The developer turned out to be really responsive (and made me feel bad having started the bashing thread on here before), see this thread: https://www.passmark.com/forum/pc-hardware-and-benchmarks/46757-single-thread-score-rating
Yeah, UserBenchmark will be pure fun to watch.
The [PassMark] developer turned out to be really responsive (and made me feel bad having started the bashing thread on here before), see this thread: https://www.passmark.com/forum/pc-hardware-and-benchmarks/46757-single-thread-score-rating
Will be interesting whether he feels the need to tweak the benchmark again to better reflect the real life differences, or if those are considered spot on.
Pretty bad form by Intel with their grand mess of "real life benchmarks" that never got resolved.If you look to the follow-up post I made on twitter then you see an Intel document which they were spreading stating that his benchmark was no good.
View attachment 32180
They were actively hurting his bussines.
I've been pondering what will be their methodology from now on. They've essentially removed multi-core results (> 4 threads) from the equation and after the release of the 3300X added memory latency as a huge part of the score.Wait for UserBenchmark scores, that's where the real fun begins.
I've been pondering what will be their methodology from now on. They've essentially removed multi-core results (> 4 threads) from the equation and after the release of the 3300X added memory latency as a huge part of the score.
Since memory latency doesn't seem to have been improved much on the Ryzen 5xxx series (based on GB 4 results) I guess they'll double down on that approach.
They will probably be listing points like: "Having 15% better memory latenct makes CPU A much better than CPU B in games" (despite CPU B actually winning in games due-to much larger L3 cache).
Either that or list quicksync perfomance as an essential part of the overall score
Pretty bad form by Intel with their grand mess of "real life benchmarks" that never got resolved.
Either that or list quicksync perfomance as an essential part of the overall score
*Shakes Magic 8-Ball*Wonder if it can do all core 5Ghz OC with normal cooling. Remember last year the 5Ghz hype train collapsed, I was also a victim *facepalm*
5GHz all core isn't happening, but I do think Ryzen 5000 might be better at overclocking.*Shakes Magic 8-Ball*
"Signs point to yes"
If I were you I'd just go ahead and assume you can hit 5 GHz all-core OC with a Wraith Prism on the 5950X.
I check the web for comparisons periodically and every time I've checked so far CPU encoding overall produced better quality at the same bitrate compared to NVENC. Don't get me wrong, NVENC is perfectly fine for twitch streaming and such, but if you're quality oriented CPU AFAIK is still the better option.Actually on topic of video encoding, i think the game is over for all CPUs. Turing nvenc made amazing strides in encode quality, so encode is pretty much no longer irrelevant on CPUs unless the user is encoding some future format at 0.1FPS.
Same is steadily happening in render space, those GPU accelerators of 30TF are making CPU renders less and less relevant each year.
I check the web for comparisons periodically and every time I've checked so far CPU encoding overall produced better quality at the same bitrate compared to NVENC. Don't get me wrong, NVENC is perfectly fine for twitch streaming and such, but if you're quality oriented CPU AFAIK is still the better option.
Before Turing generation I had same opinion, but the new generation NVENC is just damn decent enough and offers incredible speed/power efficiency. I think beyond some special purist scenario it offers incredible quality.
There will always be things like new codecs (AV1 etc) where encodes are not supported on GPU and CPU is the only option. But for H264/H265 Turing+ NVENC made CPU encoding irrelevant for non-pro usage.
And for casual streaming it is a slaughter.
If you use same bit rate then the size is going to be the same. Doesn't matter what encodes it and what codec is being used.Do GPU encodes still have much higher file sizes than CPU encodes?
If you use same bit rate then the size is going to be the same. Doesn't matter what encodes it and what codec is being used.
It wouldn't shock me if no one "who matters" even knew this white paper exists but that's just me. Intel has a systemic culture and people problem both behavioral and in the hiring process but it's a great company and it's still got much to offer. Unfortunately, these issues are becoming more and more a common occurrence (Intel continuously drawing poor Optics, Negative press but that will have to take its course until bad actors are removed.If you look to the follow-up post I made on twitter then you see an Intel document which they were spreading stating that his benchmark was no good.
View attachment 32180
They were actively hurting his bussines. A 70 B$ company putting pressure on an individual developer....
This is true, but I was looking more for a comparison at the same quality. If you can set the same constant bitrate and not have a noticeable difference I'd be really impressed.
If You throw same "decent" bitrate at them, nvenc encodes fine - for example:
The problem is that NVENC does not support what software based encoders do -> two pass and encoding and image quality driven modes. So pretty much impossible to compare them apples to apples in as tight bitrate as possible modes and by definition software will end up with better quality at same bitrate.
But nvenc destroys software in efficiency and speed.
NVENC does have constant quality mode and also supports lossless encoding. Personally I never use fixed bit rate nor two pass encoding (and average bit rate).If You throw same "decent" bitrate at them, nvenc encodes fine - for example:
The problem is that NVENC does not support what software based encoders do -> two pass and encoding and image quality driven modes. So pretty much impossible to compare them apples to apples in as tight bitrate as possible modes and by definition software will end up with better quality at same bitrate.
But nvenc destroys software in efficiency and speed.
Launch is on Nov. 5th. Reviews should go live then at the latest naturally, possibly earlier, who knows.Excuse me if it has been mentioned before, because I am sure it has, but when on earth are the actual reviews going to hit and when do they hit the stores? hmmm