Видео

I am testing CPU/RAM performance. Game resolution has nothing to do with it, since higher resolution loads GPU. So to eliminate GPU bottleneck, all CPU tests should be performed at low resolution. Most reviewers test CPUs with 4090 at 1080p. 3060 is like 4 times less powerful. So even at 540p it performs much worse than 4090 at 1080p. I'd say 3060 at 540p is even slower than 4090 at 1440p. So it is pretty realistic scenario. If I had more powerful GPU back then, I would increase resolution, at least, to 1080p.

DDR4 XMP profiles are usually rated at 1,35V. If you want to achieve higher than XMP frequency with the same timings, try 1,4V. I had to set voltage to 1,43V to achieve 3800MHz CL15. Previously I used it at 3800MHz/CL16 and 1,35V was enough.

@@starstreamgamer3704 are you doing oc with xmp profile or you create new manual profile with hand writed timings ?

​@@burakahmettr8193 I started with XMP. But it does not matter much, as long as you manually tune voltage.

After v2.00 introduction CPU performance decreased noticeably (in the main game). The only thing that v2.13 brought is FSR-FG support (aka FSR 3).

@starstreamgamer3704 exactly bro, I played the game fine back in 1.6, now driving fast ruins everything

I do not use any specific software, just set Curve Optimizer to +150MHz and -18 for all cores (in BIOS). My sample works fine at these settings and holds 4600 MHz most of the time.

65W is it's rated TDP. While it's PPT is 88W. But it can consume up to 110-120W under some loads, when not constrained by any limits, depending on it's core clock and core voltage. Though the most optimal limit for this CPU is 90W. Anything higher brings negligible performance improvements, but significantly increases the CPU temperature.

@@starstreamgamer3704 then I wonder how r5 2600/3600 perform if they are forced to use more watts

​@@sownlengoc My sample was functioning at 3,8 GHz/1,22V. Next step for it was 3,9GHz/1,28V. This would increased it's power draw to 65-80W in games. At 4GHz/1,38V it was consuming ~90-100W, but it's gaming performance was just 3-4% higher then at 3,8GHz/1,22V. So 70W limit for 2600 is an optimal choice. 5600 just needs a bit more power to reach a good balance between power consumption and provided framerate. 3600 is similar to 2600 on this matter - 70W limit is an optimal choice for this CPU as well.

I wanted to make such a comparison, but this card is not fast enough to run modern games at native 4K/Max. And if you use DLSS set to "Quality", this means games are rendered at 1440p and then upscaled to 4K via tensor cores. So the results will be the same as shown in this video. For older games it might be more useful though. I tested one game at 4K (Horizon Zero Dawn) and found out that VRAM OC brings 8% more frames at native 4K. But since there is no DLAA in this game and modded DLAA (via DLSS Tweaks) performs significantly worse than TAA (and do not achieve locked 60 fps), I do not see any sense in not using DLSS Quality. So testing HZD at native 4K is kinda irrelevant.

@@starstreamgamer3704 light game turn on 4k max but modern games turn on in 4k not im max, just medium in nativ 4k

Good result. May I ask what chips is your RAM based on? And what are the settings you chose for your CPU (frequency, voltage etc.)?

which exact server is it :) I'd love to join

It is! Look at it's tech specifications on any website (TechPowerUp, for example). Memory config: Memory Size - 12 GB Memory Type - GDDR6X Memory Bus - 192 bit Bandwidth - 504.2 GB/s

Why not? There is no significant increase in either power consumption or temperatures or fans speed. So overclocking memory (VRAM) is literally free performance for very low effort. Overclocking GPU is a different story though and depends a lot on the type and size of a GPU cooler, room temperatures, PC case quality, etc. I'd rather undevoted a GPU and overclocked memory. It is a perfect combination of performance/temperatures/noise ratio.

Well, I performed another comparison, this time with enabled Frame Gen as well (both FSR 3 and DLSS 3). You can watch my latest video. Shortly: you can run this game on 4070 Ti (or any other 12GB NV card) at 1440p maxed out with FG, if there are no active background apps, that use additional VRAM. Cards with 12GB buffer are right on edge, to be honest, but they are still capable to run any game at 1440p w/o any issues.

Hi! Unfortunately, I do not have any experience with this specific board. You can check this yourself. It should be called the same.

This CPU has a TDP of 85W. So you do not need massive cooler. Any tower-type model with 3 heat pipes should be sufficient. Just check the combability with LGA 2011-3.

@@starstreamgamer3704 thanks

No problem.

You are welcome. I am glad my video was useful for you and others. If you need higher single core performance, there is not much of a choice. 1620 v3 has only 10MB of L3 cache. So you may want a CPU with larger L3 cache, which might give a small boost. Other than that, v3 Xeons can only offer more cores. But you can look at v4 models, their single core performance is a bit better due to slightly higher IPC. For example 2637 v4, which boosts to 3,7GHz and also has 15MB of L3. Or 2643 v4, which has 6 cores (with the same 3,7GHz boost), but also 20MB of L3. P. S. Make sure Dell Precision T5810 supports v4 CPUs (you might need to upgrade motherboard BIOS).

99% of games can not utilize more than 8 CPU cores. 2667 v4 clocks at 3,5 GHz, while unlocked 2699 v3 - at 3,6 GHz. Taking into account small IPC improvements in v4 CPUs, 2667 should be a bit faster. But it has only 25 MB of L3 cache, while 2699 v3 can boast 45 MB. But such a big amount of cores/threads might hurt 2699 v3 gaming performance a little bit. So, in the end, it is a tough choice. To make thing simple, I'd recommend 2667 v4. This is an easy "drop and use" solution. While with 2699 v3 you should unlock turbo boost and disable about half of it's cores (10 active cores should be enough, I suppose). This will also let this CPU hold it's core clock at 3,6 GHz most of the time.

@@starstreamgamer3704 wow its a great explanation. Now I put 2667 on my Z440. Thank you very much!👍

I am glad I could help. Happy gaming and better winning!

Nevertheless, good result. Though I think even 975 mV at ~2800 MHz would be enough to fit in 220W. At least my 4070 Ti can do it in most cases. 4070 Super is indeed power limited, unlike 4070 Ti. More expensive versions have extended power limits, but they cost almost as much as the Ti. Nvidia is a very calculating company.

Xeon e5 2680 v4 functions at 2900 MHz (all core boost), has 35 MB of L3 cache and a bit higher IPC, while Xeon e5 2680 v3 functions at ~3300 MHz (with unlocked turbo boost) and has 30 MB of L3 cache. So, I believe, overall gaming performance of these CPUs will be very similar. Since Xeon e5 2680 v4 has 2 more cores and higher RAM speed, it is a better CPU for productivity. So I would choose v4, if it costs just a bit more. You will not need to mess with unlocking turbo boost in this case, which is a plus.

@@starstreamgamer3704 nice thanks I just built a pc with the v4 and a rx580 2048 and am really enjoying it. Just wanted to see if I missed a better option thanks again for getting back to me

You are welcome!

You are welcome!

According to some independent sources, R&C: RA can utilize only 1/3 of full NVME 3.0 bandwidth (about 1200 MB/s at peak). So even the slowest NVME 3.0 drive is enough to realize the full potential of this game. There are a lot of comparisons between NVME 3.0 and 4.0 on the web, but no one could find even a slight difference. Same goes to PS5 version of this game - the slowest expandable NVME 4.0 drive (3200 MB/s) matches internal PS5 drive (5500 MB/s) perfectly.

This CPU is a pretty decent value for gaming. Though, it depends on how much cheaper v3 alternatives are currently selling for.

You are wrong. Please, show me, where the game devs claimed PCI-E 4.0 drive is required for the correct functionality of DS? Moreover, the official requirement for DS is as follows: an NVMe SSD (PCI-E 3.0 or higher).

afaik rtx io hasn't really used more than 1200mbps in any test, so how could pcie 4 possibly be obligatory? Not even the full 3.0 bandwidth is necessary...

@@luisguerreiro8045 what is your source? Where did you see that RTX i/O is limited to 1200mb.s? :-o DirectStrage is made to increase bandwidth. So to say that it is limited to 150MByte/s.

@@wizztiti1181 not really what I meant, I'm saying it hasn't gone to that level in any title yet, ratchet and clank for once hardly uses 200mb/s in those portals as you can see in the digital foundry review or simply by toggling a i/o graphic into your rivatuner and checking it yourself, at this point we only have ratchet and clank and forspoken running directstorage, I'm not saying directstorage's won't ever make full usage of the pcie4.0 bandwidth, I'm saying the games that did implement it on pc at the moment don't come anywhere close to even capping a pcie3.0 ssd, that will likely change in the future, but as of now it is what it is.

I am glad I could help.

It depends. i7 6850k is not an outstanding value for the money these days. There is a cheaper version named Xeon E5 2643 v4. It is the same 6-core Broadwell CPU with slightly lower core clock, but bigger L3 cache. Thus they will perform roughly the same. But if you have a decent branded MoBo, which has overclocking capabilities, then the unlocked multiplier of i7 6850k may be of use (even if it's price is higher).

E5 1680v4 is about the same, more cores, no overclock but. more cache

​@@JCMW-hw9jlE5 2680 have 12 cores

Yes, it is pretty easy to make similar comparison between the real 4080 and emulated 4070 Ti Super and also emulated 4080 Super, which will be, at best, 6-8% faster than base 4080, so you just need to overclock 4080 (cores by 6% and memory by 3%). As concerns 4070 Ti Super, it's theoretical performance will be 44.10 TFLOPS, which is 10% more than 4070 Ti (with it's 40.09 TFLOPS) and 9,5% less than 4080 (with it's 48.74 TFLOPS). So 4070 Ti Super is right in the middle between 4070 Ti and 4080. But it's memory bandwidth is 672 GB/s, which much closer to 4080 (just 6% less) or 33% more than 4070 Ti. Judging by one of my previous videos (ruclips.net/video/p6-axcnHFVk/видео.htmlsi=Tz-I_VBQ2Rvhcbmf), 17% of additional bandwidth provides just 5-6% more frames for 4070 Ti. This means that 33% more bandwidth will provide 10-12% more frames, at best, since the rule of diminishing returns applies in this case. Judging by 3070 Ti, which has 36% more mem bandwidth, but is just 7-8% faster than vanilla 3070 (partly due to 4,5% more cores), 10-12% is a bit too optimistic though, but since all these cards use different chips, we can not say for sure. Taking into account the fact than 4080 is ~20% faster than 4070 Ti at 1080p, ~25% faster at 1440p and ~30% faster at 4K, we can assume that 4070 Ti Super will be roughly 8-12% faster than 4070 Ti at 1080p, 13-17% at 1440p and 18-22% faster at 4K (in the best case). This makes sense as 7900 XT (depending on resolution) is usually 5-15% faster than 4070 Ti (in raster). While, according to Nvidia's plan, Ti Super should outperform this card at any resolution (in most games), just like 4070 Super and 4080 Super should outperform 7800 XT and 7900 XTX correspondingly. P. S. For now, 4070 Ti do not have any issues related to it's 12GB buffer at 1440p. While for 4K it is a bit too slow to play at native res and max settings, if we are talking about modern demanding games. But 4K + DLSS at High settings is doable, since it decreases demands for both the chip itself and VRAM. 4070 Ti Super will become a solid lower tier 4K card, just like 7900 XT, thanks to it's much higher mem bandwidth and more powerful chip, but taking into account it's features, it will be definitely be more appealing one. It will be a decent starting option for the owners of 4K screens, especially if 4080 Super is just 12-15% faster and 20-25% more expensive.

@@starstreamgamer3704 Exactly, I ve a Gigabye M27W monitor and it's 2k only, but I ve found issues with Last of Us P1 and Lords of the Fallen where VRAM spill happened. After latest update the issue with Last f Us is gone but I see texture popping in and out at long distance which I feel s a measure the programmers took to minimize GPU Memory. Same happened with Hogwarts Legacy. So I feel a top tier 1440p Ray Tracing GPU or a low tier 4K GPU. At thr pricing they r pushing that is what a 4070 Ti Super should be

​@@1989rs500 At $799 Ti Super would be a good deal (in the current market). But Nvidia will probably price it at $850-$900, which is still much better deal than $1200 4080. But time passes and tech becoming cheaper. That is a usual situation. In a year we will get 5000 series, which offers better performance for the price. 5070 will likely be priced at $600-$700, will have 16GB of VRAM and will be close to 4080 Super in terms of performance. P. S. TLoU needs only 9,3GB at 1440p/Ultra now, after it has been patched. Maybe there is still some kind of a bug related to textures. HL is still a buggy game, same as SWJS, Forspoken, Dead Space Remake etc. Unfortunately, most PC ports of 2023 were broken and many of them are still in a pretty bad shape.

@@starstreamgamer3704 5000 series as per news will not be significant upgrade to 4000 series, 5090 will be a significant upgrade but rest f it will be marginal say 10-20% over current 60,70,80 series. I for 1 wud keep 4070 Ti Super till 5xxx series goes out f production.