Comparison of my OC'd 1800x-5960x and 6700k Aida64 Benchmarks

[guskline]

Now that I have my Ryzen 1800x rig up and running I thought I would run my Aida64 Benchmark suite on all three machines so posters could see the comparisons.

All three machines are custom water cooled and OC'd. All settings are my stable settings after numerous benchmark testings such as RealBench etc. They have different ssd and video combos but all are running Win10-64 OS. Full specs in my sig below.

Specs for 1800x
OC'd to 4.0 Ghz
16G DDR4 (2x8) running at 2400

Specs for 5960x
OC'd to 4.4 Ghz
16G DDR4 (2x8) running at 2133

Specs for 6700k
OC'd to 4.6 Ghz
16G DDR4 (2x8) running at 2933

I hope this comparison of Benchmark numbers gives some idea of where the Ryzen stands.

 

[Zstream]

No, it doesn't for me. I hate when people overclock the crap out of everything and start to do comparisons. The MAJORITY of people don't overclock.

 

[Hail The Brain Slug]

No, it doesn't for me. I hate when people overclock the crap out of everything and start to do comparisons. The MAJORITY of people don't overclock.

Good for you, go look at one of dozens of reviews where they compare stock to stock.

I haven't seen a single person with a 5960x that doesn't have it overclocked. When you spend $1000 on an extreme edition that is likely to overclock 40-50%, you'd be an idiot not to. It's nice and useful to see stable overclock vs stable overclock when we're comparing top end, unlocked SKUs.

I hope you run more benches, guskline, and expand that chart. I'd be interested to see Cinebenches, Realbench, prime95 bench, and anything else you've got time for.

 

[Hitman928]

I think the memory speeds for the 1800x and 5960x are flipped in the chart. Either that or your sig and post have them flipped.

 

[DisEnchantment]

Specs for 1800x
OC'd to 4.0 Ghz
16G DDR4 (2x8) running at 2400


I hope this comparison of Benchmark numbers gives some idea of where the Ryzen stands.

I think the memory related tests could do some help with faster RAM. I have only stock and under-volted 1700X but with 3200 I get close to 49K in Memory Read/Write and latency in very low 80s. at CL16 (CL15 rounded up to 16)

Someone was running 3600Mhz at CL14 but posted only game benchmarks. Should very interesting to see.

 

[richierich1212]

I'll try and run some Aida64 Benches this weekend with my 3200 Memory

 

[guskline]

I think the memory speeds for the 1800x and 5960x are flipped in the chart. Either that or your sig and post have them flipped.

You are correct Hitman928. I corrected the master which is uplinked via tiny pics. 1800x is running 2400 speed, 5960x is 2133 and 6700k is 2933.

 

[DisEnchantment]

I ran some Benches of of my own, also I OC the 1700X to 3.9 GHz @ 1.325 V. RAM @3200 CL15/16

Latency

CPU Queen : 99778
CPU PhotoWorxx : 24649
CPU ZLib : 830.7 MB/s
CPU AES : 79391 MB/s
CPU Hash : 26763 MB/s
FPU VP8 : 7431
FPU Julia : 37425
FPU Mandel : 20077
FPU SinJulia : 14644
FP32 Ray-Trace : 7476 KRay/s
FP64 Ray-Trace : 4073 KRay/s

Cinebench R15 MT : 1689

 

[guskline]

DisEnchantment, THANK YOU!
WOW is your CPU queen score high. Holy Moly!

Is see what you mean by faster memory scores.

All in all it appears your 1700x at 3.9 is really fast.

 

[guskline]

I have some updated benchmarks for my 1800x;5960x and 6700k at their respective OC speeds, 4.0/4.4 and 4.6.

First as to Cinebench R15 the multi-cpu scores for 1800x/5960x and 6700k are:
1735/1771 and 1011
As to RealBench 2.43,
the 1800x

As to the 5960x:

Finally as to 6700k

 

[DisEnchantment]

DisEnchantment, THANK YOU!
WOW is your CPU queen score high. Holy Moly!

Is see what you mean by faster memory scores.

All in all it appears your 1700x at 3.9 is really fast.

Or could be the sleep bug?
Now I am back to stock and undervolt.

 

[guskline]

Good for you, go look at one of dozens of reviews where they compare stock to stock.

I haven't seen a single person with a 5960x that doesn't have it overclocked. When you spend $1000 on an extreme edition that is likely to overclock 40-50%, you'd be an idiot not to. It's nice and useful to see stable overclock vs stable overclock when we're comparing top end, unlocked SKUs.

I hope you run more benches, guskline, and expand that chart. I'd be interested to see Cinebenches, Realbench, prime95 bench, and anything else you've got time for.

Thanks. I'll keep working on benchies!
My MSI X370 mb is running the 1.1 BIOS. All I did was set the cpu speed to 4000 and the DRAM to 2400. I have Gskill 3200 CL16 ram so I'll have to eait for BIOS updates before I can clock it faster.

 

[guskline]

Or could be the sleep bug?
Now I am back to stock and undervolt.

What mb are you using?

 

[YBS1]

Uh, gusk.....Refresh my memory, didn't you have a 3930K as well that was replaced by the 1800X? If so what's your overall impression of the 1800X@4GHz versus the 3930K in the mid-4GHz range?

 

[guskline]

Uh, gusk.....Refresh my memory, didn't you have a 3930K as well that was replaced by the 1800X? If so what's your overall impression of the 1800X@4GHz versus the 3930K in the mid-4GHz range?

Actually the 3930k was replaced by the 5960x. It's been sometime since I owned the 3930k. It was a great chip and I actually had it clocked as high as 4.6Ghz.

The Ryzen 7 1800x is truly in it's infancy but is turning out to be a fine chip. The BIOS needs refinement to support higher memory speeds. I'm now stable at DDR4-2666 speed but I should be higher.

My 3 rigs are in my sig below.

 

[Carfax83]

I ran some Benches of of my own, also I OC the 1700X to 3.9 GHz @ 1.325 V. RAM @3200 CL15/16

Latency

CPU Queen : 99778
CPU PhotoWorxx : 24649
CPU ZLib : 830.7 MB/s
CPU AES : 79391 MB/s
CPU Hash : 26763 MB/s
FPU VP8 : 7431
FPU Julia : 37425
FPU Mandel : 20077
FPU SinJulia : 14644
FP32 Ray-Trace : 7476 KRay/s
FP64 Ray-Trace : 4073 KRay/s

Cinebench R15 MT : 1689

Holy crap! Unless Aida64 Extreme is bugged with Ryzen, then your memory latency scores are insanely high, which isn't good since lower is better. Here's my memory scores for my 6900K at 4.3ghz using DDR4 3200 CL14:

 

[Carfax83]

I think CPU Photoworxx is the best CPU benchmark, because it actually uses modern instruction sets like AVX and AVX2. The FPU benchmarks like FPU VP8 and FPU Julia use older instruction sets like SSE2 and SSSE3. This explains why guskline's 5960x and 6700K have such a large lead over his Ryzen CPU in Photoworxx, as the Intel CPUs have much higher AVX and AVX2 throughput. Here's the CPU Photorworxx benchmark for my 6900K @ 4.3ghz:

 

[guskline]

I just was able to update my BIOS on the MSI X370 Titanium from 1.1 to 1.31 which resulted in being able to run my ram at 2933 (haven't tried 3200 yet).

Here are changed Aida64 scores:

 

[guskline]

Memory scores went up and latency came down but is still much higher than my Intel rigs.

 

[Carfax83]

Yeah I was going to say, those latency scores for Ryzen are terrible. There is no way they can be accurate if you ask me, because there is no way high speed DDR4 3200 could produce such a result. I mean, that's as bad as the P4 was

 

[guskline]

I know an error message comes up on AIDA (I have the licensed version) that warns that the software may not yet be compatible with Ryzen.

 

[SamMaster]

Yeah I was going to say, those latency scores for Ryzen are terrible. There is no way they can be accurate if you ask me, because there is no way high speed DDR4 3200 could produce such a result. I mean, that's as bad as the P4 was

From what I could gather from reading many posts and watching videos (no links, late and tired), the latency is caused from the link between the "modules" on the chip, including communication between caches. That same link also runs at the RAM speed, making Ryzen have more of a speed bump with faster ram versus Intel, especially with games, but runs well in benches that do not use up much L3 cache (no need to use the link).

So, as the MB manufacturers update their firmware and allow faster stable RAM speeds, Ryzen will be faster as well. We might not see such latencies when AMD release a single "module" chip (R5 line will still be two modules apparently, including the four cores) that do not have a link between them.

 

[Carfax83]

From what I could gather from reading many posts and watching videos (no links, late and tired), the latency is caused from the link between the "modules" on the chip, including communication between caches. That same link also runs at the RAM speed, making Ryzen have more of a speed bump with faster ram versus Intel, especially with games, but runs well in benches that do not use up much L3 cache (no need to use the link).

That seems like a bad design choice then, and it will probably especially hurt them in the HPC market where the working sets are much larger and have to be loaded from system RAM rather than the cache, and the memory module speeds are much lower.

So, as the MB manufacturers update their firmware and allow faster stable RAM speeds, Ryzen will be faster as well. We might not see such latencies when AMD release a single "module" chip (R5 line will still be two modules apparently, including the four cores) that do not have a link between them.

Thing is though, the same thing applies to Intel as well. Intel's mainstream quad core CPUs benefit greatly from faster RAM, much more so than the HEDT parts like Haswell-E and Broadwell-E that have monstrous caches and quad channel memory.

 

[thepaleobiker]

Now that I have my Ryzen 1800x rig up and running I thought I would run my Aida64 Benchmark suite on all three machines so posters could see the comparisons.

All three machines are custom water cooled and OC'd. All settings are my stable settings after numerous benchmark testings such as RealBench etc. They have different ssd and video combos but all are running Win10-64 OS. Full specs in my sig below.

Specs for 1800x
OC'd to 4.0 Ghz
16G DDR4 (2x8) running at 2400

Specs for 5960x
OC'd to 4.4 Ghz
16G DDR4 (2x8) running at 2133

Specs for 6700k
OC'd to 4.6 Ghz
16G DDR4 (2x8) running at 2933

I hope this comparison of Benchmark numbers gives some idea of where the Ryzen stands.

Not sure if you are using the recent update AIDA64 version with Ryzen support - https://forums.anandtech.com/thread...-amd-ryzen-“summit-ridge”-processors.2502505/

Might change some numbers!

Regards,
Vish

 

[guskline]

VISH, thanks, yes I did update AIDA64 but like a Knucklehead I went to update to the latest BIOS (1.41), experience problems flashing (yes I used M Flash and not windows) and could not recover. Fortunately I still have my basic Asus Prime B350 Plus mb so Out came the MSI X370 Titanium and it is shipped to RMA to hopefully reflash the BIOS.

I rebuilt the machine and since the new EK AM4 EVO waterblock has the threaded screws that mount directly to the mb backplate you can use just about any new mb and it will work.

I decided to run the 1800x stock BUT with the latest BIOS update just released for the Asus B350 Plus, 5.13, my memory, Gskill DDR4-3200 TridentZ CL16 is running SOLID at 2933 speed using the D.O.C.P. option and selecting 2933 speed. 3200 will not boot.

My Aida64 CPU Queen score stock with the MSI mb was 82300 and is now 85600 with the Asus mb. Part of that might be the base clock for MSI was 99.21 while the Asus appears to be 100. The release notes on the Asus BIOS for 5.13 said improved cpu performance. I was able to boot in at 4Ghz but the MSI is a beefier mb so I'll run the Asus B350 at stock cpu speed with ram at 2933. I still enjoy the OC feature of the 1800x.

Tonight I'll run updated Aida64 scores for the 1800x stock.