Core i9-9900K, benchmarking controversy, adios mainstream hyperthreading? - The Full Nerd Ep. 71
video description
Date: 2022-03-15
Related videos
Comments and reviews: 10
brad
With all due respect, your discussion about hyperthreading is misguided and misleading. It think that the approach that you are taking is doing a disservice to the community. Your comments are assuming that all threads are created equal when they are definitely not. The second thread on each cores is allowed to run opportunistically and use the available resources that would otherwise be sitting idle as the main thread queues for memory access etc. It does not get the same amount of access to CPU resources as the Primary thread.
All you need to do is run the single and multithreaded cinebench tests. The multiplier that the application calculates for you is an indication of the performance equivalence of a CPU that only has a single thread running on each physical cpu cores that the hyperthreaded/SMT is allowing.
4c/8th performs at about the same level as 5 physical cores. That is only 25% the performance for each HT thread compared to a primary thread.
The 8700K 6c/12th performs in the range of a 7-7. 5 physical core equivalent CPU.
The 9th Gen range has certainly changed the way Intel are delivering the performance but they are stuck at 14nm so can't leverage a smaller node to gain any benefits in performance. Intel are limited in the extra frequency they can squeeze out of 14nm so the only other thing that they have left to boost performance this generation is to increase the number of cores.
The Coffee Lake Generation offered what was effectively a 4, 5, 6 and 7-7. 5 physical core equivalent range of CPUS. 9th gen Has not mentioned any 4 core chips as yet. The announced chips are offering 6, 8 and 9. 50-10 core equivalent performance.
I guess that Intel could have offered a faster 6c\12th SKU instead of the 8c/8th chips but they already have one in the 8700K that can be overclocked to 5Ghz and another 6 core SKU doesn't have the same marketing sparkle as an 8 core SKU. A new 6/12 SKU is unlikely to out perform an 8700K Who would want to buy it?
reply
With all due respect, your discussion about hyperthreading is misguided and misleading. It think that the approach that you are taking is doing a disservice to the community. Your comments are assuming that all threads are created equal when they are definitely not. The second thread on each cores is allowed to run opportunistically and use the available resources that would otherwise be sitting idle as the main thread queues for memory access etc. It does not get the same amount of access to CPU resources as the Primary thread.
All you need to do is run the single and multithreaded cinebench tests. The multiplier that the application calculates for you is an indication of the performance equivalence of a CPU that only has a single thread running on each physical cpu cores that the hyperthreaded/SMT is allowing.
4c/8th performs at about the same level as 5 physical cores. That is only 25% the performance for each HT thread compared to a primary thread.
The 8700K 6c/12th performs in the range of a 7-7. 5 physical core equivalent CPU.
The 9th Gen range has certainly changed the way Intel are delivering the performance but they are stuck at 14nm so can't leverage a smaller node to gain any benefits in performance. Intel are limited in the extra frequency they can squeeze out of 14nm so the only other thing that they have left to boost performance this generation is to increase the number of cores.
The Coffee Lake Generation offered what was effectively a 4, 5, 6 and 7-7. 5 physical core equivalent range of CPUS. 9th gen Has not mentioned any 4 core chips as yet. The announced chips are offering 6, 8 and 9. 50-10 core equivalent performance.
I guess that Intel could have offered a faster 6c\12th SKU instead of the 8c/8th chips but they already have one in the 8700K that can be overclocked to 5Ghz and another 6 core SKU doesn't have the same marketing sparkle as an 8 core SKU. A new 6/12 SKU is unlikely to out perform an 8700K Who would want to buy it?
reply
Outkast
PT knew exactly what they were doing, if your in the business of testing/benchmarking and still cant manage to do it right after 15 years i think its time to step down and stop wasting peoples time and money, the interview with steve (gn) was dog shit, he didn't answer anything he was asked in full, all he did was fumble around with the questions until he came up with an im sorry excuse for it. there is no excuse,
pc world, Gamers Nexus, Linus Tech Tips, jays 2 cents, Unbox Therapy, bitwit, jokers productions, pauls hardware, hot hardware, tech report, techgage.
how is it that all these people can get it right over and over and if they make a mistake they are pretty quick to mention it and correct it, yet PT cant manage to do a simple red vs blue benchmark? after 15 years? i dont buy it for a second they are scum imo
reply
PT knew exactly what they were doing, if your in the business of testing/benchmarking and still cant manage to do it right after 15 years i think its time to step down and stop wasting peoples time and money, the interview with steve (gn) was dog shit, he didn't answer anything he was asked in full, all he did was fumble around with the questions until he came up with an im sorry excuse for it. there is no excuse,
pc world, Gamers Nexus, Linus Tech Tips, jays 2 cents, Unbox Therapy, bitwit, jokers productions, pauls hardware, hot hardware, tech report, techgage.
how is it that all these people can get it right over and over and if they make a mistake they are pretty quick to mention it and correct it, yet PT cant manage to do a simple red vs blue benchmark? after 15 years? i dont buy it for a second they are scum imo
reply
Ruben
AdoredTV published a great video a couple days ago, researching the fact that Intel has both ownership in, and contributes money/developers/hardware to Principled Technologies AND the software companies that develop some of the most widely used benchmarking tools out there.
Really makes me wonder if the -don't attribute to malice what you can attribute to incompetence- narrative that reviewers are invoking when talking about PT's testing methodology should be given credence over the thought (or some may say fact, that this was basically a thinly veiled -inhouse- test done by Intel, using a their front -third party- to legitimize the results.
reply
AdoredTV published a great video a couple days ago, researching the fact that Intel has both ownership in, and contributes money/developers/hardware to Principled Technologies AND the software companies that develop some of the most widely used benchmarking tools out there.
Really makes me wonder if the -don't attribute to malice what you can attribute to incompetence- narrative that reviewers are invoking when talking about PT's testing methodology should be given credence over the thought (or some may say fact, that this was basically a thinly veiled -inhouse- test done by Intel, using a their front -third party- to legitimize the results.
reply
Benign
AMD is like France in the scope of industry and security, and Intel is like Deutschland in Der Reich. While AMD might be first to do several things, they do so poorly enough that Intel doesn't really have to conduct itself in a fair or desperate manner. In order to maintain financial equilibrium, they gotta pull this random garbage and stir up media for a few months, hoping to cash in on raw exposure and attain a greater win than otherwise. Dear Intel, just release good value products and we will be happy. As of now, so much negativity surrounds the brand and I am afraid to give my confidence to the brand in any way, shape, or form.
reply
AMD is like France in the scope of industry and security, and Intel is like Deutschland in Der Reich. While AMD might be first to do several things, they do so poorly enough that Intel doesn't really have to conduct itself in a fair or desperate manner. In order to maintain financial equilibrium, they gotta pull this random garbage and stir up media for a few months, hoping to cash in on raw exposure and attain a greater win than otherwise. Dear Intel, just release good value products and we will be happy. As of now, so much negativity surrounds the brand and I am afraid to give my confidence to the brand in any way, shape, or form.
reply
LFly
Until Shintel develops a completely new architecture from grounds up in 14 nm or 12 nm, Shintel cpu-s will continue to overheat like lava, kill VRMs and memory banks when OCing or turboing, have faulty inefficient hyper-thrading, nuclear power consumption that never alligns with TDP, absolutely no multi-core performance and all the Meltdown and Spectre security bugs on a hardware tranzistor level: -. Meltdown and Spectre OS and BIOS patches constantly lower Shintels IPC performance and will continue to do so until Shintel takes 2 years to develop a completely new architecture from grounds up: -.
reply
Until Shintel develops a completely new architecture from grounds up in 14 nm or 12 nm, Shintel cpu-s will continue to overheat like lava, kill VRMs and memory banks when OCing or turboing, have faulty inefficient hyper-thrading, nuclear power consumption that never alligns with TDP, absolutely no multi-core performance and all the Meltdown and Spectre security bugs on a hardware tranzistor level: -. Meltdown and Spectre OS and BIOS patches constantly lower Shintels IPC performance and will continue to do so until Shintel takes 2 years to develop a completely new architecture from grounds up: -.
reply
Cinnabuns2009
Only issue is, for people that have been benchmarking for so long, and then doing -all these things- incorrectly (in the favor of their commissioners no less) presumes incompetence. Why would Intel commission incompetent people to run their benchmarks? Simple answer is, they wouldn't. That is the crux of the issue. Then after Gamer's Nexus interviews the guy and asks all the questions, doesn't get any answers and the -follow up- also doesn't provide any valid answers either. Like interviewing a politician and then later getting more skirting answers in the follow up.
reply
Only issue is, for people that have been benchmarking for so long, and then doing -all these things- incorrectly (in the favor of their commissioners no less) presumes incompetence. Why would Intel commission incompetent people to run their benchmarks? Simple answer is, they wouldn't. That is the crux of the issue. Then after Gamer's Nexus interviews the guy and asks all the questions, doesn't get any answers and the -follow up- also doesn't provide any valid answers either. Like interviewing a politician and then later getting more skirting answers in the follow up.
reply
Bigdog302V8
Intel over the years has not changed at all, Remember the Athlon 64 vs Pentium 4? they actually contacted benchmark makers in the 2004-2005 era to program things that gimped AMD products, Cennebench is one of those that comes to mind of the offenders to score higher on Intel than AMD, Intel also released the Pentium 4 based on the Zeon based Gallatin cores called the Extreme Edition, it was more like Emergency Edition really. Nvidia itself is like Intel where they are anti competitive and cheats regularly to win.
reply
Intel over the years has not changed at all, Remember the Athlon 64 vs Pentium 4? they actually contacted benchmark makers in the 2004-2005 era to program things that gimped AMD products, Cennebench is one of those that comes to mind of the offenders to score higher on Intel than AMD, Intel also released the Pentium 4 based on the Zeon based Gallatin cores called the Extreme Edition, it was more like Emergency Edition really. Nvidia itself is like Intel where they are anti competitive and cheats regularly to win.
reply
Yamil
-Gordon, Brad
We have to be reasonable regarding that resolution testing thing. Who buys a $300-$500 CPU with a $1000+ GPU to pair it with a 1080p monitor? I would argue that testing at 1440p and 4k is more useful for regular gamers who would actually buy those parts, since it sounds extremely rare to buy that kind of hardware for lower resolutions.
Testing at 1080p is necessary to see the worst case scenario that a user could potentially see, but using that alone is misleading at best.
reply
-Gordon, Brad
We have to be reasonable regarding that resolution testing thing. Who buys a $300-$500 CPU with a $1000+ GPU to pair it with a 1080p monitor? I would argue that testing at 1440p and 4k is more useful for regular gamers who would actually buy those parts, since it sounds extremely rare to buy that kind of hardware for lower resolutions.
Testing at 1080p is necessary to see the worst case scenario that a user could potentially see, but using that alone is misleading at best.
reply
Benjamin
The didn't have the memory timings at stock for the Intel test benches they loaded XMP and dropped the ram ratio to get the 2666 clock. They used significantly tighter timings on the Intel systems than they did on the AMD rigs. On the subject of what Intel might have told PT because they are or are not allowed to do something, Intel was not allowed to pay off OEMs to not use their competitor yet they did that very thing for some years.
reply
The didn't have the memory timings at stock for the Intel test benches they loaded XMP and dropped the ram ratio to get the 2666 clock. They used significantly tighter timings on the Intel systems than they did on the AMD rigs. On the subject of what Intel might have told PT because they are or are not allowed to do something, Intel was not allowed to pay off OEMs to not use their competitor yet they did that very thing for some years.
reply
Aquilam
All the testing proved is -if you have infinite money, you can build a better gaming machine with Intel-.
The i9-9900K is selling for $530 (cheapest, which is $570 after tax) on amazon, $570 on newegg and B&H. without a cooler. Minimum you can argue is it's $600 CPU (with a -really- cheap cooler. AMD's 2700X is $280 at microcenter (with a cooler. you can buy a CPU and decent GPU for the same price as the 9900K.
reply
All the testing proved is -if you have infinite money, you can build a better gaming machine with Intel-.
The i9-9900K is selling for $530 (cheapest, which is $570 after tax) on amazon, $570 on newegg and B&H. without a cooler. Minimum you can argue is it's $600 CPU (with a -really- cheap cooler. AMD's 2700X is $280 at microcenter (with a cooler. you can buy a CPU and decent GPU for the same price as the 9900K.
reply
Add a review, comment
Other channel videos
