voltages. I always see people obsess over voltage when ocing. Is there any useful info there?  Ie if my voltage is high going to the cores even at stock speeds i can determine that i did just lose the silicon lottery?

 

hwmonitor will tell me the highest voltage reached for each core when im gaming or benchmarking. 

[unclewebb]

All active cores get their voltage from the same voltage rail. Different cores are not being fed different voltages. HWMonitor differences are because it is sampling at slightly different times.

 

Some  CPUs that require lots of voltage can still overclock just fine. The difference between 5000 MHz and 5200 MHz is 4%. The difference between an average overclocker and a really good overclocker will be barely noticed in any real world use.