Quite sad that Anand is leaving AnandTech, he's one of the few people to do architectural 'deep-dives'. He mentioned that he's been distancing himself from writing for a while. Lets hope they can uphold the quality of the write-ups we're used to seeing.
And it's not long since David Kanter went part-time at RWT. I'm not too sure of other places to look for the same standard of writing, any thoughts?
I would be interested in new places to find that standard of writing as well because I currently use Hexus and AnandTech. The others I know of are in the same category as Hexus so not worth looking at in addition to Hexus.
I'm quite a fan of TechReport these days - they don't always go into the arch detail as much as Anand did, but their empirical testing is rather good - for eg their task based power draw figures (http://techreport.com/review/26977/i...sor-reviewed/5)
CAT and others have pointed out some weirdness with some of TR's comparisons though so I'm wary of them. The power draw/time graphs are interesting though; Tom's also seem to have a decent power analysis setup but again there seems to be some irregularities in their testing methodologies IIRC.
Also I remember seeing a site recently which started using a 'scope to measure power consumption but was just posting raw scope screenshots which aren't terribly useful as you just see a load of spikes which can be miles higher than 'real' power draw - that's one reason why caps are used in power supply circuitry. I thought it was Tom's but can't seem to find it, and it doesn't seem to be the case with some of their recent reviews.
They sometimes make choices with their choice of competition or benchmarks that are a bit odd, but nothing worse than say Hexus. However as long as you apply some discernment to read what they're testing it's usually good - the review I linked paints the FX8350 in fantastic light if you look at the results - it's second only to the most premium of intel chips.Originally Posted by watercooled
Yeah I suppose so; it's just a case of looking at the results whilst bearing in mind their testing methodology etc and working it out for yourself.
BTW it was Tom's I was thinking of: http://www.tomshardware.co.uk/amd-a1...w-33021-8.html
How are those 'high resolution' graphs supposed to be useful to readers? Granted, they do tell you the averages, but because of the zoom level, you tend to get an imaginary line far higher up than the important averaged power draw.
It's also important to know what you're doing when using a scope for stuff like this, taking into account sample rate, scope bandwidth, the characteristics of the load you're measuring in conjunction with these, and in Tom's case, inductive effects of their current clamps.
One really bad-case example would be measuring something like a PWM load with a scope set to a perfectly synchronised sample rate - you might end up measuring all 'ons' or all 'offs', ending up with zero or 100% duty figures rather than the actual draw.
I'm not suggesting the websites don't know what they're doing (or at least I certainly hope they do), but to a lot of readers the raw data might be very misleading.
Last edited by watercooled; 31-08-2014 at 07:37 PM.
I think raw data is more useful than limited data that is explained badly. I don't read reviews to take what is said in them at face value and as such I find I don't read conclusions anymore. The data and the method of getting it is more important to me, the discussion of that data is interesting but the conclusion is less than useful most of the time.
So I read reviews like a picture book after I've read through AnandTech's discussion of their testing and Hexus' article plus the comments in the forum thread that we all post to. I find more value on the Hexus forums than out of both articles to be honest and it is the reason Hexus is my go to tech site; the reviews, giveaways and care@Hexus scheme are all great things to have but without the community it would be another data source for me, not a place to discuss and learn about technology from its members.
Other places are great, as sources of data that I read and bring back here to discuss. Their conclusions are farts in the wind to me and I find it weird that people place so much value on other people's conclusions, it's better to make your own.
Well it's not really raw data, just a graph compiled from it; knowing all of the raw data plus settings could provide useful information, but the graphs shown don't really do that as I explained, and could be more misleading that anything to readers not familiar with scope traces.
I've been waiting for this timer to expire to see the announcement. http://centtech.com/
I know it isn't and I wouldn't be able to understand the raw data just like the majority of people reading the article. Anyone can look at raw data but very few can compile it into an intelligible article; in other words very few people can read and understand raw data so it is useless to present it without an argument to help make it intelligible; an intelligible article, however, can help people unable to understand the raw data get an idea of why the raw data is important, which is far more helpful to a reader.
So I go back to what I said; reading the arguments, or points made, about data that is collected and presented in an intelligible manner is very helpful for myself and other tech enthusiasts that have not worked in the industry or studied the subject at university. Raw data, and conclusions, are less than helpful in comparison to discussion about the presented data, even if the data presented is accompanied by subjective reasoning that is influenced by personal preference (in other words bias opinion). Humans can handle personal preference and subjective reasoning more easily than raw data because we think subjectively and thus can process opinions more easily. Right or wrong, bias or not, opinions are the way we convey information from one person to the next; we need to work with them.
I work with opinions by reading two articles from authors I trust, ones I know to be consistent even if bias, and then scan through other articles to get an idea of how many points made in the primary articles match the ones made elsewhere to give myself an idea of which are more likely to be accurate and helpful. Through this a better understanding is achieved and I would hesitate to recommend that anyone reads only a few good articles or reads the full article of every person writing about the same technology.
I read Hexus and AnandTech- I then "read" The Tech Report, Guru3d, TechSpot and bit-tech. There are others I go through when interested in buying something specific but general reading is limited to those at the moment which is why I am interested in hearing about other sites used by Hexus members.
No I think you might be missing my point about the example I linked. They do give average numbers which is generally the useful figure, however the graphs and I suspect the 'maximum' figures too, aren't telling people what they'll think it is, and hence they're likely to misinterpret them.
The graphs shown are not averaged like we normally see e.g. in that Tech Report article linked by kalniel, which filters out the large but short-lived spikes. It's not a great comparison, but another example of where you're going to see very large current spikes is PSU inrush current; this is often in the area of 40A or even higher. That's the sort of power a high-power electric shower will pull. But even though this current massively exceeds the rating of the fuse in the plug, and even the the 20/32A MCB in the consumer unit, it doesn't trip them as it's so short-lived. Using this inrush current to calculate 'maximum' power would obviously be hugely misleading.
My argument isn't against using graphs, just *those* graphs, which aren't useful to most people because of the lack of averaging, and aren't really useful to people more familiar with scope readings either because of the timescale used. The same goes for the 'maximum' figures shown, which I suspect are peak instantaneous figures, not something which is going to show up on the AC side.
Wonder when the NDA drops on the 285? Today is supposed to be the release date, but we usually see stuff like this around midday.
NDA is supposedly over at 8AM EST or 12 noon GMT.
Sounds about right.
Still not impressed with that Centaur website; I was looking forward to seeing some news on their Isaiah II!
8370, 8370e (reviewed) and 8320e. http://www.anandtech.com/show/8427/a...ew-vishera-95w
Slightly disappointed in the 8370e - looks like it's really not turboing all that much in tests so still some way behind the 8350.
R9 285 reviewed: http://techreport.com/review/26997/a...-card-reviewed
Last edited by kalniel; 02-09-2014 at 01:28 PM.
The e Vishera CPUs seem to have pretty impressive power efficiency considering they're still on the same node. Look at TR's numbers, they're even approaching Kaveri! Although I must admit I'm suspicious of their methodology considering the 4790k is faster than the 5960X in their x264 benchmark, and very close in the Handbrake one. The postage stamp video strikes again?
The AnandTech results also show a remarkable drop in power consumption over their release 8350! It's a pity about the platform it's still stuck on though, AM3+ really isn't brilliant in terms of idle consumption vs newer platforms like 115x/FMx.
Also, reviews of the 285 are starting to trickle out, seems a bit meh overall. Granted, it offers new features and driver optimisations might improve performance further, but it's pretty much the same as the 280 it replaces in performance, and around the same area in terms of power efficiency. The fact all the reviews I've seen so far seem to be testing OC models probably isn't helping efficiency though.
There are currently 55 users browsing this thread. (0 members and 55 guests)
Posting Permissions
LinkBack URL
About LinkBacks
Reply With Quote
