Userbenchmark does more shady stuff

[GDRRiley]

Just now, ShrimpBrime said:

You know damn well you'd cry running a 1.8ghz mobile cpu and wonder why your frame rates sucked just as much if he cleaned this thread.

hey I've got a few core 2 duo laptops I've used and my currently daily laptop has a 3500U which isn't much faster

[leadeater]

1 minute ago, GDRRiley said:

I'm just giving you the option if you feel the need to. I'm not going to sit here crying that a mod cleaned a thread I made.

Well since I'm commenting I won't be moderating the thread. It just amazes me how willing a place such as Userbenchmark is at denying reality. Ryzen 4000 mobile is a legitimate important step forward in that market, it's important that it succeeds not because I like AMD but because Intel will have to compete with it. If Intel can better it then great, but lets not make it unnecessary to have to.

[Guest]

Just now, GDRRiley said:

hey I've got a few core 2 duo laptops I've used and my currently daily laptop has a 3500U which isn't much faster

I can't stand laptops. It's a good way to throw out money. I found that out with core 2 laptops lol.

[mr moose]

So to stick witht he 4900HS and ignore the rest here is exactly what userbenchmark had to say:

https://cpu.userbenchmark.com/SpeedTest/1084184/AMD-Ryzen-9-4900HS-with-Radeon-Graphics

Quote

Ryzen 4000 Mobile CPUs offer benchmark busting multi-core performance on the go, but marketing hype aside, it’s unclear how this will translate to real world performance gains for laptop users. Gamers are better off with low latency CPUs. At launch, the top GPU available in a 4000 series laptop is the RTX 2060. Since the GPU is largely responsible for overall gaming performance, the Ryzen 4000 laptops will offer mid tier gaming performance at best. Pairing stronger GPUs would be sub optimal because the Zen latency bottleneck becomes increasingly severe with more powerful GPUs. Streamers and media producers, who have historically benefited from CPU cores, are better off using the GPU (NVENC or QuickSync) for encoding. Leading media creation applications including both DaVinci Resolve and Adobe Premiere Pro are largely GPU bound. Sixteen threads can indeed be useful in corner case workloads (UserBenchmark 64-core, Cinebench, Handbrake (CPU mode), Blender (CPU mode)) but for the majority of consumers most of the threads will remain idle. With low power consumption and high core counts, the 4000 range, on paper at least, is a perfect fit for the datacenter. AMD should focus on delivering a platform that offers performance where end users actually need it rather than targeting inexperienced gamers with the same old "moar cores" mantra. From a gamer’s perspective, the best feature of the 4000 series laptops is the absence of 5000 series GPUs. ^{[Mar '20 CPUPro]}

Looks to me like they are talking directly to gamers here.  First is anything they are saying actually wrong?

[leadeater]

1 minute ago, ShrimpBrime said:

I can't stand laptops. It's a good way to throw out money. I found that out with core 2 laptops lol.

Yea I will only buy the worst pile of crap laptop possible and use it to remote in to a better system.

[leadeater]

3 minutes ago, mr moose said:

Pairing stronger GPUs would be sub optimal because the Zen latency bottleneck becomes increasingly severe with more powerful GPUs.

Factually wrong.

 

3 minutes ago, mr moose said:

it’s unclear how this will translate to real world performance gains for laptop users. Gamers are better off with low latency CPUs.

Also nope.

 

 

Here is an example of a more powerful GPU performing worse because it actually is being limited by the CPU, to the extent the "slower" system is faster. There is no game and GPU where the 4900HS will hold it back more so than an Intel mobile CPU, it will in fact do so less.

[GDRRiley]

1 minute ago, mr moose said:

So to stick witht he 4900HS and ignore the rest here is exactly what userbenchmark had to say:

https://cpu.userbenchmark.com/SpeedTest/1084184/AMD-Ryzen-9-4900HS-with-Radeon-Graphics

Looks to me like they are talking directly to gamers here.  First is anything they are saying actually wrong?

the more cores line.

apparently 8 is a high core count data center chip

apparently the slightly higher latency means these game bad cus a few of the older and esports games they used don't likely ryzens cores.

[Guest]

What makes Ryzen 4000 laptop cpus appealing is performance efficiency. 

Instructions per clock cycle at a lower watt and all about that battery life advertisement. 

Throw a 2060 in there, it starts to look powerful. And in a video processing sense it is. 

The Cpu just simply lacks the Frequency more so than anything.

[GDRRiley]

Just now, ShrimpBrime said:

What makes Ryzen 4000 laptop cpus appealing is performance efficiency. 

Instructions per clock cycle at a lower watt and all about that battery life advertisement. 

Throw a 2060 in there, it starts to look powerful. And in a video processing sense it is. 

The Cpu just simply lacks the Frequency more so than anything.

they do have the frequency 4.4 isn't bad and with the better IPC they fight even the top 9900HK in 90W mode

[mr moose]

15 minutes ago, leadeater said:

Factually wrong.

 

Also nope.

 

 

Here is an example of a more powerful GPU performing worse because it actually is being limited by the CPU, to the extent the "slower" system is faster. There is no game and GPU where the 4900HS will hold it back more so than an Intel mobile CPU, it will in fact do so less.

 

The 4900HS with 2060 is not as good at gaming as other CPU's with better GPU's and the only way to prove it being sub optimal with larger GPU's is if we can put them on a see first hand yes? does this mean the issue is the use of the term latency?

 

EDIT: for the record can we get links to anything other than hardware unboxed, I hate that guy and it's not helping that the entire thread seem to link him as the only source of information.

[GDRRiley]

3 minutes ago, VegetableStu said:

i pity first time PC hardware hobbyists because newcomers tend to look for easy comparison numbers first. UB would probably be doing them a disservice for metrics and overviews that not many in the review industry fully agree with

all the comparison sites screw with them. from ----boss to UB.

no one who does reviews rights agrees with them

[Guest]

Just now, GDRRiley said:

they do have the frequency 4.4 isn't bad and with the better IPC they fight even the top 9900HK in 90W mode

 

It's 4.3ghz max boost (XFR single core) and would not apply to a Cinebench argument. https://www.amd.com/en/products/apu/amd-ryzen-9-4900hs

I suspect the All core sustained boosts around 3.8ghz to stay within the thermal design point for the cooler.

But yes the 4900HS is a beast. It seems to be about as fast as my 2700x even! 

 

 

 

 

 

[xAcid9]

20 minutes ago, mr moose said:

Looks to me like they are talking directly to gamers here.  First is anything they are saying actually wrong?

They're not entirely wrong but just showing one side of the story is pretty dick move. Unethical. 

 

 

 

 

As a user of DaVinci Resolve, not once it got bottleneck by my GPU.

 

[GDRRiley]

Just now, ShrimpBrime said:

It's 4.3ghz max boost (XFR single core) and would not apply to a Cinebench argument. https://www.amd.com/en/products/apu/amd-ryzen-9-4900hs

I suspect the All core sustained boosts around 3.8ghz to stay within the thermal design point for the cooler.

But yes the 4900HS is a beast. It seems to be about as fast as my 2700x even!

single core result in there too. yep it is putting 2700s to shame. I'm sorry my 2700 time to get you to 4.2 or higher.

[Guest]

Just now, GDRRiley said:

single core result in there too. yep it is putting 2700s to shame. I'm sorry my 2700 time to get you to 4.2 or higher.

Love my itx board. Performance enhancer Level 4 runs 4300mhz sustained. Set it and forget it. (gotta keep temps down though)

 

Sitting on my hands. Waiting on DDR5 desktop. Ryzen 3000 series upgrade to 3700x for me if 4000 desktop drives prices low enough. I can't keep buying processors at 350 bucks every 1.5years lol.

[GDRRiley]

4 minutes ago, mr moose said:

EDIT: for the record can we get links to anything other than hardware unboxed, I hate that guy and it's not helping that the entire thread seem to link him as the only source of information.

https://www.anandtech.com/show/15708/amds-mobile-revival-redefining-the-notebook-business-with-the-ryzen-9-4900hs-a-review

[mr moose]

2 minutes ago, GDRRiley said:

https://www.anandtech.com/show/15708/amds-mobile-revival-redefining-the-notebook-business-with-the-ryzen-9-4900hs-a-review

So userbenchmarks mistake here was assuming it's latency would be the same as other ryzen processors?

[leadeater]

It's probably also worth pointing out when it comes to Ryzen Mobile 4000 the HS is more than just a model SKU it's a design standard OEMs have to abide by to be allowed to use that CPU. So it may well be that AMD is not allowing anything higher power than an RTX 2060 Max-Q to be paired with an HS or requires a larger battery when using a RTX 2070 Max-Q but the OEM was not able to meet other design rules to accommodate it.

 

We will see higher end GPUs in laptops with H series Ryzen Mobile CPUs but those don't exist yet. If you want a power and efficiency optimized laptop, not the best possible gaming laptop, then current HS laptops are a good choice. What is not a good choice is HS CPUs for gaming focused laptops because that is not what the HS CPUs are for.

 

We are all, including Userbenchmark, commenting on incomplete data and therefore lack direct evidence to prove or disprove certain claims. However we are not lacking in usable evidence to make educate observations to comment on things we expect to see.

 

And anyway when it comes to laptops design power and cooling design means more than the sum of parts in it, laptops with the same CPU and GPU can perform quite differently.

[leadeater]

13 minutes ago, mr moose said:

The 4900HS with 2060 is not as good at gaming as other CPU's with better GPU's and the only way to prove it being sub optimal with larger GPU's is if we can put them on a see first hand yes? does this mean the issue is the use of the term latency?

That is not what they said, they said ALL Ryzen Mobile 4000 are worse and should not be paired with high power GPUs and base that statement off a single select data point which is of desktop class products.

 

They took one data point, made a factually incorrect statement because it's not at all latency (that is the factually incorrect part), and made a broad all encompassing claim.

 

In that entire statement I don't think a single thing said was correct, other than showing that Fortnite at that time did run better on Intel architecture just not for the specific reason they stated.

[GDRRiley]

1 minute ago, mr moose said:

So userbenchmarks mistake here was assuming it's latency would be the same as other ryzen processors?

in part yes. they also didn't seem to make the conclusion that when CPU bound it out performs a intel chip with the same GPU meaning it has more CPU power so if it shouldn't be paired with anything better than a 2060 nether should any intel mobile chip. all according to them

 

[leadeater]

17 minutes ago, mr moose said:

EDIT: for the record can we get links to anything other than hardware unboxed, I hate that guy and it's not helping that the entire thread seem to link him as the only source of information.

Doesn't matter if you don't like him that data is actually accurate and the testing methodology they use has proven to be so. You can pick any source you like the results are the same with one exception, Userbenchmark.

 

If you are the odd one out then maybe look at yourself and not others first. Comment to Userbenchmark.

[mr moose]

Just now, leadeater said:

That not what they said, they said ALL Ryzen Mobile 4000 are worse and should not be paired with high power GPUs and base that statement of a single select data point which is of desktop class products.

They said it was due to latency,  they then linked to a 3600X bench as proof, but the 3600X has worse latency than the 4900HS which has better latency again over Intel.  That whole statement might be an observation (seeing Intel CPUs with better GPU's perform better) that was put down to an assumption (the cause of the performance was due to latency). 

 

Just now, leadeater said:

They took one data point, made a factually incorrect statement because it's not at all latency (that is the factually incorrect part), and made a broad all encompassing claim.

They only have 6 test for the 4900HS.  And they clearly didn't look into the design or latency before posting.

Just now, leadeater said:

In that entire statement I don't think a single thing said was correct, other than showing that Fortnite at that time did run better on Intel architecture just not for the specific reason they stated.

 

With number of samples when used in comparison at a whopping 0.01% I am sure there is room for improvement. 

[leadeater]

6 minutes ago, GDRRiley said:

in part yes. they also didn't seem to make the conclusion that when CPU bound it out performs a intel chip with the same GPU meaning it has more CPU power so if it shouldn't be paired with anything better than a 2060 nether should any intel mobile chip. all according to them

 

If we are going to cherry pick then lets just drive right in to the deep end, go big or go home.

The 4900HS is over 3 times faster than the i7-9750H, that's right you heard it here first!

 

[mr moose]

5 minutes ago, leadeater said:

Doesn't matter if you don't like him that data is actually accurate and the testing methodology they use has proven to be so. You can pick any source you like the results are the same with one exception, Userbenchmark.

 

If you are the odd one out then maybe look at yourself and not others first. Comment to Userbenchmark.

It does matter, because in an effort to avoid my own biases,  I like to get information from sources that don't piss me off.

[mr moose]

1 minute ago, leadeater said:

If we are going to cherry pick then lets just drive right in to the deep end, go big or go home.

The 4900HS is over 3 times faster than the i7-9750H, that's right you heard it here first!

 

I haven't got to that bit yet, it's a big article and I keeps stopping to ask questions.