timecop1818 yea.. and intel is better ?? you forgetting all those quad core chips intel kept giving the mainstream, while telling every one we dont need more then 4 cores for the mainstream.. or how about their 95 watt cpus using up to 200 watts ??
Even if tile-based rendering isn't often done on laptops, what in that makes this *benchmark* irrelevant?
Benchmarks aren't always chosen based on what lots of people use, but rather than by their ability to produce useful metrics - in this case multi-threaded computing when restricted by only the execution resources, power and thermals.
I swear some of the most rabid fanboys have a few words they Ctrl-F from the page and react to - without even reading the entire post or article. And that goes for both camps - I've seen it from some people, to defend AMD against some imagined slights (a few words taken out of context).
Yup. You can guarantee that whatever is actually being said, somebody will carry it off-topic to a talking point they feel more secure in - even if it's totally irrelevant and they're not actually right about that, either. Tends to be how most political discussions go, too...
Yeah, there have been lottsa scandals about doctored benchmarks. Oh wait....
AMD under Lisa, have a rep as very straight shooters. Intel only tell the truth for practice. They have zip cred ATM - a laughing stock.
"Absolutely excelling" is a tautology btw. I highly doubt a semi literate finds paid work to justify a fancy laptop.
Nobody makes a living running benchmarks either - any pc - any brand. Benchmarks means "indicators", & are used by both - perhaps innappropriately, but no side can set the rules. they are simply what consumers have come to expect
I think a large part of this stems from confusing benchmarks with real word performance (at a given task that is different than the benchmark).
Benchmarks usually highlight one aspect or side of a product, and you use multiple benchmarks to prong it from different sides.
Cinebench and other tile-based renderers are used to gauge the maximum achievable multi-threaded performance under full load, when there is no inter-thread communication required that would affect the compute efficiency. You use other benchmarks to get at the single- and lightly threaded workloads, and do some real world tests.
It's only the totality of different kinds of tests that tells (enough of) the whole story. Since reviewers have a large audience, they can only give general recommendations on the kind of workloads a product is suitable for, and the kinds it's the best or one of the best for. It's on every individual to consider the tests that are relevant to their specific use case and make their own decisions. (Also: Follow multiple review outlets that do the kind of testing that caters to your use case.)
I think the big whinging from the Intel fans comes because, when Zen first released, Cinebench was very much a best-case scenario for AMD's chips - especially Threadripper. It showed them in the best possible light, while performance elsewhere wasn't so hot.
AMD have sorted most of those issues now, but Cinebench has become an easy way to compare their product generations - and sure, it still shows AMD's thread-heavy products in their best light. But the trolls don't have new arguments because Intel don't have new products, so they're going back to the same old ones.
Its also not just up to AMD. As it was noted in various independent tests, developer optimisations for hw can make or break results in both games and data centre space.
Most developers implement Intel coding which of course does tend to behave somewhat better on Intel hw, and AMD is left to 'reach' Intel from brute forcing alone (which doesn't exactly makes the playing field fair to begin with).
In most recent tests of HEDT and data centre CPU's such as ThreadRipper and EPYC, it was noted that when developers optimised a program for Zen uArch, performance improved by over 50%.
So, we need to bear in mind that in the software field, performance can radically vary as code can be selective. We need devs to write code (or have AI write code) which can execute as efficiently as possible (and make use of anything the said hw has to offer) on any hw without discrimination.
I know that some people don't think there are people who do serious gaming and productivity work on laptops, but there are.
For such people (like me), we do like high performing multithreaded performance which can be sustained indefinitely (especially for things such as 3d Studio Max which easily maxes out all the cores/threads when rendering animations).
Then there's video-editing, and occasional gaming.
We also need these systems to be portable, so people like me actually like systems such as Acer Helios 500 PH517-61 which have desktop 2700 and Vega 56 with powerful cooling. I know its a desktop replacement with not so great battery life, however, it IS highly portable (infinitely so in comparison to a desktop), not to mention, quiet, and can easily sustain maxed out CPU/GPU performance indefinitely while remaining quiet and cool (cooler than some desktops even).
Anyway, I like the fact that 4800H for example seemingly comes within a spitting distance of 3700x (performance-wise) in just 45W TDP envelope, and has a capable iGP to boot which would enhance battery to a high degree (plus, depending on which dGPU is used, it would also be a capable gaming machine).
Emphasis on single threaded performance is not a big deal for lots of people since Zen uArch has been quite capable of doing this nicely. Zen 2 is another ballpark though, so its certainly welcome that we are now going to have real mobile hw with some serious performance punch.
However, as you know, OEM execution will be key. AMD can have superior hw, but if OEM's don't include capable cooling and then mismatch other internal components and cut corners, its going to be a problem (but this is not something AMD has influence over sadly).
I just hope OEM's stop treating mobile users as second class citizens and do things competently with high quality control on AMD hw this time around.
Real world performance should reach the advertised benchmarks/numbers if the cooling is done competently... if its not, we know OEM's will be to blame.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
406 Comments
View All Comments
Korguz - Tuesday, January 7, 2020 - link
timecop1818 yea.. and intel is better ?? you forgetting all those quad core chips intel kept giving the mainstream, while telling every one we dont need more then 4 cores for the mainstream.. or how about their 95 watt cpus using up to 200 watts ??Hul8 - Tuesday, January 7, 2020 - link
Even if tile-based rendering isn't often done on laptops, what in that makes this *benchmark* irrelevant?Benchmarks aren't always chosen based on what lots of people use, but rather than by their ability to produce useful metrics - in this case multi-threaded computing when restricted by only the execution resources, power and thermals.
Spunjji - Tuesday, January 7, 2020 - link
They've got their trigger words, and by heck they're going to spin with them."Synthetic benchmark!"
"Vendor-provided!!"
"Efficiency!!!"
Meanwhile the rest of us can be casually excited and wait for the reviews before deciding what's not good enough.
Hul8 - Tuesday, January 7, 2020 - link
I swear some of the most rabid fanboys have a few words they Ctrl-F from the page and react to - without even reading the entire post or article. And that goes for both camps - I've seen it from some people, to defend AMD against some imagined slights (a few words taken out of context).Spunjji - Wednesday, January 8, 2020 - link
Yup. You can guarantee that whatever is actually being said, somebody will carry it off-topic to a talking point they feel more secure in - even if it's totally irrelevant and they're not actually right about that, either. Tends to be how most political discussions go, too...msroadkill612 - Tuesday, January 7, 2020 - link
Yeah, there have been lottsa scandals about doctored benchmarks. Oh wait....AMD under Lisa, have a rep as very straight shooters. Intel only tell the truth for practice. They have zip cred ATM - a laughing stock.
"Absolutely excelling" is a tautology btw. I highly doubt a semi literate finds paid work to justify a fancy laptop.
Nobody makes a living running benchmarks either - any pc - any brand. Benchmarks means "indicators", & are used by both - perhaps innappropriately, but no side can set the rules. they are simply what consumers have come to expect
Hul8 - Wednesday, January 8, 2020 - link
I think a large part of this stems from confusing benchmarks with real word performance (at a given task that is different than the benchmark).Benchmarks usually highlight one aspect or side of a product, and you use multiple benchmarks to prong it from different sides.
Cinebench and other tile-based renderers are used to gauge the maximum achievable multi-threaded performance under full load, when there is no inter-thread communication required that would affect the compute efficiency. You use other benchmarks to get at the single- and lightly threaded workloads, and do some real world tests.
It's only the totality of different kinds of tests that tells (enough of) the whole story. Since reviewers have a large audience, they can only give general recommendations on the kind of workloads a product is suitable for, and the kinds it's the best or one of the best for. It's on every individual to consider the tests that are relevant to their specific use case and make their own decisions. (Also: Follow multiple review outlets that do the kind of testing that caters to your use case.)
Spunjji - Thursday, January 9, 2020 - link
Nailed it.I think the big whinging from the Intel fans comes because, when Zen first released, Cinebench was very much a best-case scenario for AMD's chips - especially Threadripper. It showed them in the best possible light, while performance elsewhere wasn't so hot.
AMD have sorted most of those issues now, but Cinebench has become an easy way to compare their product generations - and sure, it still shows AMD's thread-heavy products in their best light. But the trolls don't have new arguments because Intel don't have new products, so they're going back to the same old ones.
deksman2 - Wednesday, January 15, 2020 - link
Its also not just up to AMD.As it was noted in various independent tests, developer optimisations for hw can make or break results in both games and data centre space.
Most developers implement Intel coding which of course does tend to behave somewhat better on Intel hw, and AMD is left to 'reach' Intel from brute forcing alone (which doesn't exactly makes the playing field fair to begin with).
In most recent tests of HEDT and data centre CPU's such as ThreadRipper and EPYC, it was noted that when developers optimised a program for Zen uArch, performance improved by over 50%.
So, we need to bear in mind that in the software field, performance can radically vary as code can be selective.
We need devs to write code (or have AI write code) which can execute as efficiently as possible (and make use of anything the said hw has to offer) on any hw without discrimination.
deksman2 - Wednesday, January 15, 2020 - link
I know that some people don't think there are people who do serious gaming and productivity work on laptops, but there are.For such people (like me), we do like high performing multithreaded performance which can be sustained indefinitely (especially for things such as 3d Studio Max which easily maxes out all the cores/threads when rendering animations).
Then there's video-editing, and occasional gaming.
We also need these systems to be portable, so people like me actually like systems such as Acer Helios 500 PH517-61 which have desktop 2700 and Vega 56 with powerful cooling.
I know its a desktop replacement with not so great battery life, however, it IS highly portable (infinitely so in comparison to a desktop), not to mention, quiet, and can easily sustain maxed out CPU/GPU performance indefinitely while remaining quiet and cool (cooler than some desktops even).
Anyway, I like the fact that 4800H for example seemingly comes within a spitting distance of 3700x (performance-wise) in just 45W TDP envelope, and has a capable iGP to boot which would enhance battery to a high degree (plus, depending on which dGPU is used, it would also be a capable gaming machine).
Emphasis on single threaded performance is not a big deal for lots of people since Zen uArch has been quite capable of doing this nicely.
Zen 2 is another ballpark though, so its certainly welcome that we are now going to have real mobile hw with some serious performance punch.
However, as you know, OEM execution will be key.
AMD can have superior hw, but if OEM's don't include capable cooling and then mismatch other internal components and cut corners, its going to be a problem (but this is not something AMD has influence over sadly).
I just hope OEM's stop treating mobile users as second class citizens and do things competently with high quality control on AMD hw this time around.
Real world performance should reach the advertised benchmarks/numbers if the cooling is done competently... if its not, we know OEM's will be to blame.