Actually, for me the thing I’d like to improve on is my latency. Or is that 100% a function of the interface, and the computer doesn’t help get it down?
To be honest, while the intention of this thread is good, it is pretty useless to compare. To many variables in both software and hardware between different people to do a comparison like this. It needs to be done like the computer sites do when they test cpu performance and compare.
Same OS, same settings, same ssd/hhd, same RAM amount and speed, same BIOS settings or at least similar in case of different motherboards.
Fair enough, but to me that’d be like testing car performance by planting just the engines in the same Escort or something .-)
The thing is - like in the case of 7700K vs. 6700K - CPU to CPU the difference is in benchmarks (similar or same system) merely around 9% give or take.
Based on benchmarks and overall computer site tests, there really seems to be zero sense in upgrading a 6700K to a 7700K.
But, it’s not merely a matter of CPU calculation power; the mobo architecture and memory controller etc. are very important, and this type of blunt test is the only way to effectively showcase how much faster can a new gen system really be - even when based on CPU charts etc.the differences are next to nothing.
So this is already showing very clearly just how much faster the newer ones are.
I have this rule to NEVER upgrade anything unless the update gives me twice or more the speed of the old system, only then it is justifiable to spend the money.
And going through CPU charts and performance tests it has seemed that I would at best be gaining maybe 15-20% going to the most expensive new intel chip, which just makes no sense.
But this shows very clearly just what the gain can be to upgrade.
In fact I’d be happy with these results alone to make up my mind, but I’d LOVE to see some oldskool dual-xeon systems battle it out. How does a mega-core count old Dell or HP workstation fare? a mac pro? and a Threadripper, more ryzen chips???
To me this is priceless, and results are looking to be very clear that with every newer gen system you get a clear boost that isn’t at all obvious based on just component tests and the usual media mark tests.
Right. You have one platform and change one parameter, and then you know the importance of that parameter. But then you go and change a different one. That’s really the better way to go about it because you want to isolate one parameter at a time to understand what the implications are of changing them.
Look at it this way: If we run your test on a maxxed out system that’s expensive and you see the performance you like you might go out and buy that system. And that’s fine. But you don’t really know if you end up overpaying for it simply because you don’t know if all of the new items in that new system are necessary. Perhaps all you need is a new motherboard architecture, perhaps it’s different memory, perhaps it’s more cores, perhaps it’s faster cores… but it might not be all of those things to get the results you’re looking for. But you’ll never know because you have no way of isolating one parameter at a time if the systems are entirely different.
I don’t think you’re reading it right then. Looking at the chart of the ‘classic’ test the 6800k can run 269 plugins at 256 samples buffer. The 7960x can run 511. Now, if I understand correctly the 6700 you have is the same generation as the 6800, except with four cores instead of six. So, since scaling isn’t perfect you should have more than 66% of the performance of the hex-core. So, you should reach about 180-190 plugin instances. (511-190)/190=1.69, or a 169% increase!.. not 20%…
Obviously, whether or not that translates into export numbers is a different issue, but the charts show a tremendous increase over your generation - if my assumptions are correct of course (i.e. your quad core’s performance relative to the 6800k).
At this point it may actually be worth it for you to download and run the DAWbench (non-VI) test and see just what you get.
Both the interface and the CPU matter from what I can tell. Techreport disappointingly ran tests using a USB interface that people said didn’t have great latency performance (Focusrite maybe? I forget…) and you can actually see how it hits a wall in their testing. When I brought this up I was told it was because of the limits of the CPU architecture, but clearly there were both Intel and AMD CPUs that hit a wall, and the wall was exactly the same number. It’s so unrealistic that the bottleneck must have been elsewhere, and the only other common denominator was the interface.
But with a good enough interface / drivers a more powerful CPU should give you better capacity to run at low latency (smaller buffers). I think in general Intel chips perform better at lower latencies, and in general higher clock speed gives you a bigger improvement, or a more linear one at least, compared to more cores as far as this is concerned.
Only thing I can think of at this point in time is if there are further fixed to deal with the Meltdown vulnerability and if that harms Intel CPU performance. It’s possible.
You’re absolutely right but consider this: most people air know do not care how many plugins they can run, they have never heard of DAW bench and all they basically understand about computers is how much they cost and how quickly can they render stuff.
That’s what I mean - stuff like DAW bench and CPU Boss and such, they are hebrew to most people, only make them confused about what is the actual difference.
That’s what I mean by largely academic. It’s very very hard to have an idea how fast might a dual Xeon machine be for instance in Cubase based on numbers of percentage and even plugin running capability.
This kind of test is blunt and unscientific, but what it gives is very easily understandable data on how much time can the guy behind the desk expect to save moving from that Core2 machine to a new i9 or whatever - because to them it’s useless to quote percentages.
All I ever hear is ‘how much faster does it export’ and I can’t honestly but guess given the scale of different systems. Is more cores always better? Is Ryzen as fast as i9? Hard to start explaining about plugin instances.
So far, the results have been clear and sort of expected but also surprising.
Anyhow, how can it hurt to have as many different ways to approach testing system capabilities? I don’t really see why would you be so apprehensive about getting information that is clear and simple, even if not so exact and scientific. More info and insight is always good, no? Well I for one really appreciate the results so far and find them far more revealing than comparing various graphs. I think there’s a lot of people who do the same too.
I’m doing DAW bench as well. And I have a feeling the difference in that vs.7700 is not nearly as huge as with this test where it doubles the speed.
Basically what I’m getting at is that when you give people simple, broad and generic data they end up not perfectly emulating the system in question and if they miss one component or another they get confused and upset because their system doesn’t perform like the one they saw and thought they got. That’s why a bunch of non-specific data can lead to more confusion than clarification.
I got 129 instances in DAW bench with 64K buffer in DAW bench with the 6700K.
And yes, the point is exactly that, not to try to level the playing field by eliminating various changes in systems over the years, but get a general look with ease.
Seems to me results so far are extremely well in line with DAW bench results, but is far easier to test and quicker as well. Plus it’s much more undertandable to non-gearheads.
Still eager to hear from any multi-Xeon machine (older) to chime in.
The way I figure now, it’d cost me about 600;- to very certainly double (or more) the speed of my machine / halve export times.
The most cost-effective solution for me would be i7-8700K with a new mobo and cooler > circa 600.
Going to Ryzen, Threadripper or i9 yields diminishing returns with possibly 2-3 times the costs and not nearly as much more oomph in comparison.
I really don’t see how a Ryzen 7 1700 for example is 2-3 times more expensive than an 8700k, nor do I see how the returns are diminishing given your criteria. Of course, we’re not in the same country (?) so that may make a difference…
I mean the Ryzen 1700 is almost exact same 300;- here as the 8700K but the 8700K is about a third faster at the same money. And Threadripper is over 3 times 8700K price yet only maybe 30+ % faster so the benefit per buck decreases.
At least in audio use it’d seem the 8700K is of pretty much the best bang for buck right now.
Also if you’re willing to pay a grand for a threadripper, you can already get an i9 for the same which blows the threadripper outta water.
But that’s how it’s always been, AMD has always been a bit of an underdog for audio stuff.
If the 8700K is about 300 where you live (Finland I presume) then you’re getting it cheaper than Americans are (!). Here in the US a realistic comparison for me in New York is:
Newegg online / microcenter local store
8700k: 390 / 360
R7 1700: 300 / 250
The Ryzen includes a cooler. When you add the cooling solution to the Intel CPU you’ll be paying at least $30 I bet. So it’s more like;
8700k: 420 / 390
R7 1700: 300 / 250
So if we ignore taxes (doesn’t apply to New Yorkers for Newegg online purchases, but for Microcenter local of course) and the cost of a motherboard, the Intel CPU is actually 40% / 56% more expensive. Looking at the charts we see that the performance of a 1700X is about the same as the 8700k at stock, and the 1700 can of course be overclocked to match that 1700x. Of course, at that point it’s an unfair comparison, because you can overclock the 8700K as well at which point it performs even better. But the question is if it’s still able to perform about 40-50% better. Perhaps it is.
Threadripper is really a different animal completely, and the same goes for the i9 platform. I don’t think anybody should invest in that platform unless they absolutely need all the cores or more importantly all the connectivity. There’s simply no comparison once connectivity is included; the Threadripper is WAY better than the LGA 1151 platform. Just no comparison at all. But of course not everybody needs that connectivity so it’s possibly a moot point at which point we’re back to discussing cores instead.
But again, you’re saying that the i9 blows the Threadripper out of the water. It does… at a price. The discussion whenever alternatives to Intel comes up typically goes exactly this way, and it’s really frustrating. If your argument is valid, then there’s no reason to not buy the top-of-the-line i9 to begin with. Forget about the 8700k. It’s a meaningless purchase.
But if it for some reason does make sense to buy the 8700k then it’s probably as you say because it’s a good value, and at the right amount of money a person wants to spend. So, that same logic applies to comparing Threadripper with i9 chips as well. What is the better value, and how much can you really spend on the CPU? The answer as far as I can see is definitely not that the i9 is the better value, but instead the Threadripper. You actually get more performance per dollar if you pick the right one, compared to the Intel CPUs.
Again, this is all assuming that we can look at the DAWbench benchmarks for an indication of performance between CPUs (and we can) but that the only thing that matters is export times (of which we know nothing, supposedly).
Logic here is which CPU at least doubles the speed of my 6700K - DAW benchmark at higher latency or export speed - and both the 1700X and 8700K would do it at nearly same money (8700K roughly 60-80e more) and looking at DAW benchmarks the 8700K does quite a bit better than the 1700X for the same money.
Plus I could use my 6700K memory and such, which would reduce the overhaul price. Perhaps I could even use my existing cooler too bringing them real close to price. So it really makes no sense to settle for the 1700X.
Threadripper however is indeed triple price plus needs new memory and all, effectively quadrupling the price of overhaul. Yet isn’t that much faster than 8700K especially at higher latency settings. So it isn’t that enticing for price/performance.
Even a 7700 or 6800 would be a nice increase, but the overall price comes close to 8700K already. And CPUs in between…well I’d have to excel chart all the pros and cons, there could be a CPU easily twice as fast as my 6700K is and available at lower than ~600 for the whole overhaul. But at a short exploration the 8700K seems like the cheapest way to surely double my efficiency.
I pretty much never spend more than maybe 400 or so a year in upgrades, and I could possibly use my contacts to source a 8700K at even slightly cheaper, so I’m not that much in a rush.
Basically, when I figure out a combo at about 400:- that can double my speed, then it’s money well spent in my book.
A Threadripper or i9 seems like overkill for sure for my needs.
No, Neither a 8700K or a Ryzen 1700 will double your speed. They are faster yes, but no near double as fast than a 6700K. This is why you cannot compare it like this. Your 6700K should be around the same as my 7700K, but yet in your list it shows 13 seconds slower. Why? Because maybe your OS is more bloated taking up resources, or maybe your drive are slower or maybe this or that.
That’s why a list like this is useless unless you have the same persons test them all with same settings, clean OS install and such.
DAW bench also gives me the sama results, the 8700K being easily twice as fast. My system can run less than half the instances 8700K. 125 against almost 300 for example at low latency, which is well in line with the weakest CPU on the chart, 6800K.
Hard drive speeds don’t matter here and I’ve an nvme drive anyhow, 32GB of 2666MHz mem and such as well. Besides I’ve only reinstalled Windows clean a few weeks back as well as Cubase and made sure my BIOS settings are optimal etc. Furthermore when I might change to a 8700K I’d be using exact same settings and programs as well. So yeah, I would expect my export times to about halve and plugin running capability to about double based on either test, mine and DAW bench, upgrading the motherboard and the CPU two generations higher.
While in games or such I might only expect a 20-30% speed boost if any, the newer generation’s more effective internal processes etc.give much more noticeable boost in number crunching than is obvious.
The very thing this project has taught me is that with every CPU generation the actual speed increase in system intensive audio tasks is clearly bigger than would be obvious mere CPU charts and clock speeds indicate.
The export speeds follow the DAW bench charts very nicely which confirms that you can use DAW bench well to predict overall speed, not just plugin running power etc. Which is exactly what I wanted to find out - what tests are the most effective predicting speed in actual use. The CPU benchmarks for gamers etc.are great but don’t really reflect what happens in AV use that well, and now it’s obvious for my use that I can pretty much trust DAW bench charts for that.
And based on those, yeah, a 8700K system should be about “twice as fast” as 6700K. That’d be only in certain uses but for DAW crunching it seems that way.
Now, it might be a different matter if you built a NEW system around the 6700K using a new chipset motherboard as well - I suspect my system as well might work a bit faster with a 370 mobo instead of the current z170 Asus Maxinus 8, but then again, it makes no sense for me to only swap to a new motherboard really.
I’d be looking at about $320 with trade in of old parts and that I do believe would give me about 29% overall more oomph and about 100% more export speed and DAW crunching power based on all these tests and more.