GPU Benchmarks - 3DMark Scores Graphs (fix?) + 2080ti tips

Discussion in 'Videocards - NVIDIA GeForce' started by SoppingClam, Jun 15, 2021.

  1. SoppingClam

    SoppingClam Member

    Messages:
    12
    Likes Received:
    8
    GPU:
    RTX 2080 Ti 2130mhz
    TL;Dr but basically, the 2080 Ti scores need a bump :) - this was supposed to be a quick thread. oops ALSO just realised the linked don't show my scores as you have to be logged in. So, just take my word for it <3 eg Port Royal https://www.3dmark.com/pr/824556 TimeSpy: https://www.3dmark.com/spy/17831119 Firestrike Ultra: https://www.3dmark.com/fs/24342461

    There is a bit of an issue with benchmarks and in particular with your RTX 2080 Ti results showing lower scores in every graph by fairly significant amounts and 3000 series often higher. Verified by checking global averages in 3D Mark and also as an owner of both cards I am aware of the potential of the 2080 Ti and they often get compared against the 3000 series which influences buying decisions. In reality, a RTX 2080 Ti tend to be at least 10% faster in any graph shown. Even non-overclocked or underclocked power limited 2080 Ti's are faster.

    The most obvious way to check/verify is by simply searching any of the benchmarks on the 3D Mark webpage for any card. This provides the Global Average Scores and for this example the RTX 2080 Ti is below the scores and the 3000 series are above the average. Considering there are over 100,000 submissions minimum for each cards result and these cards are often compared. It is worthwhile taking note. Example search my 2080 Ti combined result vs RTX 3080 in Timespy see average, me, highest etc. Or Timespy Graphics score vs RTX 3070s with over 327,000 results. Or Timespy my card vs other 2080 Ti's

    It might be easier in the future, especially for older cards, rather than retesting by using the Global Average when using Graphic Card result graphs when there are over 100,000 submissions in any particular benchmark by most cards. I assume you may be using Nvidia's Jenson scores or benchmark results from a very power limited RTX 2080 Ti card which then gets recycled in other websites.

    Unlike the conclusion that shows the 2080 Ti being equal or worse than a 3070, the reality is when comparing globally 3D Mark benchmarks vs RTX 2080 Ti's the 3070 and even the Ti version of the 3070 they are generally a bit slower and don't have the OC potential. While my RTX 2080 Ti card is almost 30% faster in Port Royal than all 3070's submitted and 17% faster than 2080 Ti's on average. with a score of 10,554 compared to average score of 8282 by 3070's and avg 2080 Ti's with 9012. In fact, globally it is almost 10% faster than the very fastest submitted liquid nitrogen cooled 3070.

    This matters as someone who bought into the hype prior to gaining knowledge of how nerfed most 2080 Ti's are and various benchmarks and discussions where Jensen described the 3000 series should be utterly dominating the 2080 Ti.

    Reality is: RTX 2080 Ti sits between the RTX 3080 and RTX 3070. Example even with Port Royal ensuring to use only 1x GPU the average score for a 3070 is 8282 and the RTX 2080 Ti is 9012 from over 100,000 submissions. While my 2080 Ti get's 10,554 with the fastest score from any RTX 3070 being 9705.

    Reading benchmarks of the fastest RTX 3070 Ti that has been released. Gigabyte GeForce RTX 3070 Ti Gaming OC by comparing with Guru3D's benchmarks it does provide a fairly accurate differential vs a non-ti. This is helpful to know as your scores are similar to everyone elses and because 3DMark doesn't have enough 3070 Ti results to do comparisons yet. Just wouldn't be telling people to spend $1500+ AUD on a 3070 and board the hype machine.

    *Review graph shows above 3DMark global average 3070 score by >500 points (Global for 3070 = 12635. Review graph shows stock 3070 with overall score of 13145 while it shows only 12,636 for the 2080 Ti which is <1000 points less than the global average*

    Your scores for this Gigabyte 3070 Ti OC. Improvement of 13% over the average 12635 global overall score RTX 3070 or 5.7% faster than the global average 2080 Ti but 13% faster than the 2080ti used in your graph)
    3D Mark Timespy Overall (14283) and Graphics 14920

    Global benchmarks the average Timespy score for a 3070 is 12635 and the average Graphics score is 13741 from 325,150 submissions.
    Vs. 3070 (non-ti)

    Your graph has average score for a 3070 overall 13,145 (510 points higher than global average). Your graph has he RTX 2080 Ti lower than average by almost 1000 points at 12,636 (RTX 2080 Ti Global Average is 13,501). My RTX 2080 Ti Overall score 16,301 and Graphics score of 17,181


    (9552) 3070 Ti OC
    (9634) 2080 Ti OC <- (mine)
    (8465) 3070 non-ti (graph)
    (8605) 3070 non-ti (global)
    (8518) 2080 Ti (global)

    (8117) 2080 Ti (graph)

    This benchmark is interesting with the global average is 8605 for the RTX 3070 and slightly less for the 2080 Ti. However, some people have managed to crack 15k with the RTX 2080 Ti with a single GPU and plenty over 10k.. Which I doubt will be possible with any 3070.

    As a person that uses websites like Guru3D to pre-purchase hardware and so forth. The RTX 2080 Ti benchmark numbers are below the average provided by all users. Even people with super crappy CPU's giving bad Physx / CPU scores dragging them down.

    Example: Timespy my humble Asus RTX 2080 Ti dual fan 2.7 slot card gets a total score of 16301 vs all the 3070's in the world is ranked 3rd. With a graphics score of 17181 which is about 800 points higher than the fastest 3070.

    3000 series a bit higher than they should be. By simply doing a search on the 3D Mark website for one of the tests and a particular GPU you can see all the submitted scores in the world, thus providing an average score. From this you can then use it to say compare your card with a card you are considering to buy. Make sure search for 1x GPU only and can say look at only graphic scores only so as the CPU score doesn't influence the actual performance of the GPU.

    Purely due to the fake RRP provided by Nvidia's leather jacket man in the kitchen a long while back and lack of stock I still have my Asus Dual Fan RTX 2080 Ti. Shortly prior to the 3070 & 3080 being released a lot of people were telling me to tell my RTX 2080 Ti ASAP as it will get smashed and ill lose money etc. Fortunately I didn't as I still wanted to have a video card before selling, just in case. -> fortunately I didn't sell as the waiting list here in Aus was a minimum of 3 months and the USD to AUD conversion from $499 or $699 USD ended up being a lot more $ than that.

    TIPS FOR ALL RTX 2080 Ti Owners
    Firstly, they are of a high build quality. If running hot and fan is blowing hard, repaste and reseat. Cousin has same card as mine and he couldnt run 300w or stay over 2050mhz it would quickly go over 70 degrees. He bought some thermal grizzly, cleaned it, plonked it back together with the 2.7 slot HS/Fan and it idles basically at room temp and rarely makes it over 65 degrees and uses the 373w power limit bios. Lives in the 2100mhz all day.
    One thing I would like to point out is in relation to the RTX 2080 Ti. While the majority of cards are 2x 8pin power inputs (providing potential max wattage of 2x 150w + 75w PCIe slot = 375w) and are air cooled. The 2080 Ti I have noticed have considerable performance differences even with the same PCB style. I only really took note of this while awaiting my RTX 3080 but especially during the pre-launch to 3000 series benchmark figures and Nvidia Leather Jacket man Mr Jensen mentioning how even the 3070 is faster than the 2080ti. But prior to that the FPS figures from games like Doom Eternal in where I had just completed the game at max graphics in 4K with HDR. When watching the YouTube video comparing with a 3080 I knew it was way slower than how mine ran. So, I ran many of the same sections shown and noted my 2080ti was basically getting the fps of the 3080 and was smashing the 2080ti. So, I thought hmm.

    That's when I found out about Power Limits and how each card has one set in its bios and not anything to do with heat or it's capacity to use it. My Asus 2080 Ti Dual fan has a 2.7 slot heatsink but the fan only starts to move once it has reached 55 degrees. Problem is once these cards hit 40 degrees GPU Boost 3.0 begins the downclocking process.

    After checking in GPU-Z I saw it had a power limit of 280w where the OC model of my card has a 312w to 338w and the fans start working earlier and faster giving it the extra performance. So, after 'upgrading' my bios and with no physical difference I was getting a free 100-200+mhz without touching the overclocking. Purely from not being power limited by the less expensive card bios and it also running cooler.

    Learning from this I have put the 380w bios on the card and have it running at 125% instead of 126% max power so it never goes over 375w. Furthermore, my case is cool and the fans are quiet. So, instead of running 1525mhz it happily clocks between 2100-2160mhz +1000mem with it yet to crash. So, compared to stock numbers my card is cooler yet is 42% faster than a completely stock card.
     
    Last edited: Jun 15, 2021

Share This Page