Benchmarks vs reviewers at stock

[drakioned]

I replaced my CPU and motherboard today (3700x + B450 -> 5700x + B550) and ran Cinebench to check it was working.

My results at were 12,380 on multicore, temps were reasonable, power consumption at 77W, so it seemed to be working fine.

If I look at review sites:
- techpowerup: 13,474
- cpu-monkey: 14,230
- kitguru: 13,200
- custompc: 13,431
- nanoreview: 14,150

I do notice that CPU-Z benchmarks show comparable to what's on that.

Now, I'm used to having lower benchmarks than average, and I had always attributed it to running benchmarks under "normal" conditions (with Discord, Logitech/Steelseries/whatever, anti-virus, and Steam running in the background) rather than reviewers test conditions.

Does this sound like a reasonable explanation? Obviously I can enable PBO, etc, etc if I wanted higher numbers, but I guess I want to validate that it's working as "expected".

 

[Armageus]

Same ram as the benchmarkers? Same bios/agesa versions?

I'd discount the cpu-monkey and nanoreview results personally, the others you mention are far more well known and reputable and all seem very similar.

There are way too many variables to get an accurate result (especially if you are using an actual existing windows install rather than a clean reproducible setup like professional benchmarkers use)

Did you benchmark before the upgrade, was the score similarly lower than review scores?

 

[drakioned]

Same ram as the benchmarkers? Same bios/agesa versions?

I'd discount the cpu-monkey and nanoreview results personally, the others you mention are far more well known and reputable and all seem very similar.

There are way too many variables to get an accurate result (especially if you are using an actual existing windows install rather than a clean reproducible setup like professional benchmarkers use)

Did you benchmark before the upgrade, was the score similarly lower than review scores?

No, and I figured it would be lower due to an actual existing Windows install… but I guess I’m just looking for some reassurance that nothing is very obviously wrong. I really would prefer not to reinstall Windows to find out.

Not using same RAM (3600mhz RAM) and latest BIOS… so like, there’s definitely causes for variability

 

[kute]

Just curious why wouldn't someone enable PBO? I don't mean that snarkily, I'm new to AMD and have a 7800X3d and the first thing I did was enable it I believe

 

[Tetras]

Just curious why wouldn't someone enable PBO? I don't mean that snarkily, I'm new to AMD and have a 7800X3d and the first thing I did was enable it I believe

This article is one example, but YMMV:

AMD Ryzen 7950X: Impact of Precision Boost Overdrive (PBO) on Thermals and Content Creation Performance

The new AMD Ryzen 7000 Series of processors bring terrific performance across the board, but have been criticized in many reviews due to the fact that they often hit CPU temperatures of 95 Celcius under heavy loads. However, we have found that they only operate at these high temperatures when...

www.pugetsystems.com

There's also:

Because Precision Boost Overdrive enables operation of the processor outside of specifications and in excess of factory settings, use of the feature invalidates the AMD product warranty and may also void warranties offered by the system manufacturer or retailer.

Understanding Precision Boost Overdrive in Three Easy Steps

Precision Boost Overdrive (PBO) is a powerful new feature of the 2nd Gen AMD Ryzen⢠Threadripper⢠CPUs.â1 Much like traditional overclocking, PBO is designed to improve multithreaded performance. But unlike traditional overclocking, Precision Boost Overdrive preserves all the automated...

community.amd.com

PB2 is not overclocking.

 

[drakioned]

Found these on YouTube:
12724 - https://www.youtube.com/watch?v=_QXHgolDDss
12846 - https://www.youtube.com/watch?v=jmHlNiPQ9sQ
13326 - https://www.youtube.com/watch?v=4eStGbTgj0Q

So I assuming I'm still probably within the expected range, especially with the non-clean system.

 

[nlel1975]

I wondered the same thing with my 7950X3D in Cinebench 2024, I get multicore score of around 2000 when review sites get around 2100. My Windows 11 install dates back to October 2021 when I was running a Ryzen 5 2600 and B450 motherboard.

I replaced that with a 5700x in June 2022 and now have a new motherboard, CPU and RAM using the same Windows 11 install. I wasn't too concerned however as I'm getting roughly 95% of the reviewers score.

Don't really want to reinstall Windows 11 as when I tried with the new system it hung at 84% at copying files for install. This was off a bootable DVD, so if I did try to reinstall again I'd probably try to install off a USB stick.

I had to restore my system from backup after I attempted that fresh install and consider myself lucky that I am up and running, solving that 84% hang might have been a nightmare.

 

[eeii]

It’s always worth keeping your chipset drivers up to date. Especially when upgrading cpus using the same socket.

 

[humbug]

Reviews will vary a lot due to tech reviewers using a very wide range of coolers, the scores you get from using different cooler can vary by as much as 10 or 20%

It works on temperature, the better you cool your CPU the higher it will score, i've seen tech reviewers use the AMD box cooler, even where the CPU doesn't come with a box cooler they pull one out of storage, slap it on and call it "real world results" despite using the most expensive coolers they can find when benchmarking Intel.
Its why AMD stopped shipping Box Coolers with every CPU. tech jurnoes is why we can't have nice things...

Ryzen 5800X +360mm CLC

Stock.

CPU-Monkey is the one i usually refer to as they tend to use proper cooling, the 5800X never came with a box cooler.

CPU-Monkey result 15228.

AMD Ryzen 7 5800X Benchmark, Test and specs

The AMD Ryzen 7 5800X was released in Q4/2020 and has 8 cores. The processor can process 16 threads simultaneously and uses a mainboard with the socket AM4 (PGA 1331). In the Geekbench 5 benchmark, the AMD Ryzen 7 5800X achieved a result of 1,728 points (single-core) or 11,465 points (multi-core).

www.cpu-monkey.com

My result 15862
With PBO: 16124

I don't get a huge gain from PBO because my cooling is already pretty good.

All these stock results are true and proper.

You scored 12380 with a 5700X, that to me doesn't look out of the ordinary, it depends on your cooler, and if you have other applications running in the background, as a sort of "reality check" then its certainly not unusual.

PBO

 

[FredFlint]

Reviews will vary a lot due to tech reviewers using a very wide range of coolers, the scores you get from using different cooler can vary by as much as 10 or 20%

When I first tested my 7950X, I used a Wraith Prism that came with my 3900X, it got ~38.6K Cinebench at ~94C. I then installed an ARCTIC Liquid Freezer II 360 AIO and the score was the same. This was a surprise; I expected a better score as it’s a much better cooler but the only thing that changed was the noise level.

 

[humbug]

When I first tested my 7950X, I used a Wraith Prism that came with my 3900X, it got ~38.6K Cinebench at ~94C. I then installed an ARCTIC Liquid Freezer II 360 AIO and the score was the same. This was a surprise; I expected a better score as it’s a much better cooler but the only thing that changed was the noise level.

7000 series is different, they are designed to run full tilt at 95c.

 

[nlel1975]

This was my Cinebench r23 results when running my previous 5700x air cooled with a Vetroo V5 CPU Cooler. With PBO on and off. Higher score PBO on obviously.

I didn't run it with PBO on as the temps where in the 85C range.

 

[drakioned]

So for cooling I'm using a ARCTIC Freezer 36, none of the core tests exceed 62C on any core/package during the test, so I don't think I'm being temperature throttled.

 

[humbug]

So for cooling I'm using a ARCTIC Freezer 36, none of the core tests exceed 62C on any core/package during the test, so I don't think I'm being temperature throttled.

Same cooler i'm using, tuning PBO would make little difference, and if you're running Discord, Steam ecte... those things can have a small impact on scores, like i said your scores look normal.

 

[nlel1975]

So for cooling I'm using a ARCTIC Freezer 36, none of the core tests exceed 62C on any core/package during the test, so I don't think I'm being temperature throttled.

Yeah as you can see from my results above, at stock settings I achieved a multicore score of 13130 with a max temp of 67C with my Vetroo V5 air cooler.

With PBO enabled I got 14302 but temps reached 85C. It wasn't worth the extra heat to run with PBO enabled.

 

[drakioned]

Alright, thanks all!

 

[humbug]

Yeah as you can see from my results above, at stock settings I achieved a multicore score of 13130 with a max temp of 67C with my Vetroo V5 air cooler.

With PBO enabled I got 14302 but temps reached 85C. It wasn't worth the extra heat to run with PBO enabled.

What you could do is use a Curve Optimiser, you will find it in the PBO settings, set it: All Core > Negative > 10, 10 is usually a safe bet but can go as low as 30, the Curve Optimiser is effectively undervolting the CPU, but its more than that, simply undervolting takes the voltage off right across the range, what Curve Optimiser does is monitor the CPU at 1000Hz poling to see where it can take more volts off or where it needs to retain some volts, it does this on the fly during operation, its very effective.
It wont change the scores all that much from just running PBO but it will run cooler.

 

[FredFlint]

7000 series is different, they are designed to run full tilt at 95c.

I think they work the same, boost clocks are limited by power and heat. In my case, the chip was probably maxing the stock(170W) power, so heat had less impact. With PBO on it might have been different but I did not try it (unless it’s on by default?). Anyway, it was surprising.

 

[humbug]

I think they work the same, boost clocks are limited by power and heat. In my case, the chip was probably maxing the stock(170W) power, so heat had less impact. With PBO on it might have been different but I did not try it (unless it’s on by default?). Anyway, it was surprising.

PBO should be off by default tho if Asus i wouldn't trust there is another setting in there that's similar to MCE and turned on by default.

 

[humbug]

Its worth posting this every now and then....

It explains exactly what PBO and Curve Optimisation is, because its not like one would understand normal overclocking and undervolting, this is, or was 2020 and technology has moved on.

Its 10 minutes, its interesting if you have a Ryzen CPU, and if you came over from Intel you should really watch this.