Are Tom's Reviews Biased?

This has probably been discussed before. I will just say that I agree with you. Because of the nature of this thread, and the probable flames that will accompany it, I will just leave it at that.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =

I am completely neutral when it comes to processors and I would say that you should simply ignore the comments made and concentrate only on the graphs. The comments are completely irrelevant.

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor

P.S. if your an avid overclocker, get the 1.6A or 1.8A instead... you will achieve a much more rewarding overclock and have to pay far less to do so.

"I came, I saw, I overclocked", Julius 'Smokin CPU' Caesar

We just make an informed decision and build what we can afford.
Best buy for the buck.

<font color=blue>You get what you pay for.... All advice here is free.</font color=blue>

While I have noticed Tom Pabst turning pro-AMD especially after all of the flak he caught from Intel on his RDRAM (i820) and P3 1133 exposés. I would too if I were under that kind of fire.

But, Frank Völkel in his latest articles (without the Tom byline) is pulling the same things pro-Intel, especially today's <A HREF="http://www.tomshardware.com/cpu/02q1/020225/index.html" target="_new">article on the P4 533MHz bus</A>. It seems obvious to me that to be fair he would have had to put the 133MHz P4s up against 166MHz Athlons. This would be easy through additional overclocking and 166 FSB Tbreds have been hinted at (but not officially) by AMD.

In addition, he consistenlty labels RDRAM as fast and DDR as slow. While his table indicates the reality; shipping DDR sticks are faster than shipping and upcoming RDRAM, he hardly makes passing mention of the fact that RDRAM platforms for P4 perform better only because they have the benefit of dual-channel.

My point? Well, Tom's can be infuriating when your team doesn't win the latest review praise. Sometimes it can even make you change teams - score one for truth. Overall, I think Tom's is the best review site on the net, but I still want a second, third, forth, fifth...opinion.

I thought a thought, but the thought I thought wasn't the thought I thought I had thought.

I don't quite get what you're driving at. If you are being sarcastic, I would agree that RDRAM should not be at a disadvantage due soley to its 16bit/32bit dual-channel nature; it should be able to ramp to higher bandwidth quicker than DDR, but the fact remains that it hasn't. I don't think it will. Check out <A HREF="http://www.anandtech.com/chipsets/showdoc.html?i=1588" target="_new">Anand's latest article</A> on the new E7500 dual-channel DDR chipset for the P4 Xeon proc. If RDRAM were better and faster, why would Intel pick DDR for its mission critical servers? Maybe the wool will finally be lifted?

Oh, and I agree about the rest (except that quite a feat biased part...It would only have to run at 3400 to rip that 5GHz P4 to shreds - and will - the clawhammer).

Flame war here we come...who will be the first to call the other moron, idiot, etc.?

I thought a thought, but the thought I thought wasn't the thought I thought I had thought.<P ID="edit"><FONT SIZE=-1><EM>Edited by ath0mps0 on 02/25/02 10:43 AM.</EM></FONT></P>

I just want to add my 2 cents.

From my observations, the comments (and often times the conclusions as well) in THG reviews are often biased towards the underdog, just as sheps observed. Even at the best of times there still always seems to be a biased undertone.

Granted, it may be subconscious. It may not be intentional. THG staff may not even be aware of it. However, myself and plenty of other readers have noticed it, so there must be some truth to it.

In my not-so-humble opinion, personal opinions and biases don't belong in reviews. I see it as a matter of professionalism. (Something which THG has at times lacked.)

ath0mps0, I have to say that your comment about to be fair the review would have compared the new P4s to a 166MHz bus Athlon, well, comical. Has THG recieved an engineering sample from AMD set to a 166MHz bus? THG reviwed an AMD platform using the KT333 chipset and PC2700 memory. All considered, it was the best they could offer. THG doesn't compare vaporware to actual hardware.

Further, your statement of, "In addition, he consistenlty labels RDRAM as fast and DDR as slow.", has no more forethought than the previous comment. His conclusions are clearly drawn for two reasons based on comments in the review. 1) The performance gain seen from new RDRAM systems over new DDR SDRAM systems is quite evident. (While I admit that making such a conclusion is a bit premature given that THG has not successfully run PC2700 at CAS2, it is at least based on what hardware is currently capable of, which is all that anyone can ask. Anything else is just theory.) 2) The P4 clearly performs better when running RDRAM. This is due to synchonous clocks and bandwidth available through chipset/motherboard support.

Then ath0mps0, you continue to show an unusual penchance for tunnel-vision by asking, "If RDRAM were better and faster, why would Intel pick DDR for its mission critical servers", after poiting out a review on a new Xeon chipset. The simple answer is that Intel started work on an RDRAM solution FIRST and DDR second. Besides, just because something is new doesn't mean that it is the only or the best.

FatBurger, you make a good point that there will always be those who consider THG biased just because they review the best-of-the-best as being the best and these people can't accept that their side has just lost face.

However, even when THG reviews Intel as the new performance leader, there is still an air of AMD favoratism. Yet when THG reviews AMD as the new performance leader, AMD 'crushes' Intel. It seems to me that it there were no overall bias, then clear-cut winners would ALWAYS be declaired as such. No more, no less.

<pre><b><font color=orange>AROOOOOOOOOOOOOOOOOOOOOOO!</font color=orange></b></pre><p>

Heh heh. LOL @ FatBurger.

Yes everyone at Rambus does appear to be an idiot.

A more truthful statement though would be simply how many hours and dollars have gone into DDR SDRAM research and development compared to how many have gone into RDRAM r and d. RDRAM hasn't ramped up because nowhere near the resources are going into it compared to DDR SDRAM. (In fact, for as much as goes into DDR SDRAM, you would think that we should have quad-rate x 200MHz modules by now.)

<pre><b><font color=orange>AROOOOOOOOOOOOOOOOOOOOOOO!</font color=orange></b></pre><p>

I wonder.....
You say its insane to think a 166mhz bus XP should be compared to the new 133/533 intel, and you are partially correct. However you also mention that there is no vaporware review, but there is! Intels 133/533 setup wont be available for 2 more quarters! By that time there will be new AMD chips and setups so I dont think this article relly means much yet.

As for RDRAM, I dont think there is a real comparison between it and DDR cuz there is no dual channel DDR for P4. There is also no RDRAM for AMD. If there was either we could see a true comparison, but as it is we are comparing apples to oranges.

I myself think that each person should see what they are doing on the machine and look at benches of those apps. Once you know what you need just buy the best value. I does not matter if you "only" need a 1ghz pc or the fastest available, the price/performance in what you use is the most important, esp keeping in mind that 90% of users wont notice the diff between a 2 ghz (or 2000+) CPU and a 1ghz in thier browsing and office apps =/

Jesus saves, but Mario scores!!!

I didn't find anywhere in the Tom's article that stated that they had received engineering samples of 133MHz P4s; in fact (apart from the dramatic license at the beginning) the voltage overclocking exposé led me to believe that they had just overclocked NWs and used PC1066 RDRAM. Oh, and Tom's does often benchmark overclocked CPU's To support his claims that "Our benchmark results clearly prove that if Intel changes the FSB and memory clocks (to 133 MHz and 533 MHz, respectively), this will put it quite a distance ahead of its competition from AMD, as well as its own series of processors." Frank should have applied this same logic and overclocked equivalent AXPs.

Correct. The reason I object to Frank's handling of this topic is that he consistently points to the type of RAM as the determining factor of performance while the actual root of performance is the chipset and synchronous clocks as you stated.

Actually what you call tunnel-vision is just being well informed and knowing some facts ahead of time; Intel let its channel partners know that they would not be supporting RDRAM on their server chipsets about the time the i840 was released. The two primary reasons were 1) Price of RDRAM - although now invalid, as anand states in his review of the chipset, 4-16GB would have been impossible for all but the largest organizations. 2) Memory Latency - with RDRAMs already high latency, it about doubles for every stick you add - it is a serial channel. They would have had to implement quad or octa-channel to overcome this issue that is an extreme factor in large active databases. IMO, they only stuck with RDRAM for the P4 i850 desktop chipset due to their contract with Rambus and to keep their hands on those, now practicaly worthless, Rambus stock warrants. In addition, they had done all of their development on the RDRAM platform and it would have delayed the lauch of the P4 - not that that would have hurt them too badly - the high cost and poor performance of early P4s was enough to convert many to Athlon.

In MY not-so-humble opinion, I think that is what I said....

All in all, I only gave these examples (in my previous posts) to illustrate the contra-point to the initial article. Yes, Tom's Hardware often portrays a biased view from multiple perspectives - something that would be eradicated in a purely objective environment - the "perfect review site." I think that Tom's provides the closest thing I've found to this - but $.05 is closer to $1.00 than $.01....

I'll be the first to admit my bias for AMD; this has developed over time (from my first 80286 that had Intel and AMD stamped on the chip). After the first P-ratings failure I became a staunch Intel-only proponent. It was actually Intel's false and frankly, inept, handling of the i820 fiasco that drove me to AMD.

Now I am biding my time waiting for Intel to finally shrug off the Rambus monkey (the company) and use whatever technology is best for the consumer, not just best for the shareholder.

I thought a thought, but the thought I thought wasn't the thought I thought I had thought.

My AXP 1600+ @ 1750MHz w/166MHz FSB (~2133+) does exist!!! And it is fast.

I thought a thought, but the thought I thought wasn't the thought I thought I had thought.

That's an overclocked chip, though. You can't really review overclocked products and say that they're the same as the retail version.

<font color=orange>Quarter</font color=orange> <font color=blue>Pounder</font color=blue> <font color=orange>Inside</font color=orange>

OK, I'll admit that they probably have an actual factory sample. But, the factory sample is not a retail version either. AMD has also indicated that they will go to the faster FSB when it makes sense. Just like Intel can't officially do 533 until PC1066 is available in quantity and for reasonable prices - the RDRAM killer of the past, AMD won't *force* a move to DDR333 until it is available in higher quantity. That doesn't mean I can't unlock my chip and crank up the FSB to verify Tom's reviews.

I thought a thought, but the thought I thought wasn't the thought I thought I had thought.

Other than when they say, "we have the new P4 platform with 533 MHz Rambus memory", or when they show the picture of their engineering sample CPU <A HREF="http://www.tomshardware.com/cpu/02q1/020225/images/cpu.jpg" target="_new">http://www.tomshardware.com/cpu/02q1/020225/images/cpu.jpg</A> If they had just simply overclocked a CPU, they would have said so.

And when THG does this, they make it blaringly obvious that what they are doing is overclocking.

I won't debate Frank's dubious lack of clarification on the subject. However, since no platform yet exists which proves DDR SDRAM can completely outperform RDRAM, it becomes a moot point as to why a complete system designed around one memory type is better than another if all that the consumer needs to concern themself with still is the actual memory type.

That is just your opinion though. The truth is, had the world adopted RDRAM better and put even 1/4th the man-hours and dollars into research and development for RDRAM as was put into DDR SDRAM, then right now you'd be singing a completely different tune. What Intel did was take a gamble with a technology that could have significantly outperformed SDRAM's capabilities. But gambling is gambling. Sometimes you lose. Rambus went and ruined everything by being sneaky under-handed greedy jerks and what could have been a promising technology is now barely able to scrape up enough r and d to keep on par with the rather signal-noise touchy SDRAM.

It won't matter for too much longer though as sooner or later someone is going to completely replace both RDRAM and SDRAM with something better.

Indeed. Intel really screwed themselves there. Had they kept RDRAM out of their systems until the launch of the P4 and actually waited to launch the P4 until they could manufacture it even somewhat close to it's original specs, Intel would have done much better for themselves. But they let the pressure from AMD push them into making one mistake after another, and they're still paying for those mistakes. AMD certainly caught Intel off-guard for a while there.

And what would you do if in the end the best technology for the consumer turned out to be based on something from Rambus? What if there was proof that Intel's actions weren't purely motivated by the shareholder, but by the potential for technological superiority? The simple fact that you don't aknowledge these possabilities says a lot for just how open your mind really is to the possabilities of the future.

<pre><b><font color=orange>AROOOOOOOOOOOOOOOOOOOOOOO!</font color=orange></b></pre><p>

In case you weren't aware, cache running at full CPU clock is not the only requirement to clock the FSB higher. All of the components in the CPU (ALU, FPU, etc.) have to be capable of being harmonious at the new FSB frequency, or else the signals start coming in other than just at the peak and trough points of the wave and by the time the components can respond, the signal is garbled. Not every owner of the latest AMD chip can overclock to a 166MHz FSB flawlessly.

Thanks for the good laugh. Do you honestly believe that <i>if</i> AMD were capable of reliably kicking out stable 3-year waranteed 166MHz FSBed CPUs they would in any way hesitate to do so just because of the higher price of PC2700 SDRAM? Do you really believe that there aren't enough people and businesses out there demanding a higher-performance AMD system who don't want to void any warantees by overclocking for AMD to rake in tons of cash with an official 166MHz CPU revision?

AMD is <i>clearly</i> starting to lose it's hold as king of the hill in the x86 CPU performance world. The shock value that they gave us back when Intel was fumbling around with the limits of the P3 is over. Intel may have been cought off-guard for a while, but they're coming back strong now. AMD needs an upgrade, and they need it bad.

The fact that AMD isn't giving customers any serious upgrade while Intel continues to hype up a storm about their next P4 revisions and releases higher and higher clocked CPUs that are starting to beat the Athlon's even without overclocking says a LOT about AMD's current state of affairs.

My best educated guess is that AMD is clearly dumping close to all of their efforts into their Hammer line, and if this line doesn't turn out to be better than anything Intel can offer, AMD is going to go back to being a third-rate budget CPU manufacturer like VIA currently is. The Athlon line, while having an excellent FPU, is going to start reaching their clock speed limitations soon. If Intel's flagship were the P3, AMD would have plenty of room to breath. However, the P4 is designed to be a clock-speed monster, and it's a lot further from it's limits. AMD needs something new and exciting soon, or they're going to get bowled over, which would be a shame.

<pre><b><font color=orange>AROOOOOOOOOOOOOOOOOOOOOOO!</font color=orange></b></pre><p>

Of course it isn't, but up until the TBird it was the limiting factor. I didn't say you could overclock the FSB to the full CPU clock.

One of the specific design elements of Intel and AMD processor is that their internal timings are set by two factors: 1)the FSB 2)the clock multiplier. All internal circuits use these two factors to determine their funtional timings. ALUs and FPUs run at full clock, so external FSB has no effect. It is the circuitry external to the actual chip that is the limiting factor.

Agreed - only because they would have to know exactly what they were doing and unlock the mutliplier to lower it. Otherwise the chips are all capable. OK, every one I've seen, and that's a lot.

I thought a thought, but the thought I thought wasn't the thought I thought I had thought.

So then you are simply biased against the company. How can you claim to judge memory technology in an unbiased manner with this kind of attitude?

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =

Damn straight I am biased against Rambus - as I have stated several times and illustrated my reasons in previous posts. Things like fraudulent (adjudicated in two separate courts) dealings with JEDEC and the entire computer memory industry designed to force me (and you) to buy RDRAM or pay more for DDR. While you may enjoy the occasional S&M action, I like to make my own choices and I don't like to pay royalties.

When was the last time you bought or rented a betamax video? Sony's high royalties virtually destroyed an arguably better platform. Now just about everyone has royalty free (or nearly) VHS. This is just another example of what royalties do to a mass market product.

The memory & chipset manus are hard at work trying to get an new, royalty free, product out the door. Yes, it could be argued that RDRAM forced the memory manus to step-up R&D on DDR, and for that I thank Rambus.

I never claimed to judge it in an unbiased manner. Yet, I feel I do judge technology from an objective point of view; I just try include all the facts, not just the spin shoved at me by one vendor or the other. And, any new facts and enlightenment are welcome.

I thought a thought, but the thought I thought wasn't the thought I thought I had thought.

1GB memory modules are usually available to the motherboard manufacturers before support for them can be incorporated. While it may (or may not, as I have not bothered to check) be true that the motherboards were released before the memory modules went public, I can assure you that the 1GB memory modules physically existed before any motherboards that supported them.

Be patient. RDRAM has not been out for very long compared to DDR-SDRAM.

Since when has memory bandwidth been measured in terms of modules? The industry standard is to measure bandwidth per pin. It is the pin that takes up real-estate on your motherboard. You really want the most bandwidth per pin that you can obtain from your memory. You can always stick more pins on the motherboard but you cannot always crank up the bandwidth on each pin. That is much more difficult. Current PC800 RDRAM offers 100MB/s of memory bandwidth per pin. PC1066 will offer 133.3MB/s per pin. In comparison, current DDR333 modules offer 41.6MB/s of memory bandwidth per pin.

Adding more pins is not that difficult and is in fact on the roadmap. RDRAM modules are scheduled to move from 16 pins up to 32 pins later this year and finally to 64 pins. This will eventually make it equal in pin count to DDR-SDRAM modules. At that point, the memory bandwidth differential between RDRAM and DDR-SDRAM will be much more obvious.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =

Im sure the 533fsb p4 they used was designed from the ground up to be a 533fsb chip, and not just an overclocked 400fsb p4, sure, right.

They are both vaporware IMO.

"The Cash Left In My Pocket,The BEST Benchmark"
No Overclock+stock hsf=GOOD!

Burger, do you really think that there is any difference between a 533fsb NW and a 400fsb NW other than the default multiplier?

They do not have a seperate line for 400's and 533's just like the tbird did not have a seperate line for 200 and 266mhz fsb's.

Its a simple matter of lowering the multi and raising the fsb, hardly even an engineering sample!

"The Cash Left In My Pocket,The BEST Benchmark"
No Overclock+stock hsf=GOOD!

This does not mean its an 533mhz fsb engineering sample, its probably an unlocked NW, to which they could set any multiplier and run it at 533fsb.

"The Cash Left In My Pocket,The BEST Benchmark"
No Overclock+stock hsf=GOOD!

They would not not release it due to the cost of good ddrsdram, they would not release it to avoid eating their future markets, its the same reason they havent released a 2100+ yet, because the speed increase is NOT needed, but if you support a dual 133 and 166 fsb line, you wind up with wasted and unsold parts and more confusion in the marketplace.

The tbird B and C had to be done to make way for pc2100, but how many tbird B's did they make which never will be sold due to the lower price of the c model.

All tbirds are produced from the same line, and it is extremely rare to find one which wont do 166fsb, or even higher, if the clock multiplier is kept low within cpu speccs.

You are misinformed silver.

Intels fastest chip is the NW 2.2, it is just a hair faster than the axp2000+ for twice as much, AMD is in a GREAT position cpu wise right now, when the next NW comes out, amd will release its next chip, there is no reason for amd to rush ANYTHING.

The only thing the situation says about amd is that they are good buisnessmen, and they are selling alot of 2000+'s at a premium, and they will sell alot of 2100+s and alot of tbreds, Intel has done NOTHING which should make them afraid, make them rush out faster chips, lowering prices on all the other chips, losing them money.

If amd released a 2200+ 166fsb chip, they would have to charge an arm and a leg for it to make up for the lowered prices of all the rest of their line, which would negate one of their main points over intel....price!

"The Cash Left In My Pocket,The BEST Benchmark"
No Overclock+stock hsf=GOOD!

BRAVO, EXCELLENTÉ!!
Couldn't say it better Mat, you are indeed the best when it comes to refuting agressively, if I may say so! I can always count on ya to back some people's opinions with AMD, which are true.
Hoobooy Silver I can tell you from now, YOU DO NOT WANT to start an argument fight with Matty, just see how he and Ray do theirs! Both are extremly informed people and refute extremly well, so you have to come packed! Hell I hope this isn't gonna turn up to another memory argument, as Ray and Athompson popped! Uh oh...

--
For the first time, Hookers are hooked on Phonics!!

Matisaro, while I will grant you that you have a good mind, you still leave some holes in your arguments. Forgive me if I start out sarcastic, but your use of misinformation deserves a humerous response.

Because it isn't like AMD <i>ever</i> did anything like release the AthlonB (100MHz) and AthlonC (133MHz) version of the Thunderbird core.

Except for that tiny matter of the connected bridges to unlock the multiplier. Completely indistringuishable. No one could ever tell just by looking at it.

Again, because it isn't like AMD has ever run two different FSB chips simultaniously, resulting in confusion. Of course, I find the unsold parts on AMD's end difficult to believe. Besides the fact that they only need to stop producing the lower FSB chip in order to not end up with excess unsold CPUs, there is also the simple fact that <i>most</i> of the people who buy AMD chips, do so for their overclocking potential. And if, as you suggest, <i>nearly every</i> Athlon will run stable at a 166MHz FSB, then tons of people will be buying the cheaper and older 133MHz FSB chips, just as they bought up tons of AthlonBs when the AthlonC was released.

I never argued against this. I know that most can have a FSB that high. However, I also know that in a small (but noticable) number of cases, this actually degrades system stability. And as of yet, no one has ever proved that in most cases doing so will still keep the CPU running flawlessly 3 years from now. Yes, you can overclock the FSB. That isn't what I debate. I debate the commonality of the CPU's stability and longevity after having done so.

Right. Because AMD never did anything like change the Athlon naming convention to compete with the Pentium 4. Their line never went Athlon(A), AthlonB, AthlonC, Athlon4. Certainly AMD never incorporated a debatable CPU rating naming convention which is based on performance compared to the Pentium 4 and NOT on the actual clock speed. And it is not like AMD has ever announced a price drop immediately after Intel announces one. No, you're right. AMD had <i>never</i> shown any kind of fear of Intel whatsoever.

If AMD released a 2200+ 166MHz FSB chip, they <i>could</i> charge an arm and a leg for it and still be cheaper than the 2.2GHz P4. Right now (according to Pricewatch) AMD performance-rated CPUs are on average 79.87610069% cheaper than Intel clock-rated CPUs. With the P4 2.2GHz pricing at $467, this would allow AMD to price a 2200+ 166MHz FSB Athlon at $373.02 to retain their price competitive ratio. I don't think that anything there sounds unfeasable or unreasonable at all. In my opinion, AMD has no <i>good</i> reason not to <i>if</i> they can mass-produce said product <i>and</i> provide their typical warantee on the retail-sold versions.

To me, the fact that AMD isn't doing anything of the sort currently says a lot about the possabilities that they cannot actually achive one of these stipulations that makes the idea reasonable.

And worse, the fact that VIA is beating AMD in innovations by supporting PC2700 DDR SDRAM in their KT333 chipset is frightening. If the whole motherboard can now be running with a 166MHz bus, then why isn't AMD producing a CPU that will actually support this bus? To not do so verges on insanity. There is more than enough market waiting (nearly screaming) for it.

The Hammers are aimed at home/small-office servers, not at typical home users. As such, expect a big pricetag. (At least big for AMD's standards.)

Which sounds about right, but is achived through a reduction in etching size, which Intel can also use to expand the speeds of their 'Celerons' and P4s. (I laugh at what they call a Celeron these days though, since it's nothing more than a hacked-up P3 to ensure performance worse than a P4.) This won't keep AMD's CPUs competetive because Intel will be doing the same thing to their CPUs.

It's the same reason that Intel stopped trying to compete against the Athlon with the Coppermine. The core design simple couldn't be pushed any further to remain competetive. Now Intel has a core design that can be pushed even further than AMD's core. To remain competetive, AMD will have to actually significantly improve their core, not just reduce the etching size. And from all sources of information on AMD, it appears that they're gambling on such a concept using their Hammers. Maybe it will work. Maybe it won't. Only time will tell. However, you have to admit that if the gamble fails, it will put AMD into a very unpleasant situation.

I never said that Intel was close to doing so. I said that AMD is taking risks and putting themselves into a position where if things don't go perfectly for them, they will be in a position to be easily taken out. And as of right now, AMD is only just hanging in. They haven't 'wowed' us with any new innovations in a long time.

To end on a positive note, I do agree with you here. This year is going to be a very bright year for computers in general. We're finally going to see some decent memory speed increases. We're finally going to see some FSB speed increases. Even if CPUs don't get clocked any faster, this in itself will mean a significant performance gain for us. Plus, USB 2.0 is finally becoming a motherboard standard. Microsoft won't be able to ignore USB 2.0 for much longer. Now if we could just see some of the PCI improvements become more common on typical home-user hardware (instead of just for server use), we'd really be rocking. I just hope that the gambles that AMD are taking now don't turn out as badly for them as the gambles that Intel took with RDRAM and the initial P4 did.

<pre><b><font color=orange>AROOOOOOOOOOOOOOOOOOOOOOO!</font color=orange></b></pre><p>