| General > General Technical Chat |
| Linus Sebastion Turned Down $100M Offer |
| << < (23/35) > >> |
| EEVblog:
--- Quote from: rsjsouza on August 18, 2023, 02:51:04 am ---For one side, Gamers Nexus seemed to be quite stable in views and subscriptions and, according to Social Blade, the last week saw quite a dramatic increase in subs, which could be cast as a shadow on their motivation. Besides, I think he was butthurt with the comment against his channel regarding re-using data --- End quote --- I don't get this, what's wrong with reusing data if it's still valid and you can verify it's the same test procedure? :-// |
| Shonky:
--- Quote from: EEVblog on August 18, 2023, 03:07:40 am ---I don't get this, what's wrong with reusing data if it's still valid and you can verify it's the same test procedure? :-// --- End quote --- Steve/GN loves a good drama and then getting full effect from it. e.g. the PSU thing and other stories. |
| bw2341:
--- Quote from: EEVblog on August 18, 2023, 03:07:40 am ---I don't get this, what's wrong with reusing data if it's still valid and you can verify it's the same test procedure? :-// --- End quote --- I don't follow this area closely so I may be wrong... The GPU makers have a tight grip on the reviewers that receive cards early. They are required to use the supplied drivers and specific configurations with specific games to generate the performance numbers that the GPU companies expect. To set themselves apart, reviewers try to test with additional configurations and additional games (new and old) to generate useful comparisons. They compare these results with higher and lower grade cards and the cards from the other GPU company. They also compare historically with older generation GPUs. The problem is that there are significant changes in performance on every GPU whenever the drivers are updated. If you compare the numbers from the latest card to six-month old results from an older card, you might be making an unfair comparison. The older cards are likely to be better performing now on the new drivers. This inflates the apparent improvement of the new cards, which is exactly what the GPU makers want. Early access to cards is tightly time constrained so that reviewers are unable to do extensive testing before the embargo date. LTT Labs' initiative is trying to quickly automate testing on multiple systems in parallel so that they can deliver meaningful comparisons to older GPUs in the few days of testing possible before the embargo date of a new GPU. |
| Ice-Tea:
--- Quote from: thm_w on August 17, 2023, 09:38:03 pm ---That is his desk, the stuff is always on there. Would you spend 30mins cleaning up just to put the stuff back the next shoot? TBH don't really care if either one monetizes the video or not. --- End quote --- Give me a break. You sound like my daughter. Clean up my room? That's gonna take hhoouuuuuurrssss!!! It takes one minute and a small box to dump it in. --- Quote from: bw2341 on August 18, 2023, 05:09:41 am --- --- Quote from: EEVblog on August 18, 2023, 03:07:40 am ---I don't get this, what's wrong with reusing data if it's still valid and you can verify it's the same test procedure? :-// --- End quote --- I don't follow this area closely so I may be wrong... --- End quote --- Well put. LTT made a bit of a stink that they would retest everything all the time and then Steve "caught" them reusing data. But it were two cards that launched within days and that were tested together. So, more drama over nothing. EDIT: and by well put I don't mean the part about you being perhaps wrong but just to say it was a good recap 8) |
| hans:
--- Quote from: EEVblog on August 18, 2023, 03:07:40 am --- --- Quote from: rsjsouza on August 18, 2023, 02:51:04 am ---For one side, Gamers Nexus seemed to be quite stable in views and subscriptions and, according to Social Blade, the last week saw quite a dramatic increase in subs, which could be cast as a shadow on their motivation. Besides, I think he was butthurt with the comment against his channel regarding re-using data --- End quote --- I don't get this, what's wrong with reusing data if it's still valid and you can verify it's the same test procedure? :-// --- End quote --- New games and cards appear. Need new drivers and game patches to work properly. So the tests with older versions could be invalid. Some games are heavily patched, such as Microsoft Flight Simulator, which saw dramatic (100%+) increases in performance over its lifespan, but also erratic behaviour (poor performance only fixed by a system reinstall). I honestly believe its not easy to properly benchmark. Steve from HardwareUnboxed said its not rocket science, but it is very tedious. LTT wants to boast about automations, but then still is able to post wrong benchmark data. Some of the benchmark numbers swapped (e.g. Intel Arc A750) could be with raytracing on or off, which also has dramatic increases. The other components from GN I don't agree with at its core. Forcing an experiment outcome is IMO still a wrong approach to testing. Because that's the vibe I got from their video. Each test should be ran under the same conditions. E.g. a 300% increase between 4090 en 3090 is dramatic, but we have seen HardwareUnboxed test 8GB cards on high resolutions, and they show greater-than-expected performance retardation because of VRAM limits. Now VRAM limits cant be at fault in the 4090 benchmark (as a 3090 en 4090 both have 24GB VRAM), so that point is still valid. But for other cards.. I don't necessarily agree. The seemed to have been pointing to different charts for their GTX1060 example ("Max settings" vs "Ultra + Vulkan"). GN forced a conclusion "none of the other cards really moved" => Yes, that could be because Ultra is probably Max settings, BUT choosing between DirectX(industry standard/default) or Vulkan rendering could favour manfacturer A,B or C more. I think every reviewer would do benefit if they will write down the project and test run on each chart. And then there is the issue the part variance is huge: https://youtu.be/dGbW7orZS-A TLDR. 13x AMD Ryzen 7600 tested. CB23 MT runs in range 13.6k-14.4k. Clocks 4825MHz - 5075MHz. Load temperatures 79C - 88C. Load power 112W - 126W. Game FPS 60 to 86(!). FPS/W: 2.89 to 4.26 That latter has a 41% improvement on the same hardware bill of materials. Representing a tight cluster of measurement data based on just 1 (or even a handful) CPU or GPU under test is also misleading. That's what GN advertizes in all their charts. |
| Navigation |
| Message Index |
| Next page |
| Previous page |