It's because AMD is always the lower quality product. There is always some caveat to using them that means they need to be priced less than Intel. There's your mindset 🙂 Ezpz. I think this is mainly based on inexperience and emperical evidence in minute numbers as well as a range of uneducated assumptions.
Sadly they did it to Zen 1 so I won't be surprised if they do but on a side note I was being mostly sarcastic when I said that.
Of course 😉
Met too.
But sarcasm aside, it did happen...
I think that particular change was good, as the score didn't truly reflect real world performance, and after the change it did, for the most part.
This time however I think Zen 3 is really THAT good, it's not just a measuring error.
The fact that many don't believe it is because Intel has not change architecture for AGES now, and even the changes they did in the last 10 years were minor.
So, a "normalcy" bias was ingrained in people's minds, that it's "not possible" to have 20% plus IPC increase generation to generation, because they (Intel) only gave us 2%, 3%... or absolutely nothing (just higher clocks).
Yes, it IS possible, and more.
In theory even +100% IPC should be possible, by throwing 5-10 times more transistors at the problem... but then you run into the power usage and inability to clock high enough due to excessive energy use.
It's all a fine balancing act between more parallelism inside the core and how much energy that core uses.
And I think AMD really nailed it this time !
this is strange that the r7 5800x (8cores/16Threads) got a score of 6585 in cpu-z bench, when my r9 3900x (Stock no OC) with the same amount of cores/threads in cpu-z give me the same score.... only single core score that is higher with the r7 5800x.....
https://valid.x86.fr/ua5na1https://i.imgur.com/7nCrJdV.jpg
Looks like your RAM speed is letting you down. And from what I can see from your specs, your 3900X is definitely running much faster than mine (250MHz more)... you sure you are not manually overclocking it? What cooler are you using?
I still get mush faster in a test run... and I am just running the "AutoOC" ryzen master mode... and stock Wraith cooler...! And same Hero X570 motherboard as you. I see also you are using slightly newer BIOS (2311 (10/16/2020) vs 2206 (08/13/2020))
https://valid.x86.fr/c9jb89
It was even faster a few months ago... 526/8263
https://valid.x86.fr/ita54n
I'm getting 520-522 on my 3700X with DDR4-3733 CL16 (with tuned timings via @1usmus calculator )
Not even using PBO, CPU itself is 100% stock (as PBO seems to affect mostly just multi-core on my chip, which I rarely use)
That "480" score on a 3900X is clearly having some sort of a bottleneck.
---
Edit: I think that's an all-core fixed-clock overclock.
Which obviously will reduce single-threaded score significantly. 3900X should get 530 or so on default with decent memory.
Temperature.
Zen 2 (and probably Zen 3 too) boost higher if the ambient (and it's own temp) is significantly lower.
I did a test last winter by letting windows open on sub-zero temps outside, CPU booted with just 6 degrees at Windows desktop, ran a benchie as soon as I could and got the highest Cinebench and 3D Mark CPU scores I ever had, over 7% higher than my previous max !
I've ordered my X570 motherboard, DDR 4 3600 Ram and 360mm radiator today in preparation for the launch. Looking forward going from i7-5775c to this new chip. I think I'll be set for the next 3 to 5 years as I'll start with an 8 core and if needed down the road could always move to the 16 core.
Great stuff AMD you did well and happy to support you again as my last AMD CPU was the AMD Phenom II X6 1090T and still running to this day!
I'm getting 520-522 on my 3700X with DDR4-3733 CL16 (with tuned timings via @1usmus calculator )
Not even using PBO, CPU itself is 100% stock (as PBO seems to affect mostly just multi-core on my chip, which I rarely use)
That "480" score on a 3900X is clearly having some sort of a bottleneck.
---
Edit: I think that's an all-core fixed-clock overclock.
Which obviously will reduce single-threaded score significantly. 3900X should get 530 or so on default with decent memory.
Temperature.
Zen 2 (and probably Zen 3 too) boost higher if the ambient (and it's own temp) is significantly lower.
I did a test last winter by letting windows open on sub-zero temps outside, CPU booted with just 6 degrees at Windows desktop, ran a benchie as soon as I could and got the highest Cinebench and 3D Mark CPU scores I ever had, over 7% higher than my previous max !
Im getting the same 480 st with the 2700x so thats a huge bottleneck somewhere if 3900x is getting the same score.
As someone who understands the difficulty of writing multi-threaded software and Amdahl's Law, not everything can, will, or should be multi-threaded. However... there is hardly anything today that does any serious number-crunching and isn't multi-threaded. Sure, on the CPU side of things, games are held back by the slowest thread, but most CPUs these days are plenty fast enough to, for example, give you a good VR experience.
So, although single-threaded performance isn't meaningless, it doesn't really matter anymore, because nothing worth caring about is single-threaded and bottlenecked by modern CPUs. If you care that much about single-threaded performance, save yourself hundreds to thousands of and just buy a 7600K and overclock it. The more cores you have, the harder it is to sustain high clock speeds.
If you really insist single-threaded performance matters that much, a 5950X is a stupid choice.
Sure, on the CPU side of things, games are held back by the slowest thread, but most CPUs these days are plenty fast enough to, for example, give you a good VR experience.
But gamers, especially ones that browse forums like Guru3D, don't want a "good VR experience" they want "the best VR experience" (personally they probably don't want any VR experience, they want the best gaming experience but I'm just going with the analogy). So regardless to whether single thread performance is good or not in terms of measuring a CPU's "performance" in today's workloads, it's a massive selling point for AMD amongst gamers - many of whom are the most vocal in IT communities and stuff.
---
Edit: I think that's an all-core fixed-clock overclock.
Which obviously will reduce single-threaded score significantly. 3900X should get 530 or so on default with decent memory.
Temperature.
Zen 2 (and probably Zen 3 too) boost higher if the ambient (and it's own temp) is significantly lower.
I did a test last winter by letting windows open on sub-zero temps outside, CPU booted with just 6 degrees at Windows desktop, ran a benchie as soon as I could and got the highest Cinebench and 3D Mark CPU scores I ever had, over 7% higher than my previous max !
It does look like he is running a fixed manual (4.3GHz) overclock alright.
Wow the CPU temp was 6C ?? Where were you running that machine - in a freezing cold shed out the back garden? 😀
But for me the higher score was in summer compared to now so... although I am not running PBO at all. Must try and turn it on...
But gamers, especially ones that browse forums like Guru3D, don't want a "good VR experience" they want "the best VR experience" (personally they probably don't want any VR experience, they want the best gaming experience but I'm just going with the analogy). So regardless to whether single thread performance is good or not in terms of measuring a CPU's "performance" in today's workloads, it's a massive selling point for AMD amongst gamers - many of whom are the most vocal in IT communities and stuff.
Fine - let's go with G/Free Sync displays that can go up to 144Hz. Most modern desktop CPUs with a sufficient core count are capable of going beyond that (GPUs are usually the bottleneck but for argument's sake, let's say it isn't a bottleneck). Even though you technically can get higher frame rates, that doesn't give you the "best" experience, because you're literally not experiencing the best.
So really, the matter is "gamers want the best bragging rights" and well, those who think single-threaded performance matters today are the same sort of people. Nobody cares about your (not you, specifically) PC or your benchmark result. It doesn't matter if you can render dozens or even hundreds of more frames than your monitor can render. No amount of fancy gear is going to make you in the same league of tournament winners.
There's a difference between being an enthusiast and being vain.
Wow the CPU temp was 6C ?? Where were you running that machine - in a freezing cold shed out the back garden? 😀
In my bedroom / home office, after turning off heating and letting the windows open for about 40 mins in full early winter freeze during the night.
I was wearing boots and thick pants and jacket, gloves for my hands... and benchmarking =)
After I was done, closed things, let about 2 hours for the air to mix naturally, then turned back the heating.
Not sure if this was the one, but considering the GPU itself boosted to 2,114 (which it NEVER does on normal 23 deg ambient), it probably is.
https://www.3dmark.com/spy/9265792
Too bad no temps are stored on the online website 🙁
---
p.s. I will almost certainly repeat the operation after buying a Zen 3 CPU, just for the giggles ^.^
I have no means to do chilled water or LN2 or other stuff like various youtubers, but I can always open my windows when it's freezing outside !
In my bedroom / home office, after turning off heating and letting the windows open for about 40 mins in full early winter freeze during the night.
I was wearing boots and thick pants and jacket, gloves for my hands... and benchmarking =)
After I was done, closed things, let about 2 hours for the air to mix naturally, then turned back the heating.
Not sure if this was the one, but considering the GPU itself boosted to 2,114 (which it NEVER does on normal 23 deg ambient), it probably is.
https://www.3dmark.com/spy/9265792
Too bad no temps are stored on the online website 🙁
---
p.s. I will almost certainly repeat the operation after buying a Zen 3 CPU, just for the giggles ^.^
I have no means to do chilled water or LN2 or other stuff like various youtubers, but I can always open my windows when it's freezing outside !
"I was wearing boots and thick pants and jacket, gloves for my hands.." - That's dedication 😀
Yes the CPU overclocks in certain temp "bin" groups (15MHz is it) as Jay is always going on about. The lower the ambient CPU temp can be to start, the better.
I'd say getting a large cooling air-con unit to blow on system is probably easier than your method though.
...However... there is hardly anything today that does any serious number-crunching and isn't multi-threaded.
From daily experience I kindly disagree. There are far too many single-threaded bottlenecks happening all the time. The most stark example is image formats like JPEG, PNG, TIF and everything else. These are almost exclusively decoded on single-threads. Even worse, most software decodes these one after another even when multiple images are to be decoded.
Next example is PDF, a mostly linear format that programmers have to go out of their way to decode multi-threaded. And once it includes image data (and lots of it in the PDF files i use regularly) then you are back to the point where you can see pages being build up.
CAD software often runs on lower thread-counts plus a bottlenecking main-thread while the user actively draws a sketch. Only the last rendering step (often using the Cine engine) makes full use of all CPU cores.
Topaz AI software is mostly bottlenecked by a single thread when a new processing step begins (especially previews when you scroll around an image or change any setting.
Sure, on the CPU side of things, games are held back by the slowest thread, but most CPUs these days are plenty fast enough to, for example, give you a good VR experience.
Not everything is about frames per second, though. I literally wasted hours and days of my life waiting for Total War: Warhammer 2 to finish its turns. Every single turn-over can takes literal minutes to finish in later multiplayer campaign games. On the plus side, you can finish filling the tumbler in the meantime.
Every small (mb sized) Steam update for its games files take quite some time to finish even with a fast PCIe M.2 SSD, because the small download data is repackaged into gb sized data files.
Then there are games like World of Warcraft where some users heavily rely on third-party LUA addons to make the UI bearable. These LUA scripts are all compiled on a single thread when you enter a server world and are re-loaded again via a single thread with every load screen (instance/area chance).
So, although single-threaded performance isn't meaningless, it doesn't really matter anymore, because nothing worth caring about is single-threaded and bottlenecked by modern CPUs.
There are two main bottlenecks that cause people having to wait for their computers to finish a task: HDD/SSD throughput (especially low queue 4k access) and software waiting for its main-thread to finish some task.
To make things worse, Windows tends to shift around threads all over the place from core to core hardly allowing the CPU to even reach its maximum Turbo clock-rates unless you specifically enable core parking (introducing its own drawbacks).
We are still living in the digital stone-ages and wide-spread "current" software often is build on years and even decades old foundations. This is a reality we need to tackle in practice and high single-thread performance is the way to do so.
If you really insist single-threaded performance matters that much, a 5950X is a stupid choice.
If you insist that *both* single-threaded and multi-threaded performance matters then the choice does not look so stupid anymore. The main drawback seems to be the price-performance ratio compared to the 5900X, the premium seems quite steep.
Fine - let's go with G/Free Sync displays that can go up to 144Hz. Most modern desktop CPUs with a sufficient core count are capable of going beyond that (GPUs are usually the bottleneck but for argument's sake, let's say it isn't a bottleneck). Even though you technically can get higher frame rates, that doesn't give you the "best" experience, because you're literally not experiencing the best.
So really, the matter is "gamers want the best bragging rights" and well, those who think single-threaded performance matters today are the same sort of people. Nobody cares about your (not you, specifically) PC or your benchmark result. It doesn't matter if you can render dozens or even hundreds of more frames than your monitor can render. No amount of fancy gear is going to make you in the same league of tournament winners.
There's a difference between being an enthusiast and being vain.
I would say that people usually don't think for themselves (on any matter), they just follow some trend, and that's very sad.
https://valid.x86.fr/c9jb89 It was even faster a few months ago... 526/8263 https://valid.x86.fr/ita54n