|||

[ Edited by Velcro_SP 20.04.2011 - 06:18 ]

> let's go to the videotape

On the videotape it's a "800 GHz processor" ;-)

http://www.youtube.com/watch?v=E5sxS-muxjQ

> you said even if you were wrong you were right

I said that even if Cherrypal didn't simply lie about the CPU's clock speed but added up the clock speeds of various SoC components it was still deceptive marketing. I stand by that.
In case Cherrypal really added up the clock speeds of the CPU core, the GPU and the DSP I would have liked them stating something like "Telechips ARM11 800 MHz (500 MHz CPU + GPU + DSP) system-on-chip" as the CPU core's clock speed is an important spec to be able to compare the device with other ARM11 based devices.

> isn't your read of that text unconventional?

If it is then obviously TurboPascal's read is as well as he calls his findings a "significant mismatch" compared with the claimed specs.

> Why would the parentheses and repeated numerical symbol for addition be in
> there if Cherrypal wasn't saying that the "800 MHz" adjective was directed at
> the SoC, and that it was adding the frequencies of all three sub-units?

As I said, I didn't relate the part in parentheses to the preceding "800 MHz" part but to the succeeding "system-on-chip" part, to state that the SoC comprises a CPU, a GPU and a DSP. In that way of reading that spec the numerical symbol "+" would just mean "and".

> One person's plausibility is another's vicious rumor-mongering then.

Apropos, TurboPascal now went even farther than me, talking about the C515's successor:

"in good old Cherrypad tradition we can become very sceptical about the hardware matching the specs. To me the "Cherrypad 2" looks like a Kinpad I600. That will mean there might be some misfit between the actual CPU speed, which might rather be 800 Mhz and the announcement talked about 1 GHz, but we'll see."
http://www.cherrypal.com/support/index.php?topic=1366.msg2034#msg2034

> What if there are are multiple CPU components?

Usually, multiple CPU components mean multiple identical CPU cores at same clock speed, so this is spec'd in the form of "2x 800 MHz", which in this context doesn't equal to "1x 1600 MHz" of course.

> Cherrypal says "MHz" here and I dunno if Cyrix did that

They didn't.

http://en.wikipedia.org/wiki/Performance_Rating#Branding
http://en.wikipedia.org/wiki/Cyrix#PR_system

> How do they describe the speed of these machines with multiple cores now?

See above.

> What do other SoC-system sellers do?

THTF/LimePC for instance add up the DMIPS figures of the MPC5121e's components:

http://www.google.com/search?q=site:limepc.com+1700mips

The MPC5121e's e300 core alone delivers about 800 DMIPS at 400 MHz.
In my opinion, running an industry standard benchmark on either component and adding up the results is a slightly more truthful way than just adding up the clock frequencies, especially with asymmetric designs like SoCs, but still far from ideal.

> Buyers are supposed to pay attention to what is actually said.

Cherrypal's processor spec for the C515 is ambiguos at best.

> It'd be sneaky if the SoC had poorer performance than
> an 800 MHz CPU, but I don't know if that is the case.

Actually, if the C515's SoC's ARM11 core operates at 500 MHz it *has* poorer performance compared to an ARM11 core operating at 800 MHz as well as compared to devices using the same SoC but with the CPU core operating at 800 MHz. The CPU core's general functionality is always usable to its full extend on the consumer's part whereas the usage of other cores like DSPs depends on specific software support for them. Most software won't be able to use the DSP so I'd call simply adding its clock speed to the CPU core's clock speed sneaky indeed.

> its movie-playing ability, which I read was good.

...and would be even better with a faster clocked CPU core.

bottom line is if i pay for 800mhz... goddammit...it best be 800mhz, in this case i assume we are dealing with bait and switch, or getting rid off all the older arm chips which they aren't the only one's doing to get rid of search tigerdirect to see one. I did previously. (600mhz arm processor on a tablet i believe) The efika is clocked at 400mhz, shows up at 396 I think haven't use it lately.

|||

[ Edited by Velcro_SP 20.04.2011 - 06:19 ]

Quote:

AFAIK the clock is really 400MHz and little difference here comes from the OF which calculates the clock a bit unprecisely.

Quote:

I'm just pointing out that "adding" clockspeeds is intentional dishonest, and manufactors doing so are doing it to fool costumers.

So yes they are lying in my book.

|||



[ Edited by Velcro_SP 20.04.2011 - 06:20 ]

No, it simpler than that. In a non-SOC system we don't add the speed of the of the non CPU components to the CPU to quote speed. Just because a system is integrated into one chip does not make it right to start adding up the speeds of the different sub-systems. It IS dishonest.

[ Edited by Jim on 2011/1/30 1:55 ]

|||

[ Edited by Velcro_SP 18.04.2011 - 19:26 ]

+1 for those that believe that adding clock speeds of different parts of a SoC part is misleading and a very bad business practice.

I am surprised that anyone would try to defend this practice, but some people get stuck on a position and end up supporting things that they normally would scoff at under different circumstances, or for other companies.

Thanks Dave. I was beginning to wonder where this bizarre idea came from and why anyone would buy into it. It's deception plain and simple.

|||

[ Edited by Velcro_SP 19.04.2011 - 08:01 ]

|||

[ Edited by Velcro_SP 18.04.2011 - 19:24 ]

> like Andreas said and I had read about a long time ago, Freescale
> adds the MIPS of the components of the 5121e SoC

No, I didn't say that. Read again what I actually said about whom. And I can't see that Freescale do what you say they'd do either:

http://www.google.com/search?q=site:freescale.com+mpc5121e+mips
http://www.google.com/search?q=site:freescale.com+mpc5121e+dmips

Nowhere on these hits do Freescale add "the MIPS of the components of the 5121e SoC". Moreover, Freescale is open about the MPC5121e's components' particular clock speeds.

> that's an okay way to do it.

I still don't think so. As I said I think it's slightly more truthful than adding clock speeds, but still far from ideal.

> adding the MHz of SoCs is also an okay way IMO, and no less honest, as long
> as a manufacturer or retailer is open and tells how it is adding.

Cherrypal doesn't tell the CPU core's clock speed, which for the potential customer is the most important clock speed figure in a SoC and in a CPU in general. If Cherrypal also told the CPU core's clock speed beside the sum of all the SoC components' clock speeds I'd be okay with it. But they don't.

> Saying adding clockspeeds is dishonest is like saying dividing
> them is noble or obtaining their square root is callous.

Huh? It's obvious (see TurboPascal's posting in the Cherrypal support forum) it's not absolutely clear that the "800 MHz" figure is derived from adding up clock speeds at all. Like I told, Cherrypal's processor spec for the C515 is ambigous at best and there're at least two different ways to read it.

> It's about how accurate the result is and whether its communicated honestly.

So you think it's communicated honestly by Cherrypal, and TurboPascal, snipe and me are just not able to understand this honest communication?

> Who said anything about adding the speed of the non-CPU components to the CPU?

You said, there: "when you talk about SoC speed, there's more than just the CPU to consider".

> I referred to adding the speed of all the components of the SoC.

...which in the case of the C515's SoC apart from the CPU include a GPU and a DSP, the latter two being non-CPU SoC components in the context of this discussion. (Sometimes, the SoC as a whole is referred to as a "CPU", but clearly not here).

> Why should it be dishonest simply because it is different from non-SoC systems.

It's misleading because it's *not* different from non-SoC systems in the context of clock speeds. Clock frequencies don't work in a way that it would technically make sense to add them up across different components, no matter if SoC system or non-SoC system.

> Freescale's representation of the 5121e SoC's three cores as having
> a combined 1700 MIPS

...has been invented by you.

> It's a combined measure of performance.

DMIPS figures are measures of Dhrystone benchmark performances. While often mistaken for doing so by people who are less educated in the field of computer hardware, clock frequencies as such don't indicate performance unless the compared hardware is sufficiently similar on the core level. In the case of the C515 you could use the *real* clock speed of its ARM11 core to compare it to other ARM11 based tablets. If you know only the sum of the components' added clock speeds you can't really compare the performance to another device's performance because you don't know how the sum is portioned between the components. Theoretically speaking, if the "800 MHz" is the result of adding up clock speeds the C515's SoC could have a 100 MHz CPU, a 100 MHz GPU and a 600 MHz DSP. Knowing that a DSP, like I told you, is only usable to a very small range of software, do you anticipate the difference this would make in terms of real performance in everyday usage?

|||

[ Edited by Velcro_SP 18.04.2011 - 19:23 ]

> LimePC said so [...]. And THTF's sales agent did so

Yes, this time you got it right. They did so, and I said they did.

> Excuse me for getting the company in your comment wrong

Excuse granted ;-)

> you casually referred to an "industry standard"

Yes, that's what the Dhrystone benchmark is.

> and said it was "more truthful" to add that way

Yes, that's what I said. And in order to put this assessment into perspective I added that this way was still far from ideal.

> which threw me off.

How so?

> these links refer to a Freescale SoC, and most expressly say "Freescale."

Yes, no doubt about that, as it *is* a SoC from Freescale actually. But that doesn't mean those spec claims have been made by Freescale.

> it's not underhanded at all, not comparatively or relatively, to add the frequencies
> of components on an SoC, as long as it is made clear that is what is being done.

As I said, it's not that clear that this is what has been done with the C515's processor specs claims made by Cherrypal, else that thread wouldn't have been started by TurboPascal on cherrypal.com, speaking about a "significant mismatch". And as Jim tried to get across to you, to the usual customer it doesn't matter if it's a SoC based system or not because he wouldn't even notice from using the device without opening it. SoC or not, I find adding up clock speeds of the CPU, the GPU, a DSP and what else might be in the system and presenting this sum as spec misleading at best, unless the particular clock speed of the CPU (which I took the "800 MHz" claim for until you made me aware of how Cherrypal might have "calculated" this figure) is given as well.

|||

[ Edited by Velcro_SP 18.04.2011 - 19:22 ]

> You alluded to a company that combined MIPS measurements

"Allude to" in the sense of "explicitly naming", yes.

> provided a websearch with "Freescale" among its search terms.

You seem confused. The websearch with the "freescale.com" term was provided by me only *after* you falsely claimed that I was referring to Freescale, to prove to you that it's *not* Freescale adding DMIPS numbers. The websearch in my prior posting which you didn't read properly only contained the terms "limepc.com" and "1700mips", nothing more.

> Seeing no reason why adding MIPS should be any "more truthful" than adding MHz

Just in case you don't see it yet (in case you do see, feel free to ignore this paragraph): Dhrystone is a performance benchmark, which can be used to compare different types of cores with each other. Clock frequencies are not directly related to any kind of performance and can only be used to indicate performance with hardware that is similar on a core level. CPUs, GPUs and DSPs are very different kinds of hardware on a core level, but if you run actual code on either one (compiled from the same source code that is), you're able to compare their actual performance in terms of the benchmark that consists of said code. So it actually has *some* relevance regarding the SoC's overall performance in terms of Dhrystone benchmark to be able to state the added up DMIPS figures. Of course, without telling each core's stake it's not of that much use, but still better than the absolutely meaningless clock frequency sum that is derived from adding up clock frequencies of sub-units.

> I took your expression "industry standard" to refer to the combining of MIPS
> measurements. But you meant the MIPS benchmark only, not the combining of it.

Actually, by "industry standard" I did not only "mean" the Dhrystone benchmark alone, I also *wrote* it exactly that way:

"running an industry standard benchmark on either component and adding up the results"

> As has been clearly stated now, it is actually THTF's agent and LimePC
> that have done this combining

Yes, stated by me first and then reiterated by you.

|||

[ Edited by Velcro_SP 18.04.2011 - 19:20 ]

> there's no reason to get bent out of shape.

It's just that I object to attempts at rewriting the discussion's process.

> I thought you meant Freescale.

Yes, that's obvious. What isn't obvious to me yet is how you came to this conclusion when I explicitly named who I meant from the very beginning.

> If Freescale wasn't among the search's terms. It was among many of the results.

All 7 results went to limepc.com (due to the "site:limepc.com" part of the search). Only *one single* of those 7 results mentions the term "Freescale":

http://www.google.com/search?q=site:limepc.com+1700mips+freescale

> Its SoC was among all of the results

Yes, that was the purpose of my websearch link after all.

|||

[ Edited by Velcro_SP 18.04.2011 - 19:18 ]

> The one for the X1 refers to Freescale too, though it misspells it.

Yes, you're right there, thanks. That makes it 2 out of 7 results

> I just described the discussion as I recalled it in response to your question

So it seems you recalled it incorrectly. Fair enough. Sometimes it's better to actually re-read a part of a discussion instead of relying on flaky memory.

> Seems like I saw Freescale referred to elsewhere in that search too.

You mean in others of those 7 results than the 2 with the "Freescale"/"Freescal" mention?

> Google's results for identical search terms can vary.

Usually not in a search where "site:" is specified, like I did. So I doubt that anybody (including you) doesn't get exactly those 7 results, of which exactly 2 contain the term "Freescale" or a derivation of it. But even if all 7 results contained the term "Freescale" or a derivation of it they would still reside on limepc.com, not on freescale.com or any other site.

> But okay that's not "many."

Thanks. I'm glad this is clarified.