which report is that?
There’s a thread here somewhere about the 2080Ti. Sorry j can’t find it on my phone ATM
On it someone put up a 30min Spanish benchmark review.
Somewhere in that the guy tests the 2080Ti on FO4VR and another game.
Someone in the thread translated the Spanish.
It’s quite astonishing if true.
Nvm I found the video, from 23:40 onwards https://youtu.be/Xxl-whBL8do?t=1422 Cheers for the info! Shame he didnt say what the supersampling was set to.
1 Like
Good man. And thanks for linking here for others.
Only trouble is I don’t speak Spanish. I also we always thought it was just made up sounds to confuse people but apparently it’s a real language because someone translated it.
The numbers seem crazy though. Perhaps he’s comparing with 1080. I’d love several people to clarify what this guy is saying.
1 Like
The reviewer isn’t even using the fastest RTX card, in a few weeks the EVGA FTW3 / Kingpin card overclocking benchmarks should come around. Those cards might be able to achieve a stable overlocked 2150Mhz GPU core speed or over before throttle, seeing as people can get those speeds now at 100% fan speed overclocking on other cards that dont have as much power draw as the FTW3 and Kingpin cards will have.
Wow. You think it’ll only be a few weeks? And how much faster would it be you think? I thought the difference between normal and better cards was just a little bit. Maybe I’m wrong.
Does the gpu is bottle neck by cpu?
My friend use i5 + 1060 and get 9,000+ score on vr mark orange room, but I use amd fx-8450 + 1070 and get only 6,000+
Really? So all us folk that want multi-GPU for rendering (with NVLink for memory pooling) will not work on the RTX 20 series? Only the Quadro’s ?
From what I understand, the founders edition cards may have binned chips meaning higher performance, but Nvidia’s cooling solutions aren’t as good as third party manufacturers.
On the other hand, the third party manufacturers may be getting worse chips but the cooling solutions are better, so you may see them clocked slightly higher and performing slightly better (although under overclock) than the FE cards.
Also, I’m pretty sure the RTX bios is locked to prevent any overvolting or additional power draw. GPU performance is pretty much directly translated by GPU clock speed, (higher speeds need higher voltage, which needs more cooling, at risk of burning out the components) But for around 25w extra you can get 100 mhz more on average, so you see some cards with power limits of 300W, others have power limits of 330W. Founders cards for the 2080ti series don’t usually go over 2000MHz GPU core clock speed under overclock, whereas the MSI Trio X can do 2150Mhz under full load on air cooling (throttle will reduce this slightly to prevent the chips from being damaged) .
Some people say “power limit doesn’t affect performance.” This is true, however, it does affect overclocking potential. GamersNexus and Jay 2 cents would both agree with me here.
You want a higher voltage to put higher speeds into a chip? You need more power to do that and more cooling.
The FTW3 2080TI is looking at like 373W vs 338 watt for the EVGA XC.
Keeping in mind performance improvement is less and less the more you overclock, I’d say that the FTW3 could reach 2200-2250Mhz under peak load and 2150Mhz stable.
The Kingpin version is going to use a hybrid Liquid/air cooler, which will have lower temps with the possibility of getting as much as 2200-2350mhz stable, (this is already being done with lots of intake/outake fans and custom waterblocks on the EVGA XC 2080ti models without LN2)
The card used in the bench you referred to was a 2080ti Duke.
“For clock frequencies, we’re looking at a 1350 MHz base frequency, with Turbo allowance towards 1635 MHz”
quoted from MSI GeForce RTX 2080 Ti DUKE review
Doing some maths, taking the 1635Mhz Turbo speed of the 2080ti Duke, dividing that by 10 we get 163.5 and multiplying it by 14 we get 2289Mhz. So that would be an improvement of over 40% in performance over the Boosted Duke 2080ti if we see a card with a stable 2300Mhz+ clock speed, which is likely in either of these two upcoming cards. In terms of frames, if the Duke is getting 80FPS then the FTW3/Kingpin could be getting as much as 108FPS. If it was 120FPS it would be 162FPS at 2300mhz. An increase in power of 100mhz in the CPU or the GPU usually equates to 5% performance gain.
Bearing in mind you’d realistically need a CPU with the first two cores clocked to like 5-5.1+ gigahertz at least to not have a bottleneck in this setup, a delidded 8700k in most cases can do this.
This can be done on air cooling or with a waterblock but CPU needs delidding. Anything beyond 5.1 ghz CPU overclock in terms of performance improvement would be negligible in most cases.
For like, 25W more on the CPU and 73W more on the GPU you’d get at least 30-40% (35%) more performance without any liquid nitrogen. a 750W PSU should be more than enough to run that, a good 650w PSU would do.
Gaming performance, put simply, is pretty much directly translated by the single core performance and core clock speeds for your CPU (first two cores most important), GPU core clock speed and RAM clock speed and RAM latency/timings (low CL is better). Single core performance is what matters, not how many cores you have or SLI, Everything else doesn’t matter, most games aren’t coded to utilise more than 1 or 2 cores, rarely 4. Even hyperthreading doesn’t utilise all avaliable threads. This is cause programmers have to code for the hardware which is applicable for the most amount of people (old games = single or dual cores), in the future this may not be so and more and more newer games are using more cores and threads.
Edit: Forgot to mention the FTW3 and Kingpin cards are going to have custom PCBs and custom BIOSs to get around the locked voltage and power limitations nvidia has set in their BIOS.
For those considering overclocking their hardware
You need a good motherboard with good lanes and 16x PCI to for there to not be bottlenecks caused by the motherboard if you’re going to overclock to these extremes, not just for the extra speed a good motherboard delivers but especially with respect to CPU and RAM overclocks, normal motherboards are not designed to tolerate so much heat and the temperature monitoring software doesn’t accurately read the heat on all the components on the motherboard because not all of them have temperature monitors. If you get a temperature gun and point it at the components on the motherboard around the CPU, you’ll get a different read to what your computer might tell you and this could destroy your motherboard. Factory GPU overclock is okay without additional worry.
If you want to overclock your CPU and RAM make sure you get a motherboard that’s designed to do it, which will have proper heatsinks on modules and circuitry which needs additional cooling when overclocking. Don’t ever overclock your CPU beyond 1.4v, 1.35v is where I take mine but with 1.4 you’re looking at turning a 20 year lasting processor into a processor that lasts for 10 years.
6 Likes
This is because you’re using an older generation AMD series CPU with lower single core performance than the i5. Yes, the CPU is definitely a bottleneck for you. For your friend, almost certainly no bottleneck for the 1060 and an i5 (although i’m not sure what kind of i5 he has, if it’s an older one, maybe.)
You definitely need to upgrade your CPU first, to an i7 8700k/ i5 8600 or Ryzen 2700X, 2600X (listed in order of preference and performance best to worst). None of those CPUs will bottleneck an 1070, but a 2080ti will most definitely be bottlenecked by anything except an 8700k or 8600k.
Wow well explained mate thanks a lot.
1 Like
sir sir, would my 4790k bottleneck a 1080ti or above?
1 Like
Just to be fair, that’s for the FE edition and other cards that use 2x 8-pin power. MSI Trio uses an additional 6-pin for power and should have some additional headroom. Might not be much though.
im not really an overclocker guru but i think the best thing to check would be get a game with a benchmark program or just play a game with task manager, bring up the performance tab or what ever tab shows the graphs and leave it behind whilst you game or benchmark and see whats hitting 100%
either cpu should be hitting 100% which would leave the gpu not hitting max or vice a versa , but 1080ti is boss,
Please anyone correct me here as its what i will do when i have money to upgrade and need to see if it’s the cpu next or the gpu
I had 4790k at 4.8 GHz with 1080 and I went 8700k.
equal fps to monitor but big increase in vr
If I sold my 1080Ti I might get £500 if I’m lucky once selling fees etc paid.
If I bought a 2080TI to replace it then it would cost me £700 for an extra 30% performance.
It cost me around £200 to upgrade from 980Ti to 1080Ti for the same increase.
I am a sane man, I will not be buying a 2080Ti,
2 Likes
it’s pretty much where im stuck at but i’ve got a 1080 so i would see a much bigger jump in performance but the cost is still a hard hitter…
though will wait for the founders (boosted) cards to dry up and go back to standard founders cards with lower clocks and hopefully lower prices, and then hoping that will lower aftermarket card costs
yes, you’re right.
But I am not sure why he can get only 9,000 score like 1060 of my friend.
I am thinking that his ryzen 7 make the gpu to get the bottleneck like this
p.s. Sorry, I remember wrong, my friend get only 7,000+ for 1060.
[quote]sir sir, would my 4790k bottleneck a 1080ti or above?
[/quote]
It would be okay for a 1080ti, maybe a slight bottleneck but not worth the upgrade if you won’t be upgrading your GPU as well
you said your friend had an i5, now a ryzen 7? i’m confused. Ryzen 7 or i5 wouldn’t bottleneck a 1060.
you will need to upgrade both cpu and gpu to run pimax. if you can’t afford an 8700k, buy the 8600k if you intend to overclock, if you don’t get the non-k version. After that i’d choose a 7700k, the 2600x R5, then the 1800x Ryzen 7.
1070 won’t bottleneck any of those cpus i mentioned but if u want a better gpu than 1070, like the 1080ti or 2080ti you should be buying the 8600k at least