Discussion Cinebench Results for 2 Intel and 2 AMD Hex Core Systems at equivalent clock speed of roughly 4Ghz

[Golgatha]

I've been busy doing some upgrades this week and wanted to get an idea of the performance differences of my hex core systems. Each system runs at around 4Ghz plus or minus 100Mhz at full load, and the 5820k is my old workhorse that's overclocked to 4Ghz. Also, I'm using 32GB (4x8GB) RAM in each system, which has XMP profiles set for max performance on each platform. The only outlier here being 3600Mhz RAM for the Ryzen 5 3600 being slightly above spec for the CPU. Overall, not a lot of progress in 5 years time to be honest, with the exception of the 3600, which kinda destroys everything else. The Ryzen 5 3600 is replacing the 5820k. I'm pretty satisfied with 35% multi-threaded and 50% single-threaded performance gains at the same clockspeed; especially so since I picked up the processor for $155. I also like being off of Intel's DMI 2.0 architecture and getting to realize the full potential of my Samsung 950 Pro NVMe drive. It benchmarks at it's rated specs now, which is nice.

Cinebench R20 Results

Ryzen 5 3600, RAM at 3600Mhz
3651
485

Ryzen 5 1600 (AF Rev, so basically a 2600 in disguise), RAM at 2666Mhz
2687
361

Intel 9400F, RAM at 2666Mhz
2267
398

Intel 5820k@4.0Ghz, RAM at 2666Mhz
2705
326


Bonus: Normalized results to 4.0Ghz for some older systems

Intel Xeon X3470, Lynnfield 45nm, 4 cores/4 threads
1204 (3.2Ghz normalized to 4.0Ghz)
252 (3.3Ghz normalized to 4.0Ghz)

Intel i7 8700k, RAM at 3000Mhz, Coffee Lake 14nm, 6 cores/6 threads
3163 (4.9Ghz normalized to 4.0Ghz)
412 (4.9Ghz normalized to 4.0Ghz)

 

[blckgrffn]

Overall, not a lot of progress in 5 years time to be honest, with the exception of the 3600, which kinda destroys everything else. The Ryzen 5 3600 is replacing the 5820k. I'm pretty satisfied with 35% multi-threaded and 50% single-threaded performance gains at the same clockspeed; especially so since I picked up the processor for $155.

Cinebench R20 Results

Ryzen 5 3600, RAM at 3600Mhz
3651
485

Intel 5820k@4.0Ghz, RAM at 2666Mhz
2705
326

I was a little salty about not getting more cores and threads replacing my 3930k @ 4.2 Ghz with the R5 3600 after 7 years.

But the addition of NVME and the vastly superior CPU performance helped me get over that really quickly

I appreciate you sharing the numbers, I didn't do a good job of measuring the difference like you did.

Gaming performance was so much superior I just spent my time gaming instead of running benchmarks

 

[Golgatha]

I was a little salty about not getting more cores and threads replacing my 3930k @ 4.2 Ghz with the R5 3600 after 7 years.

But the addition of NVME and the vastly superior CPU performance helped me get over that really quickly

I appreciate you sharing the numbers, I didn't do a good job of measuring the difference like you did.

Gaming performance was so much superior I just spent my time gaming instead of running benchmarks

My longer term game is to put a Ryzen 4000 series 8 or 12 core in my main rig, move the 3600 to the computer with the 1600AF in it, and sell the 1600AF eventually.

Side story - My Dad is getting the hand me down upgrade of the 5820k. He's currently on my old i7 920@4.0Ghz. Everyone wins in this scenario.

 

[blckgrffn]

My longer term game is to put a Ryzen 4000 series 8 or 12 core in my main rig, move the 3600 to the computer with the 1600AF in it, and sell the 1600AF eventually.

Side story - My Dad is getting the hand me down upgrade of the 5820k. He's currently on my old i7 920@4.0Ghz. Everyone wins in this scenario.

My Dad is currently rocking a 5820k system I built for him That certainly is winning!

I share your optimism on Ryzen 4000 series, although I may hold out and see if I can make my next upgrade a DDR5 based system. Probably not though

You must build solid rigs to be getting such longevity out of them

 

[VirtualLarry]

and sell the 1600AF eventually.

PM me when you do. I've got loads of AM4 mobos (decent ones, but older) that could use a CPU. IF I have money at that time, of course.

 

[epsilon84]

I was a little salty about not getting more cores and threads replacing my 3930k @ 4.2 Ghz with the R5 3600 after 7 years.

But the addition of NVME and the vastly superior CPU performance helped me get over that really quickly

I appreciate you sharing the numbers, I didn't do a good job of measuring the difference like you did.

Gaming performance was so much superior I just spent my time gaming instead of running benchmarks

Haha well if you don't actually need more than 6C/12T then why get a higher end CPU?

I still have my 8700K as my main gaming PC but 'sidegraded' to a R5 3600 for my ITX lounge PC and don't regret it one bit.

To the OP, I'm happy to throw some 8700K numbers in if for you're interested. I'm pretty sure the 3600 will smoke it at locked 4GHz, IIRC my 3600 isn't that far off my 5GHz 8700K for things like Cinebench!

 

[Shmee]

Cool comparison! Nice work.

 

[Golgatha]

Haha well if you don't actually need more than 6C/12T then why get a higher end CPU?

I still have my 8700K as my main gaming PC but 'sidegraded' to a R5 3600 for my ITX lounge PC and don't regret it one bit.

To the OP, I'm happy to throw some 8700K numbers in if for you're interested. I'm pretty sure the 3600 will smoke it at locked 4GHz, IIRC my 3600 isn't that far off my 5GHz 8700K for things like Cinebench!

A benchmark at 4.0Ghz would be nice, just to see the IPC gains between generations. Or do it at a known frequency and I'll just normalize the numbers to 4.0Ghz. Thanks!

 

[Shmee]

Another good sample to add would be a 4GHz Westemere on X58.

 

[Golgatha]

Another good sample to add would be a 4GHz Westemere on X58.

I actually should have access to an i7 920 capable of 4.0Ghz this weekend. I'll see what I can run. Performance wise it should be nearly identical to Westemere.

 

[ondma]

Haha well if you don't actually need more than 6C/12T then why get a higher end CPU?

I still have my 8700K as my main gaming PC but 'sidegraded' to a R5 3600 for my ITX lounge PC and don't regret it one bit.

To the OP, I'm happy to throw some 8700K numbers in if for you're interested. I'm pretty sure the 3600 will smoke it at locked 4GHz, IIRC my 3600 isn't that far off my 5GHz 8700K for things like Cinebench!

My overclocked 8700k gives ~3875 in CB R20 at 4.9ghz. That is with xmp enabled (3000 mhz) for the ram, but no special tuning on timings. That is with an old intstall of windows and no real attempt to minimize background tasks. Normalized to 4.0 that is about 3165. R20 is pretty much a best case for AMD though.
Other than gaming, I only use my PC for light duty, so I dont really care that the 3600 is ahead in most productivity benchmarks. I use R20 mainly to see if performance is stable over time, usually check it after each update. Havent really seen any measurable performance loss since I started monitoring about a year ago.

 

[Golgatha]

My overclocked 8700k gives ~3875 in CB R20 at 4.9ghz. That is with xmp enabled (3000 mhz) for the ram, but no special tuning on timings. That is with an old intstall of windows and no real attempt to minimize background tasks. Normalized to 4.0 that is about 3165. R20 is pretty much a best case for AMD though.
Other than gaming, I only use my PC for light duty, so I dont really care that the 3600 is ahead in most productivity benchmarks. I use R20 mainly to see if performance is stable over time, usually check it after each update. Havent really seen any measurable performance loss since I started monitoring about a year ago.

Would you mind to run single thread benchmark and give me this number as well? I'm really interested in IPC gains across generations and the single threaded benchmark seems to at least approximate this pretty well.

 

[ondma]

Would you mind to run single thread benchmark and give me this number as well? I'm really interested in IPC gains across generations and the single threaded benchmark seems to at least approximate this pretty well.

Single thread is around 505 at 4.9 ghz and ram at 3000.
I did get it up to 520 with max overclock of both cpu and ram, but it was not stable under those conditions in multithreaded.

 

[Golgatha]

Single thread is around 505 at 4.9 ghz and ram at 3000.
I did get it up to 520 with max overclock of both cpu and ram, but it was not stable under those conditions in multithreaded.

Thanks for running the benches. Updated the main thread.

 

[CropDuster]

I posted these in the big Cinebech 20 thread a while back for reference

5820k @ 4.4Ghz
MT: 2952
ST: 400

X5650 @ 3.8Ghz
MT: 1801
ST: 235

 

[Golgatha]

I posted these in the big Cinebech 20 thread a while back for reference

5820k @ 4.4Ghz
MT: 2952
ST: 400

X5650 @ 3.8Ghz
MT: 1801
ST: 235

That's good info. The margin of error is pretty small doing normalizations for clockspeed. Your 5820k would normalize to 2684 and I'm directly measuring 2705 on my system. That's pretty darn close!

 

[Shmee]

Haswell E also can have its cache/uncore OCed, so that could also help make a difference on top of the core frequency.

 

[eek2121]

Single thread is around 505 at 4.9 ghz and ram at 3000.
I did get it up to 520 with max overclock of both cpu and ram, but it was not stable under those conditions in multithreaded.

My how times have changed. My 3900X gets 523 at stock (4300 Mhz “effective” clock) and 551 if I push the clocks using the EDC bug.

 

[potato masher]

How many months does it take for a 1600AF to pay for iself in A/C bill savings after replacing a 4ghz x58 system...

 

[potato masher]

Two months.

 

[Golgatha]

Two months.

Just did the math. Assuming $0.15 per Kwh, running 10 hours a day, it would pay for itself in 567 days at my cost of $85. Due to AC bill savings, it would be less time than that even.

 

[jihe]

X58 was such a beast.

 

[scannall]

X58 was such a beast.

Might be interesting to throw a 6 core Thuban into the mix.

 

[Shmee]

Thubans, such as the 1100T, were pretty decent, but the lack of SMT will not be apples to apples comparison I feel unless we also test westmere and other Intel hexcores with HT off. Perhaps best to do a test run of each 6 core with both SMT/HT on and off, if present on chip. Would love to see the results of various 6T and 12T chips at certain clocks.

 

[DrMrLordX]

Thuban would struggle in CBR20 due to lack of AVX/AVX2 compared to modern chips.