Intel i7-9800x mini review

[custom90gt]

So I figured since there is absolutely no info out there on the 9800x I've decided I will do my own mini review here on the forum. I am far from a bencher/reviewer so you'll have to bear with me. If you have a test you would like me to run, let me know. I plan on doing this in two parts. The first is totally stock, the second overclocked. This will be a time permitting thing for me so it may take me a couple to few days to get it all done.

System Specs:
CPU - Intel i7-9800x
MB - Asrock X299 Professional Gaming i9 with bios 2.0
Memory - 4x8GB G.Skill 3200mhz CL 14 running at stock XMP profile
GPU - EVGA Ultra Gaming RTX 2080 Ti @ stock
SSD - ADATA SX8200 960GB NVMe
PSU - Corsair HX1000i
CPU Cooler - Corsair H115i PRO RGB AIO

Stock results:
Time Spy - 11986
Time Spy Extreme - 5649
Fire Strike - 23266
Fire Strike Extreme - 14501
Fire Strike Ultra - 7641
Sky Diver - 53354
Cloud Gate - 47983
Ice Storm - 182025
Ice Storm Unlimited - 191299
3DMark 11 Performance - 29722
3DMark 11 Extreme - 15504
Unigine Heaven Extreme - 199.4 FPS, 5023
RealBench - Image Editing 183,232, Encoding 169,163, OpenCL 216,931, Heavy Multitasking 153,380, System Score 180,676
CPU-Z - 476.2 single, 4690.5 multi
Cinebench R15 - 1692, 171 single
Corona - 2:00
7-Zip - 48721 KB/s Compressing, 796651 KB/s Decompressing, 63263MIPS Total Rating
PCMark 10 - 6500
PCMark 8 Home Accelerated - 4973
PCMark 8 Creative Accelerated - 6914
PCMark 8 Work Accelerated - 5914
AIDA64 cache -
AIDA64 CPU benchmarks - Memory Read 82666 MB/s, Memory Write 58854 MB/s, Memory Copy 69505 MB/s, Memory Latency 87.2 ns, CPU Queen 87816, CPU PhotoWorxx 43366 MPixel/s, CPU ZLib 750.4 MB/s, CPU AES 37311 MB/s, CPU Hash 9255 MB/s, FPU VP8 7915, FPU Julia 90424, FPU Mandel 51222, FPU SinJulia 9861, FP32 Ray-Trace 23031 KRay/s, FP64 Ray-Trace 12915 KRay/s
Y-Cruncher 25M (sec) - 7.389
Y-Cruncher 1B (sec) - 35.884
Super Pi 1M (sec) - 9.095
Super Pi 32M (min:sec) - 08:20.3
Prime95 64k Bench (iter/sec) - 45392.88
Prime95 2048k Bench (iter/sec) - 648.17


Overclocking:
quite frankly not great on this CPU. I'm at least semi-stable at [email protected] with a -3AVX2 and -4AVX3 offset. Mesh is set to [email protected] and Ram is at 3733MHz for all of the tests except for AIDA64 Cache (3800MHz). The 2080ti is still at stock speeds though.

Overclocked results:
Time Spy - 13323
Time Spy Extreme - 6245
Fire Strike - 26140
Fire Strike Extreme - 15909
Fire Strike Ultra - 8282
Sky Diver - 63049
Cloud Gate - 58419
Ice Storm - 216429
Ice Storm Unlimited - 224079
3DMark 11 Performance - 33004
3DMark 11 Extreme - 16189
Unigine Heaven Extreme - 210.6 FPS, 5304
RealBench - Image Editing 217,900, Encoding 201,672, OpenCL 229,723, Heavy Multitasking 190,749, System Score 210,011
CPU-Z - 546.9 single, 5249.5 multi
Cinebench R15 - 1985, 200 single
Corona - 1:41
7-Zip - 57764 KB/s Compressing, 913586 KB/s Decompressing, 73629MIPS Total Rating
PCMark 10 - 7698
PCMark 8 Home Accelerated - 5492
PCMark 8 Creative Accelerated - 7718
PCMark 8 Work Accelerated - 6858
AIDA64 cache -

AIDA64 CPU benchmarks - Memory Read 105700 MB/s, Memory Write 75389 MB/s, Memory Copy 90955 MB/s, Memory Latency 62.5 ns, CPU Queen 98677, CPU PhotoWorxx 55795 MPixel/s, CPU ZLib 838.9 MB/s, CPU AES 41886 MB/s, CPU Hash 10415 MB/s, FPU VP8 9012, FPU Julia 105544, FPU Mandel 59798, FPU SinJulia 11061, FP32 Ray-Trace 27171 KRay/s, FP64 Ray-Trace 15255 KRay/s
Y-Cruncher 25M (sec) - 6.849
Y-Cruncher 1B (sec) - 32.787
Super Pi 1M (sec) - 7.854
Super Pi 32M (min:sec) - 07:00.4
Prime95 64k Bench (iter/sec) - 47247.43
Prime95 2048k Bench (iter/sec) - 807.73


Conclusion:
I've made some easier to read tables (at least easier for me, lol). I think it sums up the overclocking results fairly nicely.

 

[custom90gt]

Reserved for later...

 

[custom90gt]

Sorry I haven't made much headway. I'm struggling with this 9800x. Under stock settings and a non-AVX prime95 load I am hitting ~80C on my AIO. I realize an AIO isn't optimal, but I would sure like to hit more than 4.1GHz (stock all-core) on this considering I could do 4.3GHz on a 12 core at similar temps (although that was delidded)...
I've tried 5.0ghz (jump on in kinda thing) and it won't even run cinebench at 1.3v... I should have returned this chip I think, lol.

 

[Woomack]

I think you expect too much if you want to run Skylake-X on water at 5GHz ... regardless of voltage as too low will cause instability and too high will cause overheating.

 

[custom90gt]

I think you expect too much if you want to run Skylake-X on water at 5GHz ... regardless of voltage as too low will cause instability and too high will cause overheating.

I was hoping the 9800x would be similar enough to the 9900k in terms of clocks. I did get the temps down a few degrees with a remount and setting pump speed to max instead of quiet. I will see where I end up at 1.2ish volts OC wise.

Also true to form, my cores are around 12C different in temps. Sad Intel couldn't get the STIM right on these CPUs.

 

[mackerel]

Could I request a Prime95 29.4b8 bench?

Download from https://www.mersenne.org/download/
When starting for first time, hit the stress test option then cancel that. In menu is benchmark screen, please do two runs. Uncheck "use HT" on both. First set min/max FFT size to 64k. Then same again at 2048k. Setting min/max to the same means it just does that one. If possible please let me know what clock it is running during those tests.

Would also be interested in y-cruncher 25m and 1b results (as used in hwbot, don't worry about 10b as that needs a ton of ram). Note this does use AVX-512 so it may really punish any cooling system.

 

[Woomack]

9800X is about the same as 7820X ... 100MHz higher clock, more cache and 25W higher TDP. Chip itself is not much different. 9900K is on totally different architecture ... and is getting really hot too but is easier to OC.

 

[custom90gt]

Sure I can try to run those tonight after work. Stock the 9800x uses a -3 AVX3 offset (-2 AVX2), and actually runs cooler with AVX prime95 loads.

- - - Auto-Merged Double Post - - -

9800X is about the same as 7820X ... 100MHz higher clock, more cache and 25W higher TDP. Chip itself is not much different. 9900K is on totally different architecture ... and is getting really hot too but is easier to OC.

Supposedly the 9th gen x299 cpus are 14++ as well. Also we know TDP means nothing in Intel world...

 

[Woomack]

Sure I can try to run those tonight after work. Stock the 9800x uses a -3 AVX3 offset (-2 AVX2), and actually runs cooler with AVX prime95 loads.

- - - Auto-Merged Double Post - - -



Supposedly the 9th gen x299 cpus are 14++ as well.

This is only manufacture process. It doesn't change much. Just people make a noise about all these ++ like it was changing whole architecture so much. Tests of Skylake-X refresh were around the web some time ago. Based on high core units but in comparison it was like +/- 1% difference.
With new laptop chips is about the same. No difference in performance while they heat up more. There are not many of them and if I'm right then Intel is using older process in some new mobile chips because of additional problems.

 

[custom90gt]

This is only manufacture process. It doesn't change much. Just people make a noise about all these ++ like it was changing whole architecture so much. Tests of Skylake-X refresh were around the web some time ago. Based on high core units but in comparison it was like +/- 1% difference.
With new laptop chips is about the same. No difference in performance while they heat up more. There are not many of them and if I'm right then Intel is using older process in some new mobile chips because of additional problems.

But the manufacturing process is what improves the power efficiency of the cpu right? That should allow for more OC headroom.

 

[mackerel]

Skylake-X was 14+, same as Kaby Lake and one process upgrade from Skylake desktop. We saw a small but not insignificant boost in typical OC from that. Same again when Coffee Lake came out, and if Skylake-X refresh is comparable to Coffee Lake, might an extra 100-200MHz be possible? The reviews all got the top model with all the cores so that is always going to be a power and thermal nightmare to deal with. The lower core count ones may be better able to express their clock potential.

BTW the 2048k FFT test for single worker in theory fits within the L3 of the 9800X but not the 7820X. It should show a good difference against the 8 worker result which will be massively ram bandwidth limited. 64k FFT size will fit easily in cache and I can use it as an indicator for peak AVX2 performance.

 

[EarthDog]

Side note... differences in core temps aren't always due to the TIM. Some naturally run warmer or colder due to their location to other things on the die.

 

[custom90gt]

Side note... differences in core temps aren't always due to the TIM. Some naturally run warmer or colder due to their location to other things on the die.

Good point. I had similar core temp differences with my 7920x that went down to ~3-4C with the delid, but I may have just been lucky.

 

[Woomack]

But the manufacturing process is what improves the power efficiency of the cpu right? That should allow for more OC headroom.

It's not a rule and usually it doesn't work like that as in the chip are additional changes. Let's say we expect that but all depends on the chip. In this case, OC can be limited by more cache which is heating up too and because of higher capacity, max frequency can be limited too. I can't confirm max cache clock as I had no chance to overclock these CPUs.

 

[custom90gt]

It's not a rule and usually it doesn't work like that as in the chip are additional changes. Let's say we expect that but all depends on the chip. In this case, OC can be limited by more cache which is heating up too and because of higher capacity, max frequency can be limited too. I can't confirm max cache clock as I had no chance to overclock these CPUs.

Well come on over! I mean Colorado is just a hop skip and a jump from Poland right?

On a side, I'm at 4.6ghz with 1.225v so far 30 min prime95 stable (non-avx right now). Will let this run Prime while I go to work...

 

[Woomack]

No problem if you send me a plane ticket

My 7900X could make about 4.7GHz 1.2V but it was at 90°C+ and 1.25V was about max on water without throttling. I actually wasn't expecting that but 9900K at 5GHz ~1.3V on the same cooling reaches 100°C under full load. So it's maybe not as hot as Skylake-X but also not so far for 24/7 usage. 8086K on the same cooling has about 15°C less without delidding.

 

[custom90gt]

Hah sadly I may come up short on that plane ticket fund. My hope is that when I go home for the day P95 will still be running and I will have something I can run some benches with. It may not be as high of an OC as I can get with that cpu, but at least it will give me some idea on how it runs (will OC mesh and memory as well). I was hoping for ~4.8 to 5.0 24/7 since I dumped the 7920x to get the 9800x. I didn't really need 12 cores and could use the extra core speed. Sadly going from 12 cores @ 4.4ghz 24/7 to 8 @ 4.6ghz isn't really what I had in mind.

 

[Woomack]

I pretty much told you that in other threads. I was checking new series and decided I will get the 9900k instead, and I said that too. On the other hand, 9900K is also not so special on ambient cooling. My chip can make 5GHz ~1.288V but it hits 100°C in Prime95 on this voltage and quite good water cooling. It's like max clock with reasonable temps is stock. IMC is clearly worse than in my 8086K and I mean ~4500-4600 max when 8086K passed 5000. Let's say it's a 50/50 success as I made some nice tests on dry ice but it's not so good at higher clock for 24/7 or even daily benching. I will try to check if I can improve cooling a bit but I don't expect much.
9800X is still a nice chip and you could keep the motherboard what is also good as this ASRock is a really nice mobo. Except CPU clock there are other things on this platform which can be good like at least quad channel memory.
Good luck with tests.

 

[custom90gt]

Yeah [email protected] was a no go and crashed after an hour or so. Looks like this will be my most disappointing cpu that I've had in a long long time. I'm thinking about returning it and just taking the restocking fee as a lesson. Currently running benches at [email protected]. I'll try to get some of those posted up tonight. I was hopeful about this little cpu, but at least this one seems to be worse than the average 7820x...

 

[custom90gt]

Updated some tests. Here is the Prime95 testing ran at 4.3GHz (4.6GHz with a -3 AVX offset):
View attachment 202836


I will try to re-run the prime95 test at stock, add the results to the tables above, and then create some nicer excel tables. Sadly probably will not happen tomorrow though...