"We're saying that the performance shown does not represent a defective CPU. Although those scores are from an unofficial 3rd party software that is not connected to Dell. Basically, comparing a Dell machine to other machines on the market using benchmarks is not something we can support or use as backup to replace your CPU. As far as we can see there's a difference in performance, but this is due to the setup used to run the benchmark on a different machine. If the comparison were made with another exact same machine, you'd likely get the same results as this is within the expected"
Essentially, they are saying that this CPU (i9 9900k) from any other R8s would give me the same results. Ie. 1750 in cinebench R15, below 3000 in cinebech R20, below 9000 in TimeSpy CPU score. My machine is still well within the warranty period.
What are the settings you dialed in for your 9900K when running those benchmarks? Just setting OC2 profile from AWCC and then benchmarking against a sea of enthusiasts and seasoned overclockers is asking to be steamrolled by their scores.
Have you installed Intel XTU and monitor the PL1 power limit, throttles (power/ thermal/ current). When your cores boost, are all cores hitting 4.8GHz? What about AVX offset? What is the TDP draw and Vcore when the CPU is under load? Post some images of the graphs XTU or OCCT is giving you when you run those benchmarks.
I'll post some images later. The OC2 setting is to ensure that the atleast 2 cores hit 5.0 GHz. I am not pitting this against any overclocker. The basic non OC'ed score for a i9 9900K should hit 2000 mark in Cinebench R15 (around 8800 in Time Spy - should be getting 11000 on a stock non OC'ed 9900K). The single core score is even more hilarious. I am getting half the expected numbers.
Again, I am not trying to break any records. I just want to make sure that my CPU behaves the way it should be behaving.
I am getting around 1700s in R15, and less than 3000 in R20
Our scores are too near to just sweep this under the rug. If you have HWmonitor installed like I have snapshot above, can you see what is the power draw and frequency for all the cores while the benchmark is running? Something is amiss here.
In R20, we could still debate some between 6C/6T and 8C/16T because it looks like AVX is in effect for R20. I'm saying that because my cores are blazing 4.9GHz in R15 but then gets slowed down to my AVX 4.6GHz limit in R20. AVX structures are mostly shared in multithreaded machines so you could see CPUs with more physical cores pulling ahead, but not so much with logical cores.
One other thing I can think off is the difference in time between Toms, 3DGuru, Anand writing those reviews and today. Anyone knows where to find any article of impact to Cinebench due to microcode patches related to Spectre, Meltdown, and Plundervolt security vulnerabilities?
Yes, that's exactly correct. I am getting around 1700s in R15, and less than 3000 in R20 (I should be able to hit almost 5000 on a stock non OC'ed 9900K). Just google any review (Toms, 3DGuru, anything.....).
I am not sure what Dell is talking about. This CPU is pretty much the best gaming CPU's out there... reading their response makes absolutely no sense at all (they believe this is the expected result).
Guys, if that's the case, then maybe I should pay for the i7 9700K price.. since that's what I am getting.
I did monitor the CPU using XTU during my tests. That's why I can say for sure that the CPU is defective. It never made it to 5 Ghz. The voltage did go to 1.215. I just want Dell to give me an answer... how are these results considered to be acceptable? You guys are marketing this as a performance oriented desktop.
1. The Time Spy score is too low for your CPU. I've been staring so hard at Time Spy 2 weeks ago when benchmarking my new 2080S card that I've seen the Galax words at least 200 times.
2. You may be on to something here:
I did monitor the CPU using XTU during my tests. That's why I can say for sure that the CPU is defective. It never made it to 5 Ghz. The voltage did go to 1.215.
At 1.215V, the CPU will absolutely not make it to 5 GHz. The 14nm++++++ process and Skylake microarchitecture requires somewhere north of 1.3V to hit > 4.8 GHz. There could be something capping off the voltage. Either from the supply side (think PSU/ motherboard VRM/ BIOS power management) or the sink side (think CPU throttle settings/ power management unit/ p-state bug). You need to find a way to make your CPU "ride" the voltage-frequency curve, all the way from idle sub 0.8V to high 1.35V. If you consistently see the CPU being pegged at 1.215V, there could be a defect, either in the OS setting (reformat could help), motherboard (ball in Dell's court) or you really landed a CPU that has a borked p-code engine (which if you can eliminate the earlier 2 causes, you can bring it up to Intel directly).
@RedGreenApple Yes, that's exactly correct. I am getting around 1700s in R15, and less than 3000 in R20 ...Guys, if that's the case, then maybe I should pay for the i7 9700K price.. since that's what I am getting.
And on that note, Cinebench R20 score on an Aurora R7, running an i7 with no overclock: 3368
I just adjusted the voltage to 1.3. Still about around the same score, it actually got a bit worse Getting low 2000s now in Cinebench R20 Again, nothing goes pass 4.7.
Still haven't heard back from Dell. I was told that they will call me back today.
The tech support person spent 1.5 hours with me yesterday troubleshooting. While sharing my screen, I was literally showing him Cinebenchs, and 3DMark. Interestingly enough, it was his first time seeing these programs.
He wanted to see SupportAssist... I am not sure how this tool can help me.
@Anonymous What was your CPU max temperature running the R20 test?
Yeah, I was wondering the same thing. My first thought at such a low score (and what I suspect to be the issue) is the AIO cooler block isn't fully touching his CPU or incorrect pasting, but without knowing if he was getting thermally throttled I had refrained from mentioning my speculation.
Still haven't heard back from Dell. I was told that they will call me back today.
The tech support person spent 1.5 hours with me yesterday troubleshooting. While sharing my screen, I was literally showing him Cinebenchs, and 3DMark. Interestingly enough, it was his first time seeing these programs.
He wanted to see SupportAssist... I am not sure how this tool can help me.
That interaction alone speaks for itself. Either the technical support person isn't very well versed technically or the tech support team is given instructions not to acknowledge any software other than Support Assist.
Here's the thing: Support Assist does not reveal performance bugs. I've mentioned this before: my Optiplex with GTX1050 was giving 9 FPS in Support Assist test while the iGPU was throwing out 20 FPS. Dell tech support told me there was nothing wrong with my system per their software logging. I had to point out that the discrete graphics was way behind compared to integrated graphics, and that I have a lot of experience designing chips, and I can at that moment make a detour on my morning commute to their Silicon Valley HQ before the person on the line quickly agreed to send a technician over to my place to take a look at the system.
Turns out the heat sink wasn't pasted correctly on the GPU. Which is what @Anonymous and @r72019 alluded to in their posts today. Do let us know how the temps are. Or if you're comfortable handling hardware, repaste it. Just beware the electrostatic monster. It's going rampant in some parts of the US now due to very dry weather.
Here's the temp during Cinebench run. It goes between 80 to 90 C (see screen). I think that's normal for a 9900. I am fine re-pasting the CPU, but I think Dell should be accountable for selling me a non functional unit. This is actually my first time buying a pre-assembled PC. I can't believe the way support team works at Dell.
Cinebench runs for such a short period of time, 90 is really high for a cinebench single run. It should start throttling somewhere around that temperature say +/- 5* (not sure exactly) with a critical system shutdown around 105 +/- 5*.
You can try running something more stressful in conjunction with msi afterburner and check for downspikes in the msi gpu/cpu graphs.
I suspect there is an issue with the paste job or aio cooler installation.
My cpu with 5ghz stock boost doesn't even reach 60 on a cinbench first run, and idles under 30 with watercooling, as an aside.
RedGreenApple
24 Posts
1
December 26th, 2019 10:00
I just want to put this on record:
This is literally what Dell said to me:
"We're saying that the performance shown does not represent a defective CPU. Although those scores are from an unofficial 3rd party software that is not connected to Dell. Basically, comparing a Dell machine to other machines on the market using benchmarks is not something we can support or use as backup to replace your CPU. As far as we can see there's a difference in performance, but this is due to the setup used to run the benchmark on a different machine. If the comparison were made with another exact same machine, you'd likely get the same results as this is within the expected"
Essentially, they are saying that this CPU (i9 9900k) from any other R8s would give me the same results. Ie. 1750 in cinebench R15, below 3000 in cinebech R20, below 9000 in TimeSpy CPU score. My machine is still well within the warranty period.
This will be my last purchase from Dell.
GTS81
2 Intern
•
2.2K Posts
0
December 26th, 2019 10:00
What are the settings you dialed in for your 9900K when running those benchmarks? Just setting OC2 profile from AWCC and then benchmarking against a sea of enthusiasts and seasoned overclockers is asking to be steamrolled by their scores.
Have you installed Intel XTU and monitor the PL1 power limit, throttles (power/ thermal/ current). When your cores boost, are all cores hitting 4.8GHz? What about AVX offset? What is the TDP draw and Vcore when the CPU is under load? Post some images of the graphs XTU or OCCT is giving you when you run those benchmarks.
RedGreenApple
24 Posts
0
December 26th, 2019 17:00
I'll post some images later. The OC2 setting is to ensure that the atleast 2 cores hit 5.0 GHz. I am not pitting this against any overclocker. The basic non OC'ed score for a i9 9900K should hit 2000 mark in Cinebench R15 (around 8800 in Time Spy - should be getting 11000 on a stock non OC'ed 9900K). The single core score is even more hilarious. I am getting half the expected numbers.
Again, I am not trying to break any records. I just want to make sure that my CPU behaves the way it should be behaving.
GTS81
2 Intern
•
2.2K Posts
0
December 26th, 2019 19:00
Here are my Cinebench runs on i5-9600K + 2080S. It's not a good comparison but it should serve as your lower limit. I only ran the CPU bench.
1. R15 with AWCC OC2 profile + Performance fan curve. 4.8GHz 2 cores. Remaining 4 cores 4.4GHz.
Score=1039
2. R15 with custom XTU OC profile + custom fan curve. 4.9GHz all cores.
Score = 1100
2. R20 with custom XTU OC profile + custom fan curve. 4.9GHz all cores.
Score = 2503
@RedGreenApple :
Our scores are too near to just sweep this under the rug. If you have HWmonitor installed like I have snapshot above, can you see what is the power draw and frequency for all the cores while the benchmark is running? Something is amiss here.
In R20, we could still debate some between 6C/6T and 8C/16T because it looks like AVX is in effect for R20. I'm saying that because my cores are blazing 4.9GHz in R15 but then gets slowed down to my AVX 4.6GHz limit in R20. AVX structures are mostly shared in multithreaded machines so you could see CPUs with more physical cores pulling ahead, but not so much with logical cores.
One other thing I can think off is the difference in time between Toms, 3DGuru, Anand writing those reviews and today. Anyone knows where to find any article of impact to Cinebench due to microcode patches related to Spectre, Meltdown, and Plundervolt security vulnerabilities?
RedGreenApple
24 Posts
0
December 26th, 2019 19:00
Yes, that's exactly correct. I am getting around 1700s in R15, and less than 3000 in R20 (I should be able to hit almost 5000 on a stock non OC'ed 9900K). Just google any review (Toms, 3DGuru, anything.....).
I am not sure what Dell is talking about. This CPU is pretty much the best gaming CPU's out there... reading their response makes absolutely no sense at all (they believe this is the expected result).
Guys, if that's the case, then maybe I should pay for the i7 9700K price.. since that's what I am getting.
RedGreenApple
24 Posts
0
December 26th, 2019 20:00
Thanks for posting this. This is my spec: i9 9900K, RTX 2080 TI, 32GB DDR4 (dual channel), SSD/HD combo.
Here are some of my results that I shared with Dell's folks:
The images are not working on this site, so I am using google instead.
TimeSpy (expected value for CPU should be around 11000, I am getting 8800):
https://drive.google.com/open?id=1hBHRhqRupYQRMdqeNub-Osw3GMsNNHF-
Cinebench R20 (not even getting 3000):
https://drive.google.com/open?id=1bw2QkOAXx-XjEMlzy70HrpdAb_kjqJyM
Cinebench R15
https://drive.google.com/open?id=1_5ih3TvZiBaXttyYHdC9xbZ6w56ZEmOV
Now check out the single core scores.... this is absolutely unacceptable!!!!!:
Cinebench R20 (getting 250, should be in the 500s):
https://drive.google.com/open?id=1hU0bps5iA2wqvebgVVSldixh2Qp8tZTZ
Cinebench R15 (getting 92, should be over 200):
https://drive.google.com/open?id=10jdAU9ptdVCZpoaZ8BR4YZuVUM542LMm
I did monitor the CPU using XTU during my tests. That's why I can say for sure that the CPU is defective. It never made it to 5 Ghz. The voltage did go to 1.215. I just want Dell to give me an answer... how are these results considered to be acceptable? You guys are marketing this as a performance oriented desktop.
RedGreenApple
24 Posts
0
December 26th, 2019 20:00
One thing to keep in mind is that the GPU is absolutely on point. This issue is strictly around the CPU.
GTS81
2 Intern
•
2.2K Posts
0
December 26th, 2019 21:00
@RedGreenApple :
1. The Time Spy score is too low for your CPU. I've been staring so hard at Time Spy 2 weeks ago when benchmarking my new 2080S card that I've seen the Galax words at least 200 times.
2. You may be on to something here:
At 1.215V, the CPU will absolutely not make it to 5 GHz. The 14nm++++++ process and Skylake microarchitecture requires somewhere north of 1.3V to hit > 4.8 GHz. There could be something capping off the voltage. Either from the supply side (think PSU/ motherboard VRM/ BIOS power management) or the sink side (think CPU throttle settings/ power management unit/ p-state bug). You need to find a way to make your CPU "ride" the voltage-frequency curve, all the way from idle sub 0.8V to high 1.35V. If you consistently see the CPU being pegged at 1.215V, there could be a defect, either in the OS setting (reformat could help), motherboard (ball in Dell's court) or you really landed a CPU that has a borked p-code engine (which if you can eliminate the earlier 2 causes, you can bring it up to Intel directly).
r72019
6 Professor
•
5.3K Posts
0
December 26th, 2019 21:00
@RedGreenApple Yes, that's exactly correct. I am getting around 1700s in R15, and less than 3000 in R20 ...Guys, if that's the case, then maybe I should pay for the i7 9700K price.. since that's what I am getting.
And on that note, Cinebench R20 score on an Aurora R7, running an i7 with no overclock: 3368
RedGreenApple
24 Posts
0
December 27th, 2019 07:00
I just adjusted the voltage to 1.3. Still about around the same score, it actually got a bit worse Getting low 2000s now in Cinebench R20 Again, nothing goes pass 4.7.
RedGreenApple
24 Posts
0
December 27th, 2019 16:00
Still haven't heard back from Dell. I was told that they will call me back today.
The tech support person spent 1.5 hours with me yesterday troubleshooting. While sharing my screen, I was literally showing him Cinebenchs, and 3DMark. Interestingly enough, it was his first time seeing these programs.
He wanted to see SupportAssist... I am not sure how this tool can help me.
r72019
6 Professor
•
5.3K Posts
0
December 27th, 2019 19:00
@Anonymous What was your CPU max temperature running the R20 test?
Yeah, I was wondering the same thing. My first thought at such a low score (and what I suspect to be the issue) is the AIO cooler block isn't fully touching his CPU or incorrect pasting, but without knowing if he was getting thermally throttled I had refrained from mentioning my speculation.
GTS81
2 Intern
•
2.2K Posts
0
December 27th, 2019 22:00
@RedGreenApple :
That interaction alone speaks for itself. Either the technical support person isn't very well versed technically or the tech support team is given instructions not to acknowledge any software other than Support Assist.
Here's the thing: Support Assist does not reveal performance bugs. I've mentioned this before: my Optiplex with GTX1050 was giving 9 FPS in Support Assist test while the iGPU was throwing out 20 FPS. Dell tech support told me there was nothing wrong with my system per their software logging. I had to point out that the discrete graphics was way behind compared to integrated graphics, and that I have a lot of experience designing chips, and I can at that moment make a detour on my morning commute to their Silicon Valley HQ before the person on the line quickly agreed to send a technician over to my place to take a look at the system.
Turns out the heat sink wasn't pasted correctly on the GPU. Which is what @Anonymous and @r72019 alluded to in their posts today. Do let us know how the temps are. Or if you're comfortable handling hardware, repaste it. Just beware the electrostatic monster. It's going rampant in some parts of the US now due to very dry weather.
RedGreenApple
24 Posts
0
December 28th, 2019 07:00
Here's the temp during Cinebench run. It goes between 80 to 90 C (see screen). I think that's normal for a 9900. I am fine re-pasting the CPU, but I think Dell should be accountable for selling me a non functional unit. This is actually my first time buying a pre-assembled PC. I can't believe the way support team works at Dell.
r72019
6 Professor
•
5.3K Posts
0
December 28th, 2019 08:00
Cinebench runs for such a short period of time, 90 is really high for a cinebench single run. It should start throttling somewhere around that temperature say +/- 5* (not sure exactly) with a critical system shutdown around 105 +/- 5*.
You can try running something more stressful in conjunction with msi afterburner and check for downspikes in the msi gpu/cpu graphs.
I suspect there is an issue with the paste job or aio cooler installation.
My cpu with 5ghz stock boost doesn't even reach 60 on a cinbench first run, and idles under 30 with watercooling, as an aside.