Seeing higher Ryzen performance than most benchmark sites

D

DuronBurgerMan

[H]ard|Gawd
Joined
Mar 13, 2017
Messages
1,340
Looking at the reviews for the 1600X and 1500X... most still have the 1800X and/or 1700X listed in their benchmarks for sake of comparison.

And it just occurred to me, that I'm seeing consistently higher results than almost all of them. Take PCper, for instance. They see a Cinebench 15 result of 1504 for their Ryzen 1700X at stock clocks. I consistently see 1560+ for the same at stock clocks. And I'm running my RAM at 2400 MHz, right now, because the Agesa 1004 update torpedoed my RAM overclock for whatever reason.

So I'm running slower RAM than their test bench, and still end up with a higher score. They aren't the only ones, either.
PCPer: 1504
Anandtech: 1540
Hexus: 1546
Guru3d: 1527
Gamespot: 1528
Bit Tech: 1547
Hardware Canucks: 1548

PCPer's test is unusually slow compared to the rest. I wonder what they did wrong?

In fact, I could only find one benchmark that was at my system's level:
Techpowerup has the 1700X at 1567

But all of these benchmarks were WAY faster RAM. The only thing is... I did put in 32GB in this machine, so maybe that's the difference. Either way, I thought it was kind of interesting. Cinebench isn't the only benchmark I'm seeing this effect on, either. My build scores a few percentage points above the benchmark field in almost every comparable test (obviously we can't count the GPU/gaming tests unless using a reference/FE 1080 Ti, for obvious reasons).

I wonder why? Just some luck with the parts I bought, maybe?

On a side note, finally somebody benchmarked Ryzen in Terragen 4 (something I use a lot). Ryzen does well here:
https://www.bit-tech.net/hardware/2017/04/11/amd-ryzen-5-1600x-review/4

I expected it to, as I see much faster render times on my machine since upgrading from the 2600k, but never seen a benchmark with it until now.
 
PCPer is 3% different than your system. The rest are closer than that. My 4790K, at stock, in CPU-Z shows single-threaded IPC at 2% higher than the 'benchmark' processor. Luck of the draw? Slightly different settings? 3% is pretty much close to the margin of error.
 
So the reviews online using a synthetic benchmark all show a variation of a few points that are within the same range you got but you find it odd?
 
BlueFireIce said:
So the reviews online using a synthetic benchmark all show a variation of a few points that are within the same range you got but you find it odd?
Click to expand...

No need to be sarcastic. My box should be slower, given the vastly slower RAM. I'm looking to find out what I may have done to enable this machine to run consistently faster than the benchmarks.
 
Last edited:
Algrim said:
PCPer is 3% different than your system. The rest are closer than that. My 4790K, at stock, in CPU-Z shows single-threaded IPC at 2% higher than the 'benchmark' processor. Luck of the draw? Slightly different settings? 3% is pretty much close to the margin of error.
Click to expand...

If it was one benchmark comparison, I'd ignore it, for the reasons you mentioned. But this machine is consistently higher than all of them. Just trying to find out what I did. If it's slightly different settings, what settings were they? I figure it might be useful info for Ryzen folks trying to squeeze out every last drop of performance.
 
DuronBurgerMan said:
If it was one benchmark comparison, I'd ignore it, for the reasons you mentioned. But this machine is consistently higher than all of them. Just trying to find out what I did. If it's slightly different settings, what settings were they? I figure it might be useful info for Ryzen folks trying to squeeze out every last drop of performance.
Click to expand...
Have you overclocked the BCLK? There is a time bug.
 
Romeomium said:
Have you overclocked the BCLK? There is a time bug.
Click to expand...

No. If anything, in CPUID, the bclk is running a hair under 100Mhz. Always reports like 99.8-99.9. Not that it's very significant, though.

Interested in the time bug idea, though. Where'd you hear about that?
 
Last edited:
DuronBurgerMan said:
If it was one benchmark comparison, I'd ignore it, for the reasons you mentioned. But this machine is consistently higher than all of them. Just trying to find out what I did. If it's slightly different settings, what settings were they? I figure it might be useful info for Ryzen folks trying to squeeze out every last drop of performance.
Click to expand...

Different Windows version? Different version number of the benchmark? Different video driver? Different motherboard driver? Different motherboard? I mean, Ryzen is a brand-new architecture from AMD. Experiences are going to be different. It's possible people could reproduce your performance gains. It could also be possible that people may not.

Here's a scary thought. You have a golden system and everyone else who buys a Ryzen system can't get your scores but they bought Ryzen thinking your system was typical. :ROFLMAO: This is why I'm waiting for the bugs to work themselves out before jumping in. Not to mention that my work computers are already 12c/24t and Ryzen is actually a downgrade in my work scenario and my gaming PC does what I want it to (no, I don't do work on my game system).
 
Algrim said:
Different Windows version? Different version number of the benchmark? Different video driver? Different motherboard driver? Different motherboard? I mean, Ryzen is a brand-new architecture from AMD. Experiences are going to be different. It's possible people could reproduce your performance gains. It could also be possible that people may not.
Click to expand...

Checked a lot of those. Windows 10, same Geforce drivers. Same version of the benchmark. Now... motherboards and motherboard drivers could be a thing. That'd be interesting to see benchmarked. I haven't seen a good benchmark between all the major X370 motherboards. So that could very well be the difference. That would be interesting to find out.

Here's a scary thought. You have a golden system and everyone else who buys a Ryzen system can't get your scores but they bought Ryzen thinking your system was typical. :ROFLMAO: This is why I'm waiting for the bugs to work themselves out before jumping in. Not to mention that my work computers are already 12c/24t and Ryzen is actually a downgrade in my work scenario and my gaming PC does what I want it to (no, I don't do work on my game system).
Click to expand...

I doubt it. But you could be right. My 2600k box was on the verge of death. No time to wait. But... I should have forked over the cash and bought a 6900k instead of Ryzen. I made too many compromises with the Ryzen build. Sacrifice gaming FPS. Sacrifice overclockability. Sacrifice a small amount of workstation performance (the 6900k still wins there, overall, if just barely). Yeah, Intel is overpriced. But they are still better, and I should have just gone that route.

But I have this box now, and it ain't bad. Certainly blows the doors off my old rig. So I'll keep trucking with it.
 
DuronBurgerMan said:
No need to be sarcastic. My box should be slower, given the vastly slower RAM. I'm looking to find out what I may have done to enable this machine to run consistently faster than the benchmarks.
Click to expand...

From my understanding CB15 is not a RAM limited bench. After some Googling it seems some people have reported lower scores with higher ram speeds, and those who saw higher scores were in the order of 0.4%. Which is acceptable error range.

There have also been updates to windows and BIOS, not including differences in mobos and ram itself. If you were getting consistent 15% better scores it might stand out, but you are talking 0.7% change from the other tests, and a loss of 0.4% for the last one, you are well in range.
 
Naw, I saw your original thread and I think, for your usage scenario, that the Ryzen pick was better for you. For my work scenario, I have to wait for the workstation 16c/32t chips to come out and then I need to convince my boss to get it for me. :) Considering that both my workstations are responsible for producing video worth several million dollars of revenue per year he usually doesn't second-guess my requests.
 
BlueFireIce said:
After some Googling it seems some people have reported lower scores with higher ram speeds, and those who saw higher scores were in the order of 0.4%.
Click to expand...

Interesting that higher ram speeds could lower it. But if that's true, it would explain the difference in Cinebench.
 
DuronBurgerMan said:
Checked a lot of those. Windows 10, same Geforce drivers. Same version of the benchmark. Now... motherboards and motherboard drivers could be a thing. That'd be interesting to see benchmarked. I haven't seen a good benchmark between all the major X370 motherboards. So that could very well be the difference. That would be interesting to find out.
Click to expand...


Bit-tech has around 4 X370's compared w/ 1x 350, not sure if that would help or not in your quest.
 
CAD4466HK said:
Bit-tech has around 4 X370's compared w/ 1x 350, not sure if that would help or not in your quest.
Click to expand...

Reading that offers the second piece of the puzzle: the Asus X370 Prime is the fastest board in many of those tests, and always one of the fastest. Even faster than the Asus Crosshair board, on stock clocks.

Power consumption is sh*tty though. It's one of the worst for that. Still, I think a combination of that with what BlueFireIce found, plus a little margin of error luck, would explain the difference.
 
DuronBurgerMan said:
Interesting that higher ram speeds could lower it. But if that's true, it would explain the difference in Cinebench.
Click to expand...

The different was points of a percent though, which like this is all in the error range, so it's not that it went down, but rather probably had no effect however the error range of the benchmark makes it look like it did.
 
DuronBurgerMan said:
Reading that offers the second piece of the puzzle: the Asus X370 Prime is the fastest board in many of those tests, and always one of the fastest. Even faster than the Asus Crosshair board, on stock clocks.

Power consumption is sh*tty though. It's one of the worst for that. Still, I think a combination of that with what BlueFireIce found, plus a little margin of error luck, would explain the difference.
Click to expand...

I'm thinking the Crosshair results were from release day benches on Ryzen.
If you look at the screen grabs of the EFI, they say 3/02/17, even though the review is from the 17th, probably immature BIOS and 2933 RAM made it slower.
 
CAD4466HK said:
I'm thinking the Crosshair results were from release day benches on Ryzen.
If you look at the screen grabs of the EFI, they say 3/02/17, even though the review is from the 17th, probably immature BIOS and 2933 RAM made it slower.
Click to expand...

Could be. That's one irritating thing about the AMD updates... hard to make comparisons to benchmarks from even a few weeks ago.
 
DuronBurgerMan said:
Could be. That's one irritating thing about the AMD updates... hard to make comparisons to benchmarks from even a few weeks ago.
Click to expand...

It would be a PITA to refresh the numbers every few days every time a new BIOS or driver comes out, but if your here to serve the community, that is the ONLY course of action for a competent reviewer on a PC hardware site. Brent, Dan & Kyle do. ;)
 
CAD4466HK said:
It would be a PITA to refresh the numbers every few days every time a new BIOS or driver comes out, but if your here to serve the community, that is the ONLY course of action for a competent reviewer on a PC hardware site. Brent, Dan & Kyle do. ;)
Click to expand...

Which makes me wonder why Kyle isn't cursing at AMD right now ;). That's a metric shitton of work.
 
DuronBurgerMan said:
Which makes me wonder why Kyle isn't cursing at AMD right now ;). That's a metric shitton of work.
Click to expand...

Because Kyle does a whole new metric every time he does a new review, the only time I see copy & paste with this site is when power usage graphs are being compared to new [H]ardware being reviewed. Now I don't think Brent updates his power charts unless he's benching all the cards at once, I could be wrong though.
 
I don't know how good the benchmark itself is, but these guys do have distributions of invidual scores posted, so you could possibly see where you land vs. others. You can tell that the distribution is bi-modal, which probably shows those who overclock vs. those who don't.

(edit: tri-modal, not sure what the 3 groups would be)
 
atmartens said:
I don't know how good the benchmark itself is, but these guys do have distributions of invidual scores posted, so you could possibly see where you land vs. others. You can tell that the distribution is bi-modal, which probably shows those who overclock vs. those who don't.

(edit: tri-modal, not sure what the 3 groups would be)
Click to expand...

CPU performed at 90th percentile compared to other 1700Xs. So I guess it is pretty high. I can't take credit... must of done something by accident.
On the flip side of that, my 1080 Ti is performing in the 46th percentile, so that's not so good. Weird.
 
Last edited:
DuronBurgerMan said:
Which makes me wonder why Kyle isn't cursing at AMD right now ;). That's a metric shitton of work.
Click to expand...

probably why the gaming reviews have been so delayed after AMD announced they would be releasing a new AGESA update in april a few weeks ago. but yeah i agree with the way [H] does their reviews(which in my opinion is the only way they should be done) it would of been hell updating the numbers with every AGESA/bios update, lol.


but yeah as far as the original post goes, there's to many variables in place between when the 1700/1800x reviews were done to now to definitively point the finger at one thing for why you're seeing higher numbers. there's been plenty of windows updates, bios updates, AGESA updates and so on since most of those reviews were done.
 
Last edited:
You must log in or register to reply here.
Back
Top