Adobe Lightroom CC 2015.8 AMD Ryzen 7 1700X & 1800X Performance

You do know that if Adobe fixed their TOTAL SHIT multithread performance on theri software, the 7700K will perform like shit... I think it's a bad call to buy a 7700K for work like this.

Its really hard to know what makes the difference considering how different AMD's CPU architecture is from Intel's. It could be dual vs quad channel RAM (although if that was it I would expect even worse performance from the i7 7700K) or it could be Intel-specific software optimization (but then I would expect better 7700K performance). Or maybe it is the difference between the more typical L3 cache on AMD and "Smart Cache" on Intel or even the maturity of Hyperthreading on Intel.

Maybe one of AMD's engineers could figure out what makes the difference, but in the end what really matters is actual real-world performance. It is definitely an interesting question that I would love to know the answer to, though.

For exporting, I would look at the "Export 100 images" which was RAW to JPG. We've done some testing exporting to different formats (TIFF included) and we found that while the total time of course changed, the relative performance between different CPUs doesn't change a whole lot which is why we usually only test exporting to JPG.

As for the 6800K vs 7770K, it is hard to say for certain if you plan to overclock. However, if we assume that you would be able to overclock about the same amount for each CPU you could just take our results at face value since you are really talking about relative performance rather than raw number of seconds. I would suggest looking over out "How much faster is a modern workstation for Lightoom" article as well: https://www.pugetsystems.co... . Since our testing is the same between that article and this one, you could use the i7 3570K system in that article as a baseline to gauge relative performance. Basically, we saw a total export time of 162 seconds with the 3570K, 107.6 with the 7700K and 68.8 with the 6850K. Again assuming you can overclock all the CPUs roughly the same amount, based on the 10 second per image you are seeing with the 3570K you should see about 6.6 seconds with a 7700K and 4.25 seconds with the 6850K.

So I guess the decision you have to make is if the difference between 6.6 seconds with the 7700K and 4.25 seconds with the 6850K when you export a single image is worth giving up overall faster performance (and responsiveness) for everything else in Lightroom.

9 months later I am still scratching my head at this poor performance on a ryzen 1700x with an rx580. its really bad

We aren't planning on testing with Windows 7 at this point since Windows 7 no longer has mainstream support from Microsoft and the existing supply has nearly dried up to the point that we soon won't be able to sell it on our workstations at all. Our current plan is to re-do our Ryzen testing sometime in May (probably) to give AMD time to get all their problems worked out with Microsoft and everyone else. That should be plenty of time for them to get everything fixed.

Sorry, 128gb

It is supposed to help for things in the develop module, but from what I've heard it is a bit lackluster at the moment. Lots of reports that turning off GPU acceleration actually improves performance. I did hear from somewhere that the higher resolution your photos are, the more likely that GPU acceleration will help so that might be the source of the "it is bad" vs "it is awesome" opinions.

RAM speed is something we get asked about semi-regularly, but to be honest we don't have any testing planned to cover it at the moment. The reason is that we have found higher frequency RAM to be much more susceptible to failure so even if higher frequency RAM is faster, we still probably wouldn't offer it. Right now, DDR4-2400 appears to be rock solid and even DDR4-2666 shouldn't be a problem, but going beyond that we feel is a bit of a risk. Of course, if you are building your own system it might not be a big deal since you are probably willing to do some memory diagnostics if you ever have a problem but for us, that isn't really something our typical customer wants to do in the middle of their workday.

As for more RAM vs faster RAM, i would say the most important thing is that you simply have enough. If Lightroom wants to use 18GB, but you only have 16GB that is going to be a big performance hit - much more of a hit than having faster RAM could ever cover. However, if LR needs 18GB and you are choosing between 32GB of higher frequency RAM vs 64GB of safer RAM then it is really just a call between more performance but higher risk of failure versus a bit lower performance but more reliable and safer long-term (less need for upgrades). I think that is really a call that each person has to make for themselves, but I personally would go with the higher capacity.

We generally recommend a GTX 1060 as a starting point on our systems, but if you are in a budget crunch something a bit lower should be fine. I would try to stick to a GTX 1050 or above if you can though.

We've tried to figure out a good method to benchmark exactly that, but couldn't figure out a good way. Especially with brushes, if you move the brush too slow you won't see any difference between hardware. But if you move it too fast, LR lags out and the brush starts skipping around. To make it even more difficult, depending on the hardware the point where it starts skipping around changes. That in itself might be something we could benchmark (maybe), but LR gives almost no feedback for when things like that happen. So we basically would have to do a pixel checksum across a large area to see if things changed in the exact manner we expect. Unfortunately, in the trials I've done this is really inconsistent so it isn't really a reliable way to compare hardware.

I hear you that doing more testing in the development module would be great. Every time I do some LR testing I give another few techniques a try - hopefully one of these LR updates will introduce something that I can use to improve our testing even further.

High performance mode is actually what we default to during benchmarks, but the difference between it and balanced shouldn't be much as long as the system isn't idle for too long.

SMT is something we tested across a bunch of software packages a few years ago and at the time we didn't see any significant differences. A few percent here or there but it was inconsistent whether on or off was better. Honestly, I would just leave it alone and not worry about it - you probably wouldn't notice a difference unless you got out a stopwatch and started timing things.

1:1 previews are full resolution, so 5184x3456 in our case. Smart previews are always 2560px on the longest side I believe, and ours ended up with a final resolution of 2560x1707.

There were some that we covered in our later article: https://www.pugetsystems.co... . Ryzen has seen about a 7% increase in performance since this article was published, but it looks like this is from general optimizations than anything AMD specific since Intel also saw performance gains. Especially with the launch of Skylake-X (which is also covered in the article I linked) Ryzen simply isn't a good choice for Lightroom. Intel is simply able to achieve much higher performance for similar or much lower cost.