Hyperthreading might save Itanium

I'VE BEEN FOLLOWING the letters about HT (Hyper-Threading) and thought I might have something to add.

Processor architectures are reaching the point where they can't squeeze more out of instruction streams by parallelising serial code either dynamically (as Superscalar architectures like the Pentium 1/2/3/4 do) or statically (as the Itanium does). With chips now having 100+ million gates, designers have all these spare transistors lying around and just chucking them at caching and more execution units is yielding deminishing returns.

So what is HT for? I think it is a first foray into gaining some experience at utilising parallel instruction streams. It was added to the P4 at relatively low cost and will help performance of some applications, but not others. More importantly it will give Intel the field experience to pave the way for proper multi-threaded architectures. More comprehensively multi-threaded processors could prevent processor stalls by switching contexts on cache misses and keep extra execution units busy at other times. But as we are seeing it needs to be done carefully so as not to hurt the performance of unoptimised applications wherever possible.

I think HT might actually save the Itanium architecture. Everyone seems to see it as a relatively small addition, but Itanium has all those registers and execution units sitting around half of them not doing anything - there is only so much you can do with static scheduling. HT will bring the dynamism it needs and should help boost the performance significantly.

The competing and possibly complementary approach being used by several other manufactureres are multi-core SMP-on-a-chip designs.

For both approachesthe, the key is software: not new optimising, compilers, but properly written multithreaded software. At the moment, many software engineers simply don't understand the nuances of concurrent programming - programmers find it hard because they are still using 1960's ideas about concurrency like Monitors and Semaphores rather than properly scalable concurrent methodologies like CSP channels (although this has origins in the 1970's!). Multithreading is seen as a performance overhead (and it is on a uniprocessor) that you avoid if you can. Chip manufacturers know this which is why they are still desperately trying to squeze every last ounce of serial performance out of their chips.

HT is just a toe in the water.

Regards,
Jim

I'm surprised you're surprised; some resources inside the P4 are partitioned between the two threads when hyperthreading is enabled. I believe, for instance, that the eight write buffers are split between the two threads. Even if the other thread is not doing anything, each thread only has access to four buffers each. So for applications that experience performance increases due to write buffers, for instance, the single-thread performance will decrease. It takes a chip reset in order to restore the resources back to a single thread.

As I recall, Intel is working on improving some of these limitations for subsequent rollouts.

You will notice that when you run benchmarks on this sp1 the scored are more evenly match like sandra 2002 like my 2700+ in mips scores the same as a 3000+ at the same speed also the pentium 4's dont score 9100 mips anymore there more evely match to the 2002 scores. It looks like me that maybe sisoft has been warned about bias benchmarking because now it looks more even. my oced 2700+ at 2.25 ghz beats a 3.06 ghz in mips.

Might want to check it out

The Problem
There is a problem and I can't believe it hasn't already been dealt with. The problem is microprocessor marketing. It takes form in handy jargon such as "quantispeed", "hyperthreading", and especially "megahertz", and lately "gigahertz". A while back, AMD introduced (an industry re-introduction if memory serves) the performance rating and very specifically designated it as equating current processors to the would-be performance of an Athlon with the Thunderbird core.

Now I haven't done a lot of research (none actually), and I don't know if the Thunderbird core scales perfectly linearly, or if there is a perfect model of the core's performance as clock speed increases, and so I don't know if the model rating numbers are perfectly accurate. Two things are apparent though. The first is that AMD never claimed that the performance numbers were targeting Intel processor speeds. The second is that AMD's performance ratings, though questionably implemented from 2800+ and above, give the consumer a far better idea of what they are purchasing than do megahurts.

The proof for this last statement is abundant and began with the introduction of the Pentium 4. This is supported by the on going class action lawsuit against Intel in which the plaintiffs claim that they had believed that they were buying (and paying the premium for) Intel's best and fastest CPU, only to find that their machines were bested by the "slower" Pentium 3.

Now the P4 has come a long way since. It's the fastest clocked microprocessor in the world, recent refinements have improved the efficiency of the processor (buss speed increases, hyperthreading (hypermarketing maybe?) and I will be the first to tell you that it is excellent technology.

But the current market conditions are such that there is no reliable metric by which processors are judged. Which is faster - a 3.0 ghz P4 or a 3.06 ghz P4? What about a 2.0 ghz P4-m or a 1.6 ghz Pentium-M? The situation is that AMD has been taking a lot of heat lately, and I don't think this is necessarily unjustified. But more importantly, I can't believe that Intel has not. The only way to really be certain about what you are buying is to go out and run about fifty benchmarks, take all of those results and compile them, divide this number by the price of the processor, and there is your final performance rating (or some similar method). Now you all go and tell all of your friends, relatives (think mom and dad, sons and daughters) neighbors, co-workers, and everyone you pass on the street that this is how they need to proceed. Bring a camera so you can take a picture of the look on their faces.

The Irritant
There is an irritant to the marketing problem and believe it or not it is those who we trust most. What hardware reviews sites have you visited, or perhaps even authored lately? These sites merely compound the problem. If it is one thing that hardware review sites thrive off of, it is confusion. How many hardware sites reviewed the Athlon XP 3200+? Ten? Many more than that, I'm sure. If there wasn't any confusion, if there wasn't any controversy, there would really only need to be one review, wouldn't there? That would mean that if you'd read one hardware site's review, you'd have read them all. Now I'm pretty sure that this is the last thing that the hardware review sites want. If they all had the same conclusion, the forums would be just a little bit less active, the page hits would be a little less astronomical. After all, you wouldn't need to visit all of those hardware sites, you'd only need to visit one of them.

The solution
The worst part about the solution is that it has been available to us the whole time. Here we are with the tools to evaluate and communicate at our fingertips. The benchmarks and the internet. It really is that simple. It really is that simple. I'll say it a third time - it really is that simple! The hardware sites claim that the performance rating needs to be adjusted. I say that the performance rating needs to be reborn. The hardware sites need to be the ones to bring the new performance rating into this world. They need to collaborate, come to a consensus, and establish …gasp…a standard performance rating using agreed upon benchmarking software. These benchmarks should meet the criteria of fairness as well as relevance. All of this bullcrap with BAPCO and the likes isn't going to do anyone any favors, least of all the consumer. It's okay if a benchmark shows that one processor outperforms the other, just so long as the benchmark hasn't been optimized in order that it guarantees such a result. Let all of the fanboys throw themselves on the floor and scream like the babies they are. A performance metric should never be about whose processor has the best rating. It should indicate either one or both of the following: the performance of a processor and the value of the processor (value = performance / price).

Henceforth, all Intel and AMD cpu's will carry a performance rating - the same performance rating. Sure, the processor will each be better in there niche applications, and there is no reason that the performance rating can't supply this information in the form of a categorical performance breakdown. However, at this point in time Intel's lack of a performance index is actually far more deceptive than AMD's performance rating that may be skewed by a couple of percentage points (however, AMD needs to step up to the plate and be responsible about their current system).

Dangerous times are upon the consumer. Microchip architectures are about to diverge even farther than they already have. Confusion about performance will only grow unless the hardware community begins to act like a community.

I wonder about the hardware community. Do they ever talk to each other? Would there be any harm if Anandtech, Tomshardware, Tech-Report, AcesHardware all called each other up or perhaps arrange to meet in conference with each other to set up a benchmark package that they could all run in confidence. To heck with what AMD and Intel claim about their processors, we will have our own performance rating. At the end of this article I am left feeling sad. I know that the chances that the aforementioned cooperation should ever take place are probably less than nill. I'm pretty certain that with each future processor release the confusion and controversy will only increase. AMD and Intel themselves probably enjoy this climate of ever increasing FUD. Their marketing machines feed on it. I hope however, that after reading this, the hardware community will come to believe that they have the tools and the talent, all then that is lacking is the will to 'bring balance to the force'. All of this in the name of justice, and the Fudified consumer.

Yours truly,
Jay

Subject: http://www.theinquirer.net/?article=8170

Hmmm, maybe it's me but this seems quite biased and pretty much un-true. How can he venture to imply that the only people who enjoy this game will be C&C addicts already. I personally hate RTS Games. Never liked them. Even Starcraft bored me greatly. However, i played this game and loved every minute of it.... Pretty odd for a game that only a C&C addict would like. All i'm saying is a review shouldnt be based completely on an attitude of "i don't like this so no one else will." Don't let this place become MSN Slate Magazine.... :)

Long time INQ fan, Ryan Vennell

The really qestionable thing there is that most people who have Hyperthreading capible processors- would they turn it off? Probably not, but AMD might have a point. By it's nature, Hyperthreading would accelerate threaded applications. As it is, whatever OS threads are available would be of low significance, and you're left with a benchmark application, which I cannot imagine would be all that rich in threads, so the overhead for the threads might actually slow the processor down in straight raw benchmarketing.

Does the Intel Represenative suggest we benchmark P4s while playing back MP3s so that they have enough to do?

Email address supplied

...while he does raise a number of valid points, he just doesn't address them logically. The PR rating is not the solution to the problem. The problem is to reduce the benchmarks to "tasks performed".

Of course then exactly what the task is and how the task is performed become relevant (and, inevitably cloaked in a forest of names from different benchmark developers). But, in no way shape or form does the running of two year old benchmarks solve this problem...especially not by the company that makes the chips. The PR rating just doesn't stand up to the Light of Reason. And those who defend it are nothing more than simple-minded minions and slaves of the forces of Untruth and Illogic.

Anon

Subject: Real iLoo inventor hits out at Microsoft "slur"

Please, please, please stop giving this british "inventor" credit for the idea of a computerized toliet. The idea has been around in some for or another for at least the past 30 years, when a diagram of a toliet broken apart and presented as an I/O device was circulated. Then there's this website ( ) which has been around for at least a year before Microsoft's April Fools announcment. All Microsoft was doing was rehashing an old joke. Tell the inventor that he can go flush his legal options.

Jim Schuler

Subject: Did Microsoft software crash trap finance minister in limo?

As I understand it, the 7-series doesn't need M$ to screw up - according to the press, the whole software bundle in the 7 series is wedded into one improbable tangle, rather than maintained as discrete modules. [German engineering weakness - the belief that if things can be joined together, they should be joined together.] Car and Driver published a letter from a reader who had bought a 7-series car that shortly after delivery developed the nasty habit of randomly turning on the radio, with the volume level set at ear-shattering. He said he returned the car to have the problem fixed, and it took a month for the dealer to get it back to him - all the software in the car had to be dumped and a fresh installation made - and then the complete package had to be checked out.

John

Subject: Did AMD's Athlon XP 3000+ earn its rating?

this is yet another illogical, irrational whinge from an AMD-fan, the likes of which litter the Internet.

I respect the Register enough [why are you telling us? Ed], however, to respond on point to a representative set of the issues raised in his article...I'm not going to waste breath or time responding to him directly, especially not to someone who uses an AOL account.

First off, regarding the benchmark controversy.
Yes, it's a controversy.
However this doesn't make THG and Anandtech the bell-weather hardware sites on the Internet.

Tom has made a laughingstock of himself, everyone I know who has tracked his site over the last 5+ years knows this. Nice graphs, always a good source of benchmark data...that's about the usefulness of his site when it comes to x86 testing. Anand, when we want fresh litter for our kittyboxes, we print out the reams of garbage that he writes up on an irregular basis. Nice to know his thoughts have some good use (seriously, it is very good that both are branching out of straight x86 comparison testing, into hw realms where they can learn & teach to those of us who don't know. But knowledge <> intelligence and it most certainly <> objectivity). The writer even undermines his own placement of Anand at the top of the hw-site list with his comments about Content Creation 2003. If Anand based his opinions on the 3000+ on CC2003, then the author cannot with any credibility use his opinions on the merit of the 3000+ rating...unless of course he wants to say that means that it is probably even *more* worthy of a 3000+ rating than Anand thinks it is. And if so, then why should we care what Tom thinks. One or the other is right, they both can't be right. If he wants to say that industry benchmarks are tilted towards Intel and against AMD, then he can't put weight in hw testers that use them, i.e. Tom. If he still says that Tom is the cream of the crop, then his results must hold serious value. He cannot have it both ways.

In my opinon, Aces is the most-respectable site, when it comes to comparing processors, to evaluating performance. Those guys really know their stuff, and they know how to and will put a system through its paces. They don't just run Bapco, Sandra, Winstone or whatever is coming down from ZDNET. They'll write their own benches (as I do myself) to test the performance in great detail. Of course then they do have to get "real-world" performance on "real-world" code, so even they have to publish Bapco etc results on their site. So they do. But obviously they have a lot more info there than Tom or Anand. Unless all you care about is UT performance, or some other game. Which is legitimate as well, depends on your interests. But if you take that route then you throw out entirely the argument that AMD can rate their chips based on some two year old benchmark suite.

And that's where he really falls flat on his face, hard.

Ultimately the whole point of this "opinion piece" is to say that because of the weaknesses of the benchmark community, AMD is therefore justified in putting together their own benchmark suite and using it to rate their own processors.

How the hell can that make sense, except to a die-hard AMD-fan? How can he just spend 80,000 words criticizing the community for catering to Intel, then just accept AMDs choice of benchmarking results?

It's clear on its face, why that is. I won't even bother to say it, if you can't see it, then you WON'T see it. And that's the problem.

Having said that, my opinion on statements like the following become perfectly clear:

"This straw poll puts those that did not approve of the 3000+ rating in a rather small minority." {and the number is significant, especially given the large # of loud-mouth AMD-fans that supported it!}

"CC Winstone has changed over time to incorporate newer versions of applications, more video processing, larger data files, and in the 2003 release, Newtek's Lightwave 7.5 renderer, which is optimized for the Pentium 4's SSE2 but not for the SSE or 3DNow! instruction set extensions (both of which the Athlon XP supports). I'll let you draw your own conclusions about the merits and motivations behind the changes that ZD has wrought to CC Winstone over the years. " {...that software companies have updated their products to use the Intel SSE2 instructions, which increase performance, and so the benchmark suites have increased the percentage of SSE2-enhanced tests to reflect this? And this is bad? Should they only use x86/87 instructions? Only x86/x87/3dnow, as supported by the Palomino and Duron cores? What instruction sets would this AMD-fan like to see used in a benchmark suite, and what mix of instructions, what particular instructions, would this AMD-fan like to see used?}

"This is so reminiscent of what happened to BAPCo's SYSmark, where no credible explanation has yet been given for the changes that were made to that benchmark. If this was poker or craps, you'd have a very strong suspicion that someone was playing with marked cards or loaded dice." {vs dealing from the bottom of the deck, like AMD is, by rating their own CPUs with the results of tests they run themselves using a benchmark suite of their choice?}

I mean this goes on and on. He bitches and moans about Intel-favoritism in the industry, and uses that to justify AMD-favoritism. AMD is doing the same things, and even worse, by using "old" benchmarks based on "old" apps of its choice run on its choice of hardware (both Intel and AMD) by its own testers, to rate its own chips. But the most egregiously-odious comment, and the last one I'll make on this, is this:

"When so-called benchmarks fail the definition test, the onus should be on the benchmark house to prove that the contention is false. "

A most amazing piece of logic here, classic AMD-fan BS. He's already saying that it fails his little test. Now he wants the benchmark house to tilt against the windmill of their hatred of the benchmark? LOL what rational person would waste an iota of time trying to change an AMD-fans' mind about anything related to AMD or its products. But let us not apply the same logic to AMD and its PR rating, oh no. The problem is the PR rating, that AMD determines and applies to its own CPUs. If it were not for that rating then the spotlight *would* be on the hardware community and the benchmark community, and he is certainly raising valid questions. But when AMD takes the game, puts it in its own house, referees it, and refuses to make any changes other than what AMD itself wants to make, then how can this rating be a valid mark of performance on applications from an industry which is, frankly, moving forward, in directions NOT controlled and determined by AMD, a company with a shrinking share of the X86 market? This is how we get into situations where AMD disables hyperthreading on Intel cpus, because, they say, "it runs better that way". Gee: could that be because, for at least *one* reason, the code you're using to test it with is two years old? How can this even be in AMDs best-interests? The only way it *can* be is if they develop a following that rejects anything that makes AMD look bad and Intel look good...followers who write "opinion pieces" like, hm, like this one!

This whole article is an excellent example of AMD-fan logic. For that, I appreciate the effort put into the writing and posting of this article. I now have a handy on-line reference when I want to explain "AMD-fan logic" again, always written with the short-sightedness that leads me to think that they are only thinking about today, never what happened before or what will probably happen in the future. Much like Tom did when he predicted the demise of Intel just before the Athlon was released. And last but not least, this is why we need the hw community, more than ever, to put the heat to AMDs feet when they label their chips with some "performance"-based rating. You will note that they've already abandoned that approach with Opteron, showing just how credible they think it is, themselves.

Now if we could just get them to stop with the BS-filled PR and release their chips in accordance with their roadmaps (not the latest ones, but the ones they issued last year or the year before, when the PR started), we'd be o.k.

Thanks for your time.
anon

Subject: Did AMD's Athlon XP 3000+ earn its rating?

Regarding the subject article, I wonder why many people expect AMD to change its PR system every time INTC changes its processor. On the contrary, I would think a PR system would be much more valuable and consistent if set out a methodology and stuck to it. IIRC, AMD claims its PR rating is relative to its own Athlon processor (not the P4), but even if we don't buy this, I think it's clear that AMDs ratings were very conservative when comparing to P4 pre-Northwood, about right when comparing to Northwood, and perhaps a little behind when comparing with HT-enabled P4s. This does not mean the PR is inconsistent. It just means (again) that clock speed is not the whole story for performance. It is the IPC performance of the P4 that is inconsistent.

Eric Piquette

When it comes down to it, big companies buy systems based on (in this order) (1) their own internal standards (2) the likelihood that the company they are buying from stays in business and supports the platform forever (3) software support on the platform (4) performance and cost. That's why IBM, Sun and HP (in that order) have a chance at all with their high-end systems. No one cares that much, for example, that Sun is slower on many tests, because Sparc/Solaris (1) is already used in their datacenters (2) is not going out of business (could get bought though?) and is still supporting the same boring stuff (3) has much more software available than Power/AIX and Itanium/HPUX (4) is basically in the same range for performance and cost. None of these systems is bought at list price, or individually … most sales are many-server, multi-year agreements. Sun just happens to have the unenviable position of being the current whipping boy, and companies are enjoying participating, because Sun was so arrogant and unresponsive on the hardware side for so long. IBM might be twice as fast on a per-CPU benchmark, but IBM is more than twice as expensive too. So far, Itanium hasn't proven to be competitive in the sector, but with the HP server group behind it, it stands a good chance at eventually becoming number three, and with Intel behind it, it even has a long-shot chance at displacing IBM and Sun. However, if Intel flinches on the 64-bit question (e.g. adds 64-bit support to Xeon), Itanium is dead in a minute, because it will become obvious that it will never gain economy of scale nor the necessary platform support, and IBM and Sun will generate more instant FUD than . So IBM and Sun have a temporary friend - AMD - because AMD may force Intel to deploy a rear guard (64-bit Xeon) to protect their volume (and profit) line. Any way you look at it, "little iron" (2-8 CPU servers) just got a lot more affordable as a result. I've attached an email from Appro to give you an idea of how much more affordable … the Sun equivalent would be US$30M and the IBM equivalent would be US$60M. As they say in the south, "no shit."

Email address supplied

Subject: Did AMD's Athlon XP 3000+ earn its rating?

with the release of the Athlon XP "Barton" 3200+ the discussion about AMD's performance rating has reappeared. I do not understand this discussion anymore, what do people like the Papster want? Why should AMD rate this chip 2800+? They could have named the chip even XP 4000+ or some other name. P4 and Athlon do not share the same architecture and everybody agrees on that, so Performance Rating and Clockrate is not any element of performance comparision anymore.

Tom and the other testers should imagine that AMD makes a K9 Chip with a 50 stage Pipeline running at real 8 GHz, but performance clock by clock very bad. How should a 8 GHz Chip be named? AMD K9 8 GHz, AMD K9 3000+? Who will cry FOUL??? They all should have learned by now, that GHz or PR is not a way to measure REAL performance. What should we do with the Pentium-M, Crusoe, PowerPC, Alpha and so on? Many of them are slower in some benchmarks than the P4 3.0 GHz and in others they kill it. AMD had to adapt some other naming scheme to make abstraction of the Megahurtz because customers are so stupid and Intel did profit from this. Lets see it from the other side, the new XP FSB400 seems to be around 6-7% faster than the XP 3000+ that makes a 3200+ number to rest logical in the naming scheme. At least the self-elected educators from the hardware sites should use their brain and not fall in this typical megahurtz and benchmarketing war. Nobody will critizise Audi for naming the A4 2.8 even when it is slower than a BMW 325i...

Oliver Ratzeburg

The Pro Intel sites (THG, etc.) are easy to spot as they don't cover the Itanic bug.

Mark Rainwater

If Andrew Cubitt hired a farmer to appear in his commercials, would the
farmer be Cubitt's Rube?

Then again, if he hired a man named Rick to play the farmer, and painted him red (a la La Intella's blue goons), then the character could be Cubitt's Rube Rick, a pun of another color.

:)

John

If this had happened during the summer, this guy could have died.

What have we learned from this experience? By no means should we allow Windows CE into any aircraft whatsoever. If car passengers are trapped when Windows CE crashes, you can be pretty sure that aircraft passengers would die in the event of a software crash.

Michaela Stephens

Subject: cordless review - logitech / where is BLUETOOTH?!

how could anyone in this day & age get an article by their copy editor without having Bluetooth as a CHECK-OFF feature (yes/no)?!

No body is going to start buying 10 wireless devices that each a proproetary RF transciever! ... the gaggle of USB basestations would be just as much clutter as all the wires they were supposed to replace!

plaese make sure in the future that B/T is a standard element mentioned in the feature list of all wireless products (present / missing).

thanx: david

Subject: Something strange about Windows XP (Believe it or not...) and Linux...

Ok I know this may seem weird but here's the scenario;

I was having problems with Windows XP and my Hauppauge WinTV (PCI) TV Tuner... system would reboot when I would try to restart it... if I would remove it and re-install it it would only work of one Windows session... Re-install time again. Anyways I just downloaded Linux Mandrake 9.1 (Awsome...) and decided that I'll dual boot again so I said I know Windows won't creat a Boot menu for linux (I only wish but I think pigs will fly first) so I should install Windows first then Linux. After installing Windows and applying SP1a (didn't install any drivers for the cards in my machine, which was my first mistake) and then install Linux Mandrake 9.1 both worked fine. Once Linux was done I decided to go back to Windows XP and finish the install.... need less to say I haven't installed my SB Audigy 2 drivers in a while and forgot that you need to install it from CD (I had updated drivers on my HDD and they wouldn't go unless you used the CD first...), this resulted in getting a high pitched whine from my speaker (I tried to force the drivers in.... not to self never do that again :) ).

My first reaction was something was wrong with my hardware (but that wasn't it... just needed to use Audigy CD), so two hours of troubleshooting I gave up and decided to re-install Windows since there might have been something stupid I did while doing the drivers (remember I've been installing windows/linux for about 8 years now). So this is where it gets weird, I started to boot to the CD and that worked great no problems, got to choose my partitions... Primary partition 8GB Windows XP, in the extended partition I had 20GB Backups, then 24GB Games, 8GB for Linux (3.5GB=/, 512MB=Swap, 4GB=/home) and remember I didn't remove the linux partitions at the end of the extended partition. That portion of the install worked great got to the formating portion, that worked great, then to the copying of files to the hard drive to continue the setup and configuration protion of the installation but about half way through I get this blue screen stop error telling my that my BIOS is non-ACPI compliant (WTH, I've installed XP on many systems and never seen that before). Anyway I though nothing of it, might have been that the CD was scrached or something stupid like that (even though that still made no sense) I tried it 6 more times, same error every time. One day goes by and I said to my self WTH I'll nuke the Windows and Linux partitions (still keeping my precious Backups and Games partitions) and guess what no problems, using the same CD same hardware and everything... WEIRD!!! isn't it. I guess M$ is really trying to get at linux, I don't know exactly what happended, it may have been the EXT3 partition format, could have been multiple journaling partitons.

Notty Notty MicroCrap quite palying your games... haven't you been found guilty of monopolizing the industry.... Wow funny how the legal system does nothing to those who abuse it the most... How in hell does the Governments of the world expect people to obey laws when they are not blind nor stupid.... I guess that's my 2 cents.

Dan Bastianello

Subject: Logitech tightens purse strings

using your news centre to ask for a free mouse denotes a complete lack of professionality.

Cu

Arron replies
I have something like 8 other mice here, whether one goes back or not matters not a jot to me. Logitech's behaviour is just very odd indeed.

Arron

Subject: INQUIRER launches ad-free subscription site

this is my fav site............ NUMBRE ONE just after yahoo.....
(I let Yahoo be the HighOverHead Land-Fill for the interesting CreamSkimming Architecture of The Inq)

now look guys (includes men and women) ...... I know you need those ads ..... and I live with that ..... I don't even see them. ok now I do since you mentioned them......I just put up 'the shields' and enjoy it no pop-ups....etc

but it seems to me you wont sell that many subscriptions...?...$12..?... at that rate I pay about $85/year for umlimited access internet....all the sites....dial-up....

I subscribe to NOTHING..... would like to get Salon, etc....

but this is just too much...... I disconnect cable tv in the summer months........

I NEVER CLICK ON ANY ADS..... that just means to me that I will get spam

so that is my market segment.

BUT......WHAT IF:

I base the following on assumption that you are solvent with ad-based site and currently there are no subscriptions....

WHAT IF you had subscriptions that were so small in price that even me......
a guy who does not subscribe to things .... BUT WHO WANTS TO READ EVERY DAY ... could afford it and not notice the $

sort of a micro-purchase .... say $1/month as an example ...
I read it anyway........... but enjoy it enough to pay a small fee for a slightly cleaner version.....

it could take-off ...once a migration starts and word gets out on the good stuff at The Inq you will have some volume..

I think you would have a lot of subscriptions...... maybe overcoming the $12/mo - lower subscription revenue

I guess that means a different TIER .....

just an idea.

(and yes ... I am no expert....
I don't sit and ponder the economics of mass market web service subscription all day
so I know very little... )

I will keep reading and keep thinking I would pay a little bit to have NO ADS....I wouldn't even miss the $$ (= 2 bottles of cheap wine a year.)

ANYWAY I already block them all with my Ultra MultiDistributed SeriesParallelProcessor NeuralNet PhasedArray, SalineCooled, AdaptiveProcessing Super Computer ................... My BRAIN (and it is a small 57 year old model)

again...my fav site
Tom

I read the INQ every day, and try to do my bit by periodically clicking on ads. Although the ads can be annoying at times, I happily tolerate them, recognizing the quality product I'm getting free of charge would be impossible without them.

For those that can afford it, the pay site seems a good idea, and I'm happy to see the INQ add a revenue stream to help fund the quality work that's done there. Being a graduate student, and broke as a joke, I won't be signing on, but I hope others do. The only thing I find dismaying is the mention of making some content exclusive to the pay site. This seems to me a deliberate degradation of the free site to encourage subscription. I don't see another motivation-- how could it be cheaper to maintain two sets of content rather than one?. There should be plenty of people out there willing to cough up a few bucks (quid?) without the INQ having to resort to those kinds of tactics. Just my two cents (pence?). Thanks for an informative and entertaining outlet,

Austin Albino

µ