Errors & Corrections

What Lands Here

• Data errors, such as test data that was later determined to be incorrect
• Typographical errors that could affect understanding of a product or chart, such as a data entry error resulting in a part being listed at “120 FPS” instead of “102 FPS,” as an example
• Major data or testing errors that result in the removal of videos on rare occurrences they happen (even after its removal)
• Moderate errors that may affect the understanding of a sub-topic
• Mislabeled charts or data that could affect the understanding of the information

What Doesn’t Land Here

• Errors made by manufacturers in their own specification materials, even if we reported on them. Those are not our errors. This page is intended for taking accountability of errors we make alone, so a CPU manufacturer getting its own USB IO spec wrong would not be here
• Typographical errors and other errors which have no impact on anything, e.g. “NVIDIIA” or “Phantex" and other similar examples

As we are starting this catalog page in October of 2024, we have decided to set an arbitrary start time by crawling back through our videos dating back to October of 2023 for inclusion. This is a slow and manual process. It is possible we have missed something, as there was no process for adding errors to this page until October of 2024. Older content with errors we’re aware of will already have the corrections and should have been addressed, but may not get reflected here if older than October of 2023, as we needed to choose a start date. Our process for going back through old videos mostly relied upon searching the YouTube back-end for descriptions including words like "error" and "corrections." If we missed any in 2024, you may email us at team at gamersnexus dot net. New additions will be maintained as they happen, so this will not be a challenge going forward.

Writing from first-person for a moment here: Although different people take responsibility for reviews and testing at different points in the process, ultimately, anything that makes it all the way to publication is my fault, as I still look at everything for sign-off. The upside of this is that, regardless of how the error was made originally and who made it, I am the one tasked with establishing processes to prevent similar errors in the future and help the team root-cause faults to improve as a team. I take accountability and responsibility for errors that make it to publication and am committed to the resolution process.

Going forward, any corrections we issue will now also include a standardized step to add them here. Please note that there may be a latency adding them, but we will try to do so immediately. Dates will be for the date that the content was originally published.

2024 Errors & Corrections

2024, November 17

Content Title: Intel At Its Best: Revisiting the i9-12900K, i7-12700K, i5-12600K, 12400, & i3-12100F in 2024

Error: The 12400 is listed as "i5-12400 (6P/6E/12T) [10/24]." It should be listed as "i5-12400 (6P/0E/12T) [10/24]." This is a specification listing error (name entry error) that has no impact on performance or results.

Determination: We consider this Low Impact. It affected multiple charts and could lead to a misunderstanding of the spec. We are frustrated at this getting through though and are elevating it to Low Impact. It is a typo, but of a spec, and is objectively wrong. Immediately following this, we made changes to our data export process to stop this from happening.

Correction: We added an in-video “correction card” pop-out on YouTube, pinned a comment, and updated our chart labels for the article adaptation. We have made changes to future processes, including an extra QC step from Steve at the end of future exports to sign-off on data labels.

2024, November 6

Content Title: RIP Intel: AMD Ryzen 7 9800X3D CPU Review & Benchmarks vs. 7800X3D, 285K, 14900K, & More

Error: In the video, due to a combination of an unexpected BIOS behavior and a gap in our process, we missed a BIOS setting that changes based on which CPU is installed and caused low performance unique to the 7950X3D data. 7950X3D CPU data is low for this round. No other data is affected. There is no impact to conclusions in the 9800X3D content or to any other CPUs. This does not affect the prior 7950X3D review content, as this was not an issue then.

Determination: We consider this Moderate Impact for the following reasons: (1) the error affects multiple charts; (2) the affected CPU was not central to the story or comparisons, as we generally advise either going for a non-3D 16-core (for workstation) or an X3D 8-core (for gaming). It does not escalate because it has no impact on the core of the content, which was analysis of the 9800X3D, primarily against the 7800X3D, 5700X3D or 5800X3D, and Intel competition.

Diagnosis: We ran extensive additional testing on this to fully understand the issue and leave nothing to speculation. If you would like to learn more about precisely what happened (and it would be useful for people potentially upgrading to a 7950X3D), then you can find the entire diagnostic report in our HW News episode which goes over the specific settings and what we ultimately feel is our own testing oversight.

Correction: To correct this error, we posted a Hardware News (HW News) episode entry (as it was an educational mix of issues that could help inform users) explaining the issue in-depth, provided new charts with updated numbers, updated the charts on the written review article to contain the new data, pinned a comment, updated the description, and added a YouTube in-video "correction card" pop-out. (Update 10/11/24: Changed sentence structure to past tense, as all changes have now been made).

2024, October 10

Content Title: Intel Core Ultra 285K, 265K, & 245K CPU Specs: Bending Fix, Power Reduction, & Prices

Error: At 01:03, an Intel slide presented the differences between the Ultra 265K and the 14900K as being down 5% (or -5%). We misread this slide as the Ultra 285K and misreported that it was down 5%. Intel instead claimed the 285K is down 1.1% (-1.1%) from the 14900K.

Determination: We consider this Low Impact for the following reasons: (1) -5% and -1.1% are not meaningfully different in the scope of a news report on performance; (2) the error could be cut cleanly with the YouTube trim tool; (3) there is no impact to conclusions, especially as this was a news report.

Correction: To correct this error, we trimmed about 5 seconds out of the video with the YouTube trim tool, added a YouTube in-video “correction card” pop-out at the affected timestamp, updated the description, sent a tweet, and pinned a comment. For transparency and historical accountability, all of the updates listed remain in-place, despite the error no longer being present in the video.

2024, August 24

Content Title: Actually Good: $2400 Starforge Pre-Built Gaming PC Review (Lowkey Fractal Terra ITX)

Error: At 23:50, the frequency spectrum chart X-axis label was truncated due to software charting settings and limitations. It truncates the X-axis frequency to a single 0. 

Determination: We consider this No Impact for the following reasons: (1) there is no incorrect interpretation by a user capable of using the data that would make sense; (2) all of the writing and scripting is completely accurate to the chart; (3) the testing is accurate. As such, this is a typographical error on a chart label (technically, it was typed correctly, but truncated by proximity of the axis line) with no impact to the data itself.

Correction: We updated the description with information on this and pinned a comment. There is no impact to interpretation (especially not for anyone who’d know what to do with the information). The script is correct and completely accurate. The chart label is incorrect. 

Improvement: We have fixed our chart templates for frequency spectrum plots to resolve this going forward.

2024, May 9

Content Title: HW News - Intel is a Cluster, NVIDIA Blackwell Boosts Production, Sony "Still Learning"

Error: At 07:56, a story discussing Sony “still learning with Helldivers 2” brought-up a series of breaches and hacks of Sony’s systems. We at one point misspoke the company name and transposed it with another company’s name (Intel) that was also on this list of corporate breaches. We made an error. The intent was to say “Sony.” 

Determination: We consider this No Impact for the following reasons: (1) it was clear to listeners that this was a misread, as any other word other than “Sony” would not make sense; (2) it was easily trimmed, with a fully transparent update left in place and an in-video correction card at the timestamp.

Correction: To correct this error, we trimmed about 2-3 seconds out of the video (removing the reference to the company name, allowing the rest of the sentence to remain in-tact and accurate), we added an in-video YouTube “correction card” pop-out at the affected timestamps, updated the description, and pinned a comment. For transparency and historical accountability, all of the updates listed remain in-place, despite the error no longer being present in the video.

That brings us to the end of the 2024 list. If we missed anything in our crawl back through time, you may email us at team at gamersnexus dot net. We only endeavor to add errors which had an impact on the content. Errors with no impact, such as no-impact typos, will not be added.

2023 Errors & Corrections

2023, November 20

Content Title: Crazy Efficient: AMD Threadripper 7980X & 7970X CPU Review & Benchmarks

Error: Various charts within the review contain an incorrect product name label for the older Threadripper 3960X. Only the label is wrong. No data is affected by this error.

Determination: We consider this Low Impact. It affected multiple charts and could lead to a misunderstanding of the spec and is a frustrating error that we should have caught, and so we are elevating this to Low Impact as a result, despite technically being a ‘typo.’ Fortunately, it was not for one of the parts being reviewed directly and was of a discontinued CPU.

Correction: We added an in-video “correction card” pop-out on YouTube, pinned a comment, and updated our chart labels in the immediately ensuing review. 

Improvement: After this, we added a new QC pass in reviews where a Writer / Technician performs QC on only the product labels (and nothing else) to ensure at least one person is always checking those.

2023, October 19

Content Title: New AMD Threadripper 7980X, 7970X, 7960X, & Threadripper Pro CPUs Announced

Error: At 07:17, the voice-over erroneously stated that the CPU in discussion has “68 cores,” which obviously doesn’t make sense. It should read “64 cores.” This was a misread of the script, which wrote the correct line but was vocally read incorrectly. This is an error.

Determination: We consider this Low Impact. The on-screen material contradicted the voice over and later references to the same spec were correct, helping make clear that this was an error.

Correction: To correct this error, we added an in-video “corrections card” pop-out at the timestamp, updated the description, and pinned a comment. This got by two people and we should have caught this. 

Improvement: Following this, we created a formal internal QC SOP for news videos specifically to add one additional layer of review in most cases. We have no illusions that this will catch everything forever, but it will catch most errors.

2023, October 8

Content Title: [Outdated - New Tests] Counter-Strike 2 CPU Benchmarks: E-Core Challenges & X3D Benefits

Error: During testing, we did not remove the forced 400 FPS limit. Removing the FPS limit (at that time) required a console command to be manually entered at game launch. We did not use this command. There is room for debate as to whether it should be expected to use console commands to override framerate limits (400 FPS) in benchmarking; however, the simple fact is that if we had been aware of this command, we would have unlocked the framerate to verify. As a result, we considered this a test oversight, despite technically not being “wrong” (meaning: it was accurate as tested, but we would have preferred to test it without the engine limit). This still lands on the errors table because we prefer unrestricted tests.

Determination: We consider this Moderate Impact, as it affected multiple top-performing parts, but uniquely was still true to performance under the restricted conditions. Because we did not intend to test under those conditions, though, it gets “Moderate Impact” as the rank.

Correction: This particular issue accurately represented performance under conditions without console commands, but did not represent conditions as we’d like to see (uncapped). This involved subjective judgment, as it was questioning the test philosophy of using console commands to remove limits vs. what most users would do. We removed ourselves from that judgment process as much as possible by polling the community, which voted with 83% favor to leave the video in place but publish updated charts for the most restrained CPUs when running the command. This gave us the most data: Both engine-capped and uncapped. We followed the community’s vote.

We published updated benchmarks here and here. We determined that the vast majority of results for low & mid-tier CPUs were not affected, but that the top-ranked CPUs had truncated scaling due to the limiter. Because we still felt this was a test oversight, we decided to rename the original video to [Outdated - New Tests] to further highlight the limitations of the content; we also pinned a comment and updated the description to explain the limitations of the testing, explained the technical accuracy vs. the theoretical unrestrained results, ran a lengthy and dedicated HW News segment further elaborating (which gets higher views than follow-ups do to maximize audience awareness), and posted multiple YT Community posts about the matter. We also updated Cards and End Slate links to point to the HW News episode.

Closing Notes

This page was created in October of 2024. When we created it, we went back one year for inclusion of errors. We needed a starting point and one year made sense as a semi-arbitrary timescale. We have not yet gone back before October of 2023, though of course there are errors that pre-date this list. The ones we’re aware of are logged on the videos in which they occurred. We will continue adding to this list going forward. We also plan to use it to evaluate our total error rates over time to serve as a gauge of performance. This will be an ongoing maintenance project that will stay at this URL.

The everlasting objective is zero errors. We do pretty well with most videos. Errors are, however, inevitable at any publishing volume. We firmly believe that it is the handling of those errors that matters. That’s what this page is for.

Update Log

• November 9, 2024: Added a new entry for November 6, 2024.
• October 14, 2024: Initial creation of this errors & corrections page and initial data set dating back one year. Established internal processes for maintaining and updating the page.