The issue with being an Infallible GURU

[[XC] Oj101]

Good people.​

I come to you with a great concern. One that speaks directly to what I believe and will show below to have been plaguing the tech review industry for years. Like all ills, this speaks to a minority of websites, as the vast majority at least in the capacity of content creators, are upright, transparent and reliable. However, something has caught my attention and I would appreciate your sober and objective opinion on it. I must say first, that I come to you because; believe it or not - you are the keepers, watchers of the entire industry. Your consistent effort, drive and intense competitive spirit is what keeps you at the bleeding edge of desktop computing technology. In essence, competitive overclocking is the Formula 1 of consumer computing as I have many times written. At any point where a claim is made, one has to balance that against what one sees here on HWBOT. Results that have been scrutinized and verified by many and can for the most part be confirmed and repeated. Before I go further, please note that I have explicit trust in the scientific method and its rigorous methods of inquiry. Any theory, philosophy or claim, must make reliable predictions, offer a coherent picture, have repeatable results and most importantly must be falsifiable and open to scrutiny. So here goes.

1. On the 11th of June 2015, the website, GURU3D.com, published a review of the following graphics card, the GIGABYTE GV-N98TG1 GAMING-6GD, or better stated, the GeForce GTX 980Ti G1.GAMING. That review can be found here Gigabyte GeForce GTX 980 Ti G1 Gaming SOC Review - Introduction

2. The test system upon which the results were generated are briefly stated as:

• MSI X99S XPower AC
• Intel Core i7 5960X @ 4.4GHz (on all eight cores)
• GeForce GTX 980Ti - Gigabyte G1 Gaming (In "OC Mode" as per OC GURU Software setting)
• 16GB (4x 4096MB) 2133MHz DDR4
• CORSAIR AX1200i PSU
• Windows 8.1 64-bit
• DirectX 9/10/11/12 End User Runtime
• NVIDIA GeForce series latest 353.12

The above may be found on Page 12 Gigabyte GeForce GTX 980 Ti G1 Gaming SOC Review - Test Environment & Equipment of the review

3. The author supposedly recorded these results in both synthetic and game play benchmarks. I am going to focus on the synthetic tests, mainly 3DMark (the latest edition) and second to that 3DMark11. Reason being; these are common tools for measuring performance and they provide reliable and consistent results. Here are the results briefly of this review

- 3DMark 11 P23,363
- 3DMark 11 X9,378
- 3DMark Firestrike18,152
- 3DMark Firestrike Extreme9,246

I have attached pictures of the website at the time of publication showing these scores. Author has since updated the 3DMark results in both Firestrike and Firestrike extreme.

This images will no longer be found on the current review as it was changed. Oj0 and I have since been banned for trolling. This is the link to the original review page in PDF Format

4. These results naturally became suspect as they did not in any way show the kind of performance I was experiencing with an Identical graphics card. Consistently the system I was testing on scored lower in three benchmarks listed above and substantially so in the latest 3DMark Concerning especially since these are the settings I use for testing.

• - GIGABYTE X99 SOC-Champion (F4i) which was later switched to the MSI X99A Xpower AC
• - Intel Core i7 5960X @ 4.5GHZ (45x100) | 4000MHz Uncore (40x100)
• - Gigabyte GeForce GTX 980Ti G1.Gaming SOC (The identical GPU as the one under review)
• - CORSAIR DDR4 3200MHz (4x4GiB) @ 3200MHZ CL16-18-18-36-2T
• - CORSAIR AX1500i PSU
• - Windows 8.1 64-bit (always updated) and Windows 7 SP1 64-bit (for consistency checking with Press Kit results)
• - NVIDIA GeForce series latest 353.12 driver (as downloaded from Guru3D.com)

My results are as follows with the verification links and pictures under

• - 3DMark 11 P23,577 - NVIDIA GeForce GTX 980 Ti video card benchmark result - Intel Core i7-5960X,MSI X99A XPOWER AC (MS-7881)
• - 3DMark 11 X8,969 - NVIDIA GeForce GTX 980 Ti video card benchmark result - Intel Core i7-5960X,MSI X99A XPOWER AC (MS-7881)
• - 3DMark Firestrike - P17,696 - http://www.3dmark.com/fs/5084194
• - 3DMark Firestrike ExtremeX8849 - http://www.3dmark.com/fs/5084134

You can clearly see that despite, both GPUs operating at the identical OC mode (1190MHz base clock and 1291MHz boost clock), my results are consistently lower than the ones reported on GURU3D, barring 3DMark11's performance test. Even though the host system being faster as made evident by the Physics scores. The GPU scores are substantially lower. His Firestrike Graphics score is 20,358 compared to my 19,384. Nearly 1,000 points between the two at supposedly factory OC-Mode. Firestrike Extreme shows his GPU score at 9,600 while mine is 9,121
You may have to only focus on 3DMark Firestrike and Firestrike visual validations as that is the only thing we have to go by here other than the writer's word.

The system that should be faster or score higher is scoring lower, consistently. The differences in GPU scores are remarkable, with the Graphics sub scores in each case (where visible) higher than the ones I am able to achieve. This difference is not within the margin of error, but as you know, consistent with a higher overclock.
However, since it is possible for the efficiency of my system perhaps for some reason or another to be particularly low. I proceeded to test my results against what the Press deck from GIGABYTE has as a reference results. The picture attached is from the PDF press deck.

I proceeded to configure my system in a way (via UEFI) to make use of only four cores (HT was kept on however). CPU clock set to 4GHz, two sticks of memory removed (to simulate dual-channel setup as with the 4790K) while everything else remained the same. As you can tell, the scores reproduced are in line with the press deck. So perhaps, both the GIGABYTE test environment and my test environment have equally poor efficiency compared to what GURU3D reports.

• 8235 Vs the Press Deck 8256 in FireStrike Extreme in Windows 7 x64 (new installation)

8235 link - http://www.3dmark.com/fs/5083892

Clearly, there is some parity between the press deck numbers and what I am able to achieve. If anything, the results are within the margin of error.

We then move on to the supposed overclocking results on Page 36 of the GURU3D review. As per his suggestion, I used his identical settings, highlighted in the picture below. The Boost clock on my G1.Gaming 980Ti happens to be even higher than what he suggests the actual boost clock would be. His claim is that some titles will dip as low as 14522MHz ~ 1477MHz, but as high as 1,555MHz. Since this will vary per card, mine ends up at 1529MHz and is relatively consistent at that. GPU-Z reports 1302MHz, much like on his table for Core clock when overclocked. Using the following settings.

• - Temp Target 85 Degrees C
• - CPU clock +150 MHz (from default 1152 MHz)
• - Mem clock +375 MHz
• - Voltage + 50Mv
• - FAN RPM 55% (recommended but a little more is noisy)

To verify the results again, I proceeded to run 3DMark FireStrike in accordance with the configuration he claims, which is again. A 4.4GHz CPU clock, a DRAM frequency of 2133MHZ and what we assume to be an UNCORE clock left at AUTO (At any rate it would not be over 3,500MHz as it is not possible or at least highly improbable on the motherboard he is using).
These were my results, emulating his test environment and manual OC above.

3DMark Firestrike - 19097 (His)

3DMark Firestrike - 18897 (mine) - http://www.3dmark.com/3dm/7330499

Again, my Physics score even when emulating his system is higher at 21,169 vs. his 20, 447. Yet for some inexplicable reason his Graphics Score is 21,571 vs. my 21,136

Good people
I will then present to you some more evidence, which if anything should raise your levels of suspicion.
Prior to this review, on 1 June 2015. GURU3D published a review on the reference NVIDIA GTX 980 Ti. I would like to bring to your attention, page 24 where you will find screenshots of the 3DMark runs for that review.
You will note that in the screenshots, SystemInfo is running and able to verify that the score is indeed from an NVIDIA GTX 980Ti on a system powered by the Core i7 5960X. Compare that review with the review that is under scrutiny where the SystemInfo has been disabled. Thus there's no way verify that this system was indeed running a Geforce GTX 980Ti, let alone which model. So from just the previous review, to this one (in almost two weeks), SystemInfo needs to be disabled, when the test system has not changed? Why?
Upon inquiry as to why this was done, this is what the author stated.

"Validation is disabled in my benchmarks as we review a lot of embargo stuff that we do not want to see online. "

This is particularly odd because there is nothing that could be shown by SystemInfo that would be under NDA, especially since the test system has already been listed and apparently hasn't changed from the last review.
Naturally the thread at GURU3D, with all of this has been edited and offending posts including mine and from the other individual, Oj.overclock.Me, have been removed. Fortunately, he had the foresight to save these pages prior to the deletion. You may at your leisure download the original thread prior to editing here.

If you have managed to read this far, I thank you. I leave it up to you, to have your say on this matter. As I truly do believe that these results were fabricated, possibly in several ways. Efforts have been made by the author at GURU3D to produce unreliable numbers. Furthermore to obfuscate any inquiry into this and discussion, two individuals were banned in an attempt to silence them.

Finally. I want to take you back to the original 3DMark Firestrike score of 9,246 with a Graphics score of 9,600. The first thing I saw that made me suspect this entire review. It is, exactly what betrays the falsehood of it all.
The beauty of having a 3DMark Professional/Advanced licence is that you can easily track your GPU, and GPU memory clocks during the duration of any run. However, you can't if you disable SystemInfo. As I do no such thing, in the name of transparency. I bring you the following picture.

9,247!
Let's talk about getting close to a score, well there it is.
GPU score 9,602 Just two points higher than his!
As we have seen the Physics sub-score in this test is of little to no consequence (hence the Author's ability to run a system that scores almost 1,500 points lower in Physics scores but still achieve a higher overall score. If you will, it is entirely GPU bound)
Pay close attention and you will see, that the grey dotted line above the others, is the GPU Clock. It is consistent and always above 1400MHz, in fact to be exact it is 1434.1MHz throughout the entire test. What of the memory? Well look to the bottom left of the image under Graphics adapter. You will see a memory clock of 1940MHz which is 7,760MHz 760MHz more than the stock memory frequency.



- Credit to TheOverclocker for the work involved

 

[CAPSLOCKSTUCK]

Definitely some interesting results/observations there.

i cant offer any insight ...sorry, though i am looking forward to the comments of other members here.

i have noticed anomalies while benching cpus on HWBot so this thread is of interest to me.

 

[[XC] Oj101]

Over here we have GeForce GTX 780 Ti review - Introduction, and if we look at the 3DMark 11 scores we see P13496 and X5114. That was done with a Core i7-3960X clocked at 4.6 GHz. If we look at your review of the Gigabyte GeForce GTX Titan X we see on page 8 that you're now on a Core i7-5960X at 4.4 GHz (and undoubtedly a newer graphics driver), but if we go to the 3DMark 11 scores we see the same P13496 and X5114 in the graphs. The same is true of the graphs in the Gigabyte GeForce GTX 98-0 Ti G1.Gaming review, have a look at the 3DMark 11 graphss. That isn't a fair comparison, is it? It is not representative of the true performance difference between the 780 Ti and the two newer cards, is it? Yet if we don't reference back to the 780 Ti review we don't know that it was tested on a COMPLETELY different platform.

The same can be seen with 3DMark 13, on the 780 Ti we see a score of 9,935 which is still referenced in the Titan X review (graph) and the G1.Gaming review (graph)

 

[erixx]

In overclocking contests all kind of shit will be tried in or outside of the rules. But in a product review it should be totally consumer friendly and clarifying, NEVER MISLEADING!

 

[XSI]

i had differences in few hundred points on 3dmark 06 with my system.
what i think is: guru3d had a better configured system, without unnecessary programs running in the background, maybe with higher clock for longer period of time.
i've been reading them for 10+ years, so i imagine how messed up it would be to compare all those products if some reviews would be staged.
i think everything is all right here.

 

[erixx]

All TPU folk is now visiting Guru3D forums. Wild wild world (www) baby.

 

[[XC] Oj101]

i had differences in few hundred points on 3dmark 06 with my system.
what i think is: guru3d had a better configured system, without unnecessary programs running in the background, maybe with higher clock for longer period of time.
i've been reading them for 10+ years, so i imagine how messed up it would be to compare all those products if some reviews would be staged.
i think everything is all right here.

Did you look at the results? They were replicated on a new test rig, no processes running in the background or anything. Differences of 100 points in 3DMark06 are to be expected when you're scoring way into the tens of thousands - 300 points on a 40,000 point run is a vastly larger margin than 300 points on a 9,000 point run. Additionally, if you compare what was scored with the rig used to reproduce the score you'll see it's perfectly in line with the Gigabyte press deck - a difference of only 21 points. How is the difference between our score and G3D's score THIRTY TIMES larger than the difference between our score and Gigabyte's score?

The following is also part of what got me banned:

I'm still not convinced your clock speeds are right. Efficiencies for the 980 Ti are around 6.9-7.2 points/GPU MHz on a heavily tweaked bench setup (LOD, the works). If we take the score in the review, you're showing an efficiency of over 7.5 points per MHz on something that is "out of the box" - that would mean you didn't use any tweaks? I simply don't understand ow you're getting better performance on a non tweaked setup than ANYONE has gotten out of a heavily tweaked setup?

Here's a link if anyone wants to see for themselves http://url.hwbot.org/1eaCQm9

 

[[XC] Oj101]

I'm seeing more posts being deleted in the review thread

 

[qubit]

@[XC] Oj101 I'm not surprised that you found something funny at guru3d, as it's not the first time I've heard bad things about them.

Also, I used to be a news writer for TPU a few years ago and with one of my stories about something controversial (I forget what it was about now) they copied the whole thing word for word as if it was their own and pretended it was their story with no attribution back to me or TPU. That really pissed me off. Therefore, that they are dishonest hardly surprises me and I just ignore guru3d nowadays.

Just go by TPU's reviews and you'll be alright.

 

[Bill_Bright]

It would be easy to perform reliable, consistent reviews if the computer universe was stagnant, never changing. But it's not. It is ever changing and ever advancing. Everything (hardware and software) is constantly being upgraded and revised.

The only true way to objectively conduct a valid comparative review would be to always use the exact same platform. That is, the exact same motherboard, PSU, CPU, RAM, OS, benchmark programs, and regular applications, all configured the exact same way. And ideally, with multiple copies of this platform to test multiple samples of the product at the same time. Not just one card on one motherboard.

And these test samples should be purchased off the shelf, not supplied by the maker to ensure they have not been specially tweaked to achieve a good review.

All the testing parameters and procedures would have to conducted in the exact same way.

The power and room environments would be controlled and maintained in the exact same way for every test, month after month, and even year after year.

Not going to happen.

While TPU's reviews are great, it would be a mistake to go by one review alone when making purchasing decisions if the desire is the spend your hard earned money wisely.

I NEVER EVER pay attention to just one review site when researching parts for my own builds, or when offering advice for others. So I read the reviews here at TPU, but also OverclockersClub, Toms Hardware, HardOCP, AnandTech, Guru3D and where ever else my friend Bing Google tells me a review exists. And I pay attention to them all.

The only reviews I don't pay attention to are "user reviews" at sites like Newegg and Amazon because buyer reviews often downgrade a part because FedEx dropped the box, or UPS took an extra day to deliver, or Newegg said it was blue when it really was red. I only take note of user reviews if there are many reviewers saying the exact same thing about the actual product.

 

[qubit]

Good points, Bill and for absolute consistency one would have to go to these lengths - that's the scientific method after all. However, the differences seems a little too large here and the practices over at guru3d are somewhat shady given what our OP has said and my experience of them, so I think it's fine to question them those results. If one could find another site or sites running the same tests and compare them all that would be even better. If those other sites also show lower performance figures, then it lends weight to the assertion that there's something fishy going on at guru3d with this graphics card review.

 

[64K]

While TPU's reviews are great, it would be a mistake to go by one review alone when making purchasing decisions if the desire is the spend your hard earned money wisely.

I NEVER EVER pay attention to just one review site when researching parts for my own builds, or when offering advice for others. So I read the reviews here at TPU, but also OverclockersClub, Toms Hardware, HardOCP, AnandTech, Guru3D and where ever else my friend Bing Google tells me a review exists. And I pay attention to them all.

The only reviews I don't pay attention to are "user reviews" at sites like Newegg and Amazon because buyer reviews often downgrade a part because FedEx dropped the box, or UPS took an extra day to deliver, or Newegg said it was blue when it really was red. I only take note of user reviews if there are many reviewers saying the exact same thing about the actual product.

Good advice. I don't mind spending the cash on my rig but I do take the time to go to several sites to see what reviewers have to say. I don't put much weight on user reviews either especially when it's not a requirement that you even bought the item to begin with. Take for example last year with the EVGA Titan Z at Amazon. There was something like 30 user reviews and almost very one of them was just making fun of the card. E-tailers don't monitor user reviews apparently.

 

[ShockG]

JagatReview has their first look at this card up as well
http://www.jagatoc.com/2015/06/first-look-gigabyte-geforce-gtx-980-ti-g1-gaming/
It's in Bahasa but the numbers speak for themselves. They have focused on 3DMark FSE as well.

 

[OneMoar]

guru3d has been on a downward spiral for 5 years not at all surprised
-that being said
this is probably nothing more then a cherry picked "reviewer" card vs a "retail" there is also the possibility of driver difference nvidia has been known to ship "special" drivers to reviewers
benchmarks will give you a ballpark nothing more they are not intended to provided scientific data regardless of what they claim
everything you posted is *well* within the 5% margin of error
nothing to see here people move along

 

[the54thvoid]

Umm.

Just quick scanning your result - you have as 2% deviation from his score in gfx. That's not a significant difference. 21 571 versus 21 136.

I'm quite sure some review sites can be suspect but...... meh...

 

[Bill_Bright]

Another problem with user reviews is they are requested at the time of purchase and typically posted within hours of installation.

And a problem with all reviews is they never go back a year later to see how the product has held up.

 

[MrGenius]

"Don't believe anything you hear, and only half of what you read". A method that stands the test of time. In other words...almost all "information" is wrong in one way or another...always has been...always will be...that's a given. Take all of it with a grain of salt...and a truck load of skepticism.

Yes, as well as being a Grand Master Skeptic, I'm also a cynic who leans heavily towards nihilism. Faith in mankind has never existed for me. I don't expect that to ever change. And it's not something I recommend to anyone.

 

[[XC] Oj101]

@Bill_Bright Until now now, there was only one review to go by and that was Guru3D's.

guru3d has been on a downward spiral for 5 years not at all surprised
-that being said
this is probably nothing more then a cherry picked "reviewer" card vs a "retail" there is also the possibility of driver difference nvidia has been known to ship "special" drivers to reviewers
benchmarks will give you a ballpark nothing more they are not intended to provided scientific data regardless of what they claim
everything you posted is within the 5% margin of error
nothing to see here people move along

5 % is far too large a margin of error. I get consistency to within 0.3 % between runs. The discrepancy between the scores made by Gigabyte and the system used to replicate (or attempt to replicate) Guru3D's scores was only 0.26 %. A few things though:

• IF it was a card running a custom BIOS purely for Guru3D which boosted way beyond any other card, Gigabyte would need to be taken to task, as would Hilbert for not disclosing such info should he have been aware.
• IF it was a custom driver, every 980 Ti review done with the "review" driver should show the same results, and Hilbert would need to be taken to task for saying that he used version 353.12
• Reviews should be able to give you a ballpark of performance, but if one review stands out with implausible results should it not be questioned?
• Reviews should be able to give you a ballpark of performance, but refer to post #3 here. How can comparisons be made where completely different setups were used? Should THAT fact not be disclosed? NOWHERE does it say that different setups were used, the only way you'll know is if you go to the original reviews and compare for yourself. Is that fair?

There are problems here, if you don't want to accept that the scores don't add up you have to admit that there's a flaw with the testing methodology.

 

[OneMoar]

@Bill_Bright Until now now, there was only one review to go by and that was Guru3D's.



5 % is far too large a margin of error. I get consistency to within 0.3 % between runs. The discrepancy between the scores made by Gigabyte and the system used to replicate (or attempt to replicate) Guru3D's scores was only 0.26 %. A few things though:

• IF it was a card running a custom BIOS purely for Guru3D which boosted way beyond any other card, Gigabyte would need to be taken to task, as would Hilbert for not disclosing such info should he have been aware.
• IF it was a custom driver, every 980 Ti review done with the "review" driver should show the same results, and Hilbert would need to be taken to task for saying that he used version 353.12
• Reviews should be able to give you a ballpark of performance, but if one review stands out with implausible results should it not be questioned?
• Reviews should be able to give you a ballpark of performance, but refer to post #3 here. How can comparisons be made where completely different setups were used? Should THAT fact not be disclosed? NOWHERE does it say that different setups were used, the only way you'll know is if you go to the original reviews and compare for yourself. Is that fair?

There are problems here, if you don't want to accept that the scores don't add up you have to admit that there's a flaw with the testing methodology.

seriously ... %0.3
wake me when you find reality ...
also all your other bullet-points is something the industry has been doing for YEARS,nothing new techpowerup does guru3d does it .. OCN does it ...

as much as you might like to belive other wise there is nothing going on here other then the typical NVIDIA Launch caused by Dravidian's restrictive NDA and Various Reviewer Agreements(which you aren't permitted to talk about)

when a company such as nvida tells you to review x in such a way using A B C for testing and setting you don't get a choice Or you don't get a card .

nothing remotely new everybody knows it except for apparently you

 

[[XC] Oj101]

seriously ... %0.3
wake me when you find reality ...
also all your other bullet-points is something the industry has been doing for YEARS,nothing new techpowerup does guru3d does it .. OCN does it ...
as much as you might like to belive other wise there is nothing going on here other then the typical NVIDIA Launch caused by nvidia's restrictive NDA and Reviewer agreements
when a company such as nvida tells you you review x in such a way using A B C for testing and setting you don't get a choice
nothing remotely new everybody knows it except for apparently you

Then why does no other review match up to Guru3D's? Take your pick, there are plenty of reference points available all over the internet. There are quite a few scores on HWBOT, there are two other references you can use for the exact same card, there are reviews of other 980 Tis, nothing as in nothing adds up with Guru3D's claims. His efficiency (performance per clock) is better than you'll find on the Titan X. Have a look at what you see, take the graphics score and divide it by the frequency to get an efficiency (points score per 1 MHz GPU frequency). As an example, if a run scores 8,000 points and it was done at 1,600 MHz it has an efficency of 8,000/1,600 = 5 points per MHz. This allows us to do two things:

1. We can see if a score is out of alignment with similar setups, and
2. We can accurately extrapolate scores at different frequencies

If we continue with the above example and we look at the following scores:

8,102 points at 1,615 MHz (5.017 points per MHz)
8,792 points at 1,760 MHz (4.995 points per MHz)
8,444 points at 1,672 MHz (5.050 points per MHz)
7,438 points at 1,500 MHz (4.899 points per MHz)
6,963 points at 1,212 MHz (5.745 points per MHz)
9,571 points at 1,930 MHz (4.959 points per MHz)

We can see that even though second last result is the slowest it is the one we need to question as it is out of line of the other scores. For example, at the same 1,930 MHz as the last score we can extrapolate that that person would score around 11,100. Something isn't right about that score and it isn't within a margin of error. Now then.

• Bench runs on HWBOT are generally in the range of 6.9 to 7.2 points/MHz (you can verify this for yourself - http://url.hwbot.org/1eaCQm9)
• Untweaked runs AKA "out of the box performance" are generally in the range of 6.5 to 6.8 points/MHz (as was shown by the stock run compared to Gigabyte's own results)
• Hilbert's (now quietly removed) score was over 7.5 points/MHz

How does an untweaked run score a better efficiency than any tweaked run? How does he score more than any bencher that has run the 980 Ti on HWBOT, or any other reviewer that has tested the card?

 

[ShockG]

seriously ... %0.3
wake me when you find reality ...
also all your other bullet-points is something the industry has been doing for YEARS,nothing new techpowerup does guru3d does it .. OCN does it ...
as much as you might like to belive other wise there is nothing going on here other then the typical NVIDIA Launch caused by nvidia's restrictive NDA and Reviewer agreements
when a company such as nvida tells you you review x in such a way using A B C for testing and setting you don't get a choice
nothing remotely new everybody knows it except for apparently you

This has nothing to do with NVIDIA. There is no agreement needed between NVIDIA and GURU3D to review a GIGABYTE GPU. There is no special driver involved either.
I just posted for you the review from JAGAT-Review. Their numbers for the GPU scores are 9013 vs the OP's 9121. Compare that with Hilbert's 9600.
3DMark GPU score does not deviate like that, run after run. Even in 1000 runs it doesn't. between JagatReview and the OP's post it's 1.18% deviation. With GURU3D's results it's 6.1% on a test that is entirely GPU bound.
Inquiry into the scores is shut down. There are no verification links. SystemInfo is disabled and results are out of line with all others and those of all recorded at HWBOT.
From all of this you figure there's nothing going on?
I've yet to suspect any numbers from TPU as being fraudulent. So I've no proof that they do anything GURU3D does. It is wrong to mention those two together as if there's been anything to suggest TPU has previously, in their GPU review, misrepresented hardware. This is about GURU3D, where there is evidence of tempering and it is right there.

 

[erocker]

Am I missing something here as most of the results seem to be within the margin of error/irrelevance?

 

[Bill_Bright]

@Bill_Bright Until now now, there was only one review to go by and that was Guru3D's.

What? That's a bunch of bull! That is totally wrong. Not sure what rock you've been hiding under but there are many review sites that have been around for many years.

What seems clear to me now is that Guru3D was YOUR one go-to site for reviews. And now they hurt your feelings.

TPU, for example has been making great reviews at least since 2004. Jonnyguru since 2006.

Hard|OCP has been reviewing graphics cards since 2002 and CPUs since 1998!!!!

If you think Guru3D has been the "only one review site to go by" then I am sorry, but you have tunnel vision. You cannot, and MUST NOT go by one review if you (1) truly want to be informed and (2) want to make sound purchasing decisions.

Am I missing something here as most of the results seem to be within the margin of error/irrelevance?

No you are not missing anything. This is all about one person being upset with their favorite (or only! ) review site letting them down.

 

[[XC] Oj101]

What? That's a bunch of bull! That is totally wrong. Not sure what rock you've been hiding under but there are many review sites that have been around for many years.

What seems clear to me now is that Guru3D was YOUR one go-to site for reviews. And now they hurt your feelings.

TPU, for example has been making great reviews at least since 2004. Jonnyguru since 2006.

Hard|OCP has been reviewing graphics cards since 2002 and CPUs since 1998!!!!

If you think Guru3D has been the "only one review site to go by" then I am sorry, but you have tunnel vision. You cannot, and MUST NOT go by one review if you (1) truly want to be informed and (2) want to make sound purchasing decisions.

Um, where is TPU's review of the G1.Gaming? I said there was only one site to go to for a review of the G1.Gaming. For what it's worth, G3D stopped being my go-to back in 2007 Let me rephrase that, they were never a go-to but I stopped using them as a primary resource back in 2007.

In retrospect I didn't make that point very clear, I was referring specifically to reviews of the graphics card used in this review by Guru3D, the GTX 980 Ti G1.Gaming.

 

[erocker]

Um, where is TPU's review of the G1.Gaming?

Soon!

NVM, it's up!http://www.techpowerup.com/reviews/Gigabyte/GTX_980_Ti_G1_Gaming/