[SOLVED] 3700X bottleneck ?

[Gabriel3331]

Hello,

I recently bought a prebuilt with amd ryzen 7 3700x and rtx 3070. Doing 3dmark I get around 13k overall with like 14k gpu score, 9k cpu score (had a bit worse before I overclocked my ram). The cpu seems to get average/normal pts in cinebench (1290 in single core, 12550 in multicore).

The cpu scores are maybe a bit on the lower end (could potentially be because of bad timings on my ram memory), but still in "normal" range (so according to benchmarks both the CPU/GPU are doing average/fine).

Now to my question/problem. In real gaming scenarios I've had issues with low gpu usage, at first I thought it was maybe due to badly optimized games or drivers. But now I've seen this behaviour in several games and I've used DDU to remove drivers and made sure I have both the latest gpu and chipset drivers, made no difference.

So some examples:

Cyberpunk: on ultra settings (ray tracing off, DLSS = quality) i get 120+ fps in certain areas, 70-90 in others (pretty common) and drops to even as low as ~50. At first I didn't think too much of it since I knew ppl have issues with cyberpunk. But the weird part is that the gpu usually is at like 70% usage. And even if I drop everything to low, ray tracing off, DLSS off etc, i pretty much have the same FPS. To me that sounds like some pretty heavy bottlenecking (I would expect to have 100 fps+) at all times with low settings with those specs.

Warzone: Same behaviour. Pretty much have the same fps at ultra settings as I have with low settings. Even with everything on lowest, I still drop to like 100fps at times. Also seems like a cpu bottleneck. The only thing I can do get the GPU to bottleneck the CPU is to set render scale (200% (I only have a 1080p monitor btw), raytracing on, dlss off, everything ultra), at that point i get 99% gpu usage and 70 fps. But if i go pretty much any other settings seems to be the bottleneck. And the cpu cant even seem to reach 144fps to utilize 144hz even on lowest settings.

WoW: Badly optimised game i guess, but same thing here, same fps at lowest settings as with the highest settings which is ~70-100 in cities.

Overwatch: With epic settings (and renderscale 150%) I get high GPU usage and high fps (200+) usually. This game seems to work a bit better I guess. If I recall correctly I can get higher fps than that if i lower gpu settings, so cpu seems to be able to not bottleneck as much here (at least its no problem for the cpu to achieve 200+ fps).

From what I've read most people seem to agree that a 3700x shouldnt bottleneck an rtx 3070 (at least not in any major fashion), but this is not my experience at all. The CPU cant even seem to reach 144 fps in the above titles with lowest settings (except overwatch), which I would think is not a unreasonable expectation today.

So either I've had some "bad luck" in the titles I've tried - meaning cyberpunk, warzone and WoW are all badly optimized games or the CPU actually does do some quite heavy bottlenecking if you want 144fps in certain games. Might add that I've also read some other worrying comments with people saying they switched from 3600x/3700x to more competent single core processors and have seen great improvements in warzone FPS (which also would suggest that it is actual dramatic cpu bottlenecking going on).

Any ideas? I'm quite confused by the fact that most people seem to say 3700x shouldnt bottleneck a rtx 3070 in any major way, but my (limited) experience is that it actually does, and quite heavily so if you want to achieve 144fps. What's your ideas and opinions, should I try some more games or just return the computer and buy another one with a better cpu?

 

[Afro_ninja199]

run https://www.userbenchmark.com/Software and see how your looking

 

[cdrkf]

Hello,

I recently bought a prebuilt with amd ryzen 7 3700x and rtx 3070. Doing 3dmark I get around 13k overall with like 14k gpu score, 9k cpu score (had a bit worse before I overclocked my ram). The cpu seems to get average/normal pts in cinebench (1290 in single core, 12550 in multicore).

The cpu scores are maybe a bit on the lower end (could potentially be because of bad timings on my ram memory), but still in "normal" range (so according to benchmarks both the CPU/GPU are doing average/fine).

Now to my question/problem. In real gaming scenarios I've had issues with low gpu usage, at first I thought it was maybe due to badly optimized games or drivers. But now I've seen this behaviour in several games and I've used DDU to remove drivers and made sure I have both the latest gpu and chipset drivers, made no difference.

So some examples:

Cyberpunk: on ultra settings (ray tracing off, DLSS = quality) i get 120+ fps in certain areas, 70-90 in others (pretty common) and drops to even as low as ~50. At first I didn't think too much of it since I knew ppl have issues with cyberpunk. But the weird part is that the gpu usually is at like 70% usage. And even if I drop everything to low, ray tracing off, DLSS off etc, i pretty much have the same FPS. To me that sounds like some pretty heavy bottlenecking (I would expect to have 100 fps+) at all times with low settings with those specs.

Warzone: Same behaviour. Pretty much have the same fps at ultra settings as I have with low settings. Even with everything on lowest, I still drop to like 100fps at times. Also seems like a cpu bottleneck. The only thing I can do get the GPU to bottleneck the CPU is to set render scale (200% (I only have a 1080p monitor btw), raytracing on, dlss off, everything ultra), at that point i get 99% gpu usage and 70 fps. But if i go pretty much any other settings seems to be the bottleneck. And the cpu cant even seem to reach 144fps to utilize 144hz even on lowest settings.

WoW: Badly optimised game i guess, but same thing here, same fps at lowest settings as with the highest settings which is ~70-100 in cities.

Overwatch: With epic settings (and renderscale 150%) I get high GPU usage and high fps (200+) usually. This game seems to work a bit better I guess. If I recall correctly I can get higher fps than that if i lower gpu settings, so cpu seems to be able to not bottleneck as much here (at least its no problem for the cpu to achieve 200+ fps).

From what I've read most people seem to agree that a 3700x shouldnt bottleneck an rtx 3070 (at least not in any major fashion), but this is not my experience at all. The CPU cant even seem to reach 144 fps in the above titles with lowest settings (except overwatch), which I would think is not a unreasonable expectation today.

So either I've had some "bad luck" in the titles I've tried - meaning cyberpunk, warzone and WoW are all badly optimized games or the CPU actually does do some quite heavy bottlenecking if you want 144fps in certain games. Might add that I've also read some other worrying comments with people saying they switched from 3600x/3700x to more competent single core processors and have seen great improvements in warzone FPS (which also would suggest that it is actual dramatic cpu bottlenecking going on).

Any ideas? I'm quite confused by the fact that most people seem to say 3700x shouldnt bottleneck a rtx 3070 in any major way, but my (limited) experience is that it actually does, and quite heavily so if you want to achieve 144fps. What's your ideas and opinions, should I try some more games or just return the computer and buy another one with a better cpu?

The Zen 2 cores on the 3000 series are somewhat behind the fastest cpu's available for per core performance, which is important for high FPS at low resolutions. It's worth noting, an RTX 3070 is overkill for a 1080p screen, that is a 1440p / 4k card - also the RTX 3000 series struggle at lower resolutions compared to where they should be based on the specs (that is due to nVidia sharing half the core complex between int and FP functions whereas with the 2000 series the cards had separate units - the new setup boost performance at high res).

Will you get more out of a 3070 with a faster cpu? yes. However you need to look at the top tier of gaming cpu, either an Intel 10th Gen at 5+ ghz, Intel 11th Gen or the best option a Zen 3 based Ryzen 5000 series processor. The performance difference can be as much as 20% in some titles however in others it will be virtually nothing. It's also worth noting, even with the fastest gaming cpu on the market, you may not get 100% utilisation all the time with a 3070 at 1080p - the design of the card makes it difficult to properly utilise at 1080p. The Radeon 6000 series cards are the opposite - they tend to fly at 1080p and 1440p but fall behind Ampere at 4k so tend to be a better option for high fps / low resolution gaming. The driver overhead is also lower on the Radeon cards so they would perform proportionately better with a 'slower' cpu.

This vid from Hardware unboxed goes over the driver impact between nVidia and AMD cards with different cpu's... also this vid includes Cyberpunk as one of the tested games. This gives you an idea of the performance scaling you get with different cpu / gpu combos (looking at Cyberpunk looks like your performance isn't far off the mark to be honest)...

View: https://www.youtube.com/watch?v=G03fzsYUNDU


I think the main thing to keep in mind though is a 3700x is a decent gaming cpu. Wow is very poorly optimised for AMD (Blizzard software in general really doesn't perform well on AMD processors - the game engines are old and date back to when Intel dual / quad core cpu's were king, Starcraft 2 also performs sup optimally on Ryzen). Cyberpunk and Warzone both work well on either platform in those cases it's just you are a gen behind the very fastest option.

Out of interest what motherboard is in your OEM machine? If it's based on a B550 or X570 chipset chances are you could upgrade it to a Ryzen 5000 series cpu (e.g. 5600X) which should help.

 

[Gabriel3331]

@Afro_ninja199 ;

Lenovo 90RC00CJMW Performance Results - UserBenchmark

This link is clickable right? I'm no expert, but everything is looking ok in that benchmark as well right? (66th percentile compared to others with the same components and I didn't OC the gpu)

@cdrkf ;

Thank you very an excellent comment. What I'm taking from it;

1. It can indeed be a quite heavy cpu bottlenecking going on (compared to faster single core CPU:s) in certain games, you're saying up to 20%. I think this is something worth to keep in mind when people pretty much just say "no" to the question if a 3700x would bottleneck a rtx 3070. I just did quickly check some benchmark videos online and noted some other ppl with 3600x/3700x, getting around 105-110 in warzone, so maybe it's a normal fps for combo 3700x+rtx 3070. And when I watched some other benchmark video with a intel 10700k+rtx 3070, they have 130-140fps. I would almost say the difference could maybe be even greater than 20% unless something else is going on. I think this is understated because it can't uncommon for ppl to want to reach 144fps.
2. Very interesting point that 3 series are optimized for higher resolutions. While this is a highly relevant comment, it doesn't "solve" the issue of the 3700x not being able to achieve more than ~105 fps in certain scenarios (I mean, even if I had a 1440p monitor and I set the resolution to 1440p, it's not like I would jump to the 130-140 fps the 10700k got). I guess this means that you in many cases wont get more fps lowering to 1080p, but it probably (?) does not mean that you would get HIGHER fps at 1440p.

The mobo in the machine is some lenovo proprietary chipset is b550.

If this is actually the case it's quite unfortunate because I've seen quite a lot of ppl being confused by low fps with rtx 3070+ryzen 3600/3700. I suspect ppl trivialize the actual bottleneck a 3600/3700 can put on a 3070. And yes, I do agree it's ofc not the case for all games. Weird thing is that I haven't really seen these obvious difference in bigger benchmark comparisons. Maybe it's mainly a warzone and cyberpunk thing, lets not hope this is a trend then.. I almost feel liking returning it and go for a 10700k computer instead. But maybe I'll do some more research first.

 

[cdrkf]

@Afro_ninja199 ;

Lenovo 90RC00CJMW Performance Results - UserBenchmark

This link is clickable right? I'm no expert, but everything is looking ok in that benchmark as well right? (66th percentile compared to others with the same components and I didn't OC the gpu)

@cdrkf ;

Thank you very an excellent comment. What I'm taking from it;

1. It can indeed be a quite heavy cpu bottlenecking going on (compared to faster single core CPU:s) in certain games, you're saying up to 20%. I think this is something worth to keep in mind when people pretty much just say "no" to the question if a 3700x would bottleneck a rtx 3070. I just did quickly check some benchmark videos online and noted some other ppl with 3600x/3700x, getting around 105-110 in warzone, so maybe it's a normal fps for combo 3700x+rtx 3070. And when I watched some other benchmark video with a intel 10700k+rtx 3070, they have 130-140fps. I would almost say the difference could maybe be even greater than 20% unless something else is going on. I think this is understated because it can't uncommon for ppl to want to reach 144fps.
2. Very interesting point that 3 series are optimized for higher resolutions. While this is a highly relevant comment, it doesn't "solve" the issue of the 3700x not being able to achieve more than ~105 fps in certain scenarios (I mean, even if I had a 1440p monitor and I set the resolution to 1440p, it's not like I would jump to the 130-140 fps the 10700k got). I guess this means that you in many cases wont get more fps lowering to 1080p, but it probably (?) does not mean that you would get HIGHER fps at 1440p.

The mobo in the machine is some lenovo proprietary chipset is b550.

If this is actually the case it's quite unfortunate because I've seen quite a lot of ppl being confused by low fps with rtx 3070+ryzen 3600/3700. I suspect ppl trivialize the actual bottleneck a 3600/3700 can put on a 3070. And yes, I do agree it's ofc not the case for all games. Weird thing is that I haven't really seen these obvious difference in bigger benchmark comparisons. Maybe it's mainly a warzone and cyberpunk thing, lets not hope this is a trend then.. I almost feel liking returning it and go for a 10700k computer instead. But maybe I'll do some more research first.

Ryzen 3000 series are behind Intel 10000 (at least the higher 'K' versions which typically hit 5ghz without much hassle) in cpu bound gaming, although the difference depends very much on the specific scenario. In terms of performance at 1440p - no the card won't go faster, but at 1440p (especially at higher settings) then the performance between a 3700X and a 10700K will basically be nil. It's only really at lower resolutions with high FPS targets that you will see a difference. It's also slightly tricky to call it a flat out cpu bottleneck, as typically a 3700X won't be anywhere near capacity in these situations. The problem comes down to memory latency, the construction of Ryzen cpu's (3000 series and older) means that an 8 core cpu like a 3700X is actually constructed of two quad core 'CCX' modules with separate l3 cache, whilst Intel use a 'ring bus' design which gives all cores equal access to the L3. The result is that Zen 2 can fall behind the Intel part in situations like games where memory latency is critical. The internal connection in a Ryzen cpu is tied directly to the memory frequency, this is also why faster memory really benefits Ryzen as faster memory also increases the internal fabric clock of the cpu. This isn't an issue with newer Zen 3 parts as AMD unified the L3 cache between all 8 cores, meaning a reduction in core to core and core to memory latency and effectively doubles the available cache that can be used for a single application (even though the total is the same).

With regards to benchmark comparisons between cpus, the performance difference is very game dependent, with some titles working better with the architecture than others - when you look at large benchmark comparisons it tends to average out so that there isn't much difference between the two options (although the 10700K would still be ahead overall). It's also the case that typically the large tests don't focus on 'competitive' settings, usually if they test 1080p it's 1080p ultra which will run slower anyways, or they focus on 1440 or above.

What I would question, does your 144hz monitor support adaptive sync of any type (e.g. Freesync of G-Sync)? If yes have you tried running with that on? That won't boost the performance, however it will sync up the frames which avoids the issues of having the gpu and monitor refresh rates out of sync. I would think 100+ fps should be pretty fluid overall?

 

[punkncat]

IMO, your next move should be a 2K or better monitor. Running at 1080 isn't causing the GPU any stress, and there are quite a few accounts about poor 1080 performance with Ryzen. I didn't personally experience it, per se, but will say that when I moved to 2K I was seeing same/better performance than with a 1080p monitor.

Cyberpunk is a terrible game to rate system performance to. At least one of the titles you mention are an online game such that your performance can be effected (affected? lol) by your internet speed.
I don't see you mention what kind and how much RAM. Probably more helpful to us to list full specs. For instance, are you running off a HDD? Do you have a motherboard with VRM hearty enough to handle the 3700X? Are you monitoring temps?
There could be a lot more to this than just counting on the brute force of two of the components of the build without also accounting for the rest.

 

[hotaru.hino]

I had a look at some sample footage online for CoD: Warzone to see what the 3700X can do at 1080p low. Most of the time it floated around 120FPS though one person with a RTX 3070 stayed at ~140 FPS, but the GPU stayed at 50% load. The game also doesn't seem to use more than 8 threads given that most of the videos I saw didn't show a CPU usage of over 50% most of the time. Of course, this depends on the map and how many players there are:

In any case, if your goal is to have 100% GPU utilization, then you're going to need to step up the image quality output by cranking up the details to max and using DSR if needed. If your goal is to have the maximum performance possible, then I would argue the 3700X isn't the best processor for that.

 

[Gabriel3331]

Lets see;

@cdrkf
1. Yes, my monitor is g-sync compatbile, which I use. And yes, my roughly ~110fps or so in warzone, does feel pretty fluid. Might be the case that I tried games that are pretty bad to judge things from I mean with regards to what you're saying that, on average, the difference compared to a 10700k shouldn't be that big in general. I will see if I can try on some more games.
@punkncat
2. Going for a 1440p monitor would idd be on the list if I keep the computer. As you're pointing out, in a lot of games, I would expect the fps to be pretty much the same on 1440p as 1080p. That the performance (fps) would actually be higher on 1440 sounds a bit crazy, but if you say so Fair enough that cyberpunk is a bad game to judge from. I will try a few more games and see what I think. Your comment with regards to internet speed, when I have been mentioning performance I meant fps, I guess internet speed surely should not have an effect on fps?

Specs are:
3700x
rtx 3070
ram is samsung 3200mhz (unfortunately with cl 20 according to spec, and actually running even worse at 22 cl) - because of this i've overclocked it to run at 3333mhz and timings 19/19/19/40 - the overclock did bump up cpu score in 3dmark time spy a bit actually. And CPU z is reporting that it actually runs at that speed and those timings.
ssd nvme and hdd (warzone running from hdd, cyberpunk from ssd)
PSU 550w (not sure of brand but it has the info: 200-240Vac,550W 88% PSU )

Any other spec that would be relevant to know?

Temps are usually around 65-70c for both gpu and cpu in games, might rise to like 75c, but rarely above that. So I don't think it should be much throttling going on. Tho I read somewhere that ryzen actually starts throttling boost clocks at 70c (dunno if it's true). I personally don't think those kind of temps would have that noticeable performance loss (if any)? Unfortunately I can't crank up the cpu fan manually to see if it would help because of lenovos fairly limited bios (speedfan and such does not detect the cpu fan either, probably because of proprietary MB :/).

My personal guess (correct me if you think otherwise ) is that the PSU is probably good enough (if not, lenovo did a really poor job choosing that psu), temps are prolly good enough. The thing that could drag down the performance I guess is the ram since it has bad timings, but after my overclock with it running 3333mhz cl 19 it shouldn't be THAT bad right? The cpu does perform ok in benchmarks.

@hotaru.hino
Thanks for the video. I'm not that far off his fps for the ultra settings at 1080, maybe I'm 5-10 fps lower. Tho, I'm seeing no difference between ultra and low but it looks like he is actually getting some more fps on the low settings. That's pretty interesting, maybe his ram is better which helps the cpu not bottleneck as much?

And again, thank you very much for the discussion. I do feel like I need to try a few more games, and maybe it's not that bad of an idea to keep the computer and get a 1440p monitor. I just checked a bit and the offerings for 10700k+rtx3070 computers were actually a fair bit more expensive here in sweden.

I would have considered building my own pc if it was not for the gpu situation.

 

[punkncat]

@punkncat
Your comment with regards to internet speed, when I have been mentioning performance I meant fps, I guess internet speed surely should not have an effect on fps?

Specs are:

PSU 550w (not sure of brand but it has the info: 200-240Vac,550W 88% PSU )


My personal guess (correct me if you think otherwise ) is that the PSU is probably good enough

When dealing with online based games/multiplayer your internet connection will directly impact your experience.

and:
"NVIDIA recommends a 650W PSU for the RTX 3070, but if you're currently working with, say, a 600W PSU, you might get along just fine depending on other hardware in your system. "

That PSU can not be helping at all....and you totally need to replace that RAM. Ryzen are very sensitive to RAM speeds and timing. That is high and more suited for a business class computer.

 

[hotaru.hino]

and:
"NVIDIA recommends a 650W PSU for the RTX 3070, but if you're currently working with, say, a 600W PSU, you might get along just fine depending on other hardware in your system. "

NVIDIA recommends something high to begin with to cover all bases. The RTX 3070 appears to float around 220-230W when running a heavy load. Tack on maybe another 80W for the 3700X at worst case and 50W for the rest of the system. We're sitting at around 360W. Assuming the PSU unit is a good quality one and most of the rated wattage comes from the 12V rail, this is roughly 65% usage.

I have a similarly specc'd system in terms of expected power consumption and I maybe see it go up to 330W from the wall.

That PSU can not be helping at all....and you totally need to replace that RAM. Ryzen are very sensitive to RAM speeds and timing. That is high and more suited for a business class computer.

Impact of Memory Timings on Gaming Performance in Ryzen 3000 CPUs: In-Depth Benchmarks | Hardware Times

As far as memory or RAM goes, few people know how important it is in terms of actual real-world performance. We already did a detailed post showing the impact of memory speeds and bandwidth on gaming performance a while back. In this post, we have a look at the impact of memory timings on gaming …

www.hardwaretimes.com

RAM timing doesn't appear to factor much in performance. At least between the standard JEDEC timing and what you can find on the market.

 

[punkncat]

^ I went to a power supply calculator and made some assumptions on the build and was surprised to see the recommended unit under 450W. I would not have considered that system to be so efficient.

 

[mdd1963]

The 3700X is not going to match the 5600X in 1080P gaming, much less the 5800X or 5900X.

There will be less of a difference at 1440P, and, almost no difference at 4k, when/where 4k gaming is possible with a 3070.

You must pour over comparisons to decide if the cost is worth it in your chosen gaming usage scenarios.

 

[Gabriel3331]

Thanks for the replies guys!

I take it you agree that the PSU is probably not an issue. I mean, if it was, it would've just shut down the PC (which it hasn't). It is my understanding that a PSU/the components can't down scale the power they use depending on the PSU, either it works or the PC shuts down.

Good info that the ram timings probably is not that big of an issue.

I guess that my system is probably not that far off the performance it should have, maybe maybe the cpu is running a tad bit hot at times to ensure the highest boost clocks, or my chip is not the best (I usually don't see it reach the advertised boost clock of 4400mhz, it rather just goes up to 4200-4300mhz). But since it gets decent scores in cinebench and at least OK in the cpu part of the 3dmark time spy, I'm guessing it can't be that bad.

So, what I take from all of this, if I in certain games (for example warzone) want to reach high fps (144fps), a better CPU would actually help me quite a lot (for example a 10700k or a 5800x). So a x3700 does actually do quite a bit of bottlenecking in certain games if you want to 144fps (I haven't got a very good idea how common this is tho), in other games it wont.

I haven't had time to try out that too many other games, at least I tried quickly in apex legends, and there it was no issue for the cpu getting 144 fps. As pointed out by @mdd1963 , there is no way around it, it depends on the gaming scenarios of your choice.

Anyway, I think I will keep the computer since I did some looking around and the computers available with 3070 + a more competent CPU (10700k+, or 5800x+) did actually cost a quite bit more. Since chipset is B550 I can always upgrade in the future if I feel I need to.

And as pointed out, getting a 1440p monitor, so that I can actually utilize the GPU a bit more , is probably going to be my next move!

I'm new to the forum, I should probably mark the thread as solved but I can't figure out how :S

 

[OllympianGamer]

Thanks for the replies guys!

I take it you agree that the PSU is probably not an issue. I mean, if it was, it would've just shut down the PC (which it hasn't). It is my understanding that a PSU/the components can't down scale the power they use depending on the PSU, either it works or the PC shuts down.

Good info that the ram timings probably is not that big of an issue.

I guess that my system is probably not that far off the performance it should have, maybe maybe the cpu is running a tad bit hot at times to ensure the highest boost clocks, or my chip is not the best (I usually don't see it reach the advertised boost clock of 4400mhz, it rather just goes up to 4200-4300mhz). But since it gets decent scores in cinebench and at least OK in the cpu part of the 3dmark time spy, I'm guessing it can't be that bad.

So, what I take from all of this, if I in certain games (for example warzone) want to reach high fps (144fps), a better CPU would actually help me quite a lot (for example a 10700k or a 5800x). So a x3700 does actually do quite a bit of bottlenecking in certain games if you want to 144fps (I haven't got a very good idea how common this is tho), in other games it wont.

I haven't had time to try out that too many other games, at least I tried quickly in apex legends, and there it was no issue for the cpu getting 144 fps. As pointed out by @mdd1963 , there is no way around it, it depends on the gaming scenarios of your choice.

Anyway, I think I will keep the computer since I did some looking around and the computers available with 3070 + a more competent CPU (10700k+, or 5800x+) did actually cost a quite bit more. Since chipset is B550 I can always upgrade in the future if I feel I need to.

And as pointed out, getting a 1440p monitor, so that I can actually utilize the GPU a bit more , is probably going to be my next move!

I'm new to the forum, I should probably mark the thread as solved but I can't figure out how :S

On the video you showed where every core is working but usage hits a cap of 50%, that would suggest to me multithreading/smt is possibly turned off, could be in the bios, A ryzen software or powerplan issue.

 

[Gabriel3331]

On the video you showed where every core is working but usage hits a cap of 50%, that would suggest to me multithreading/smt is possibly turned off, could be in the bios, A ryzen software or powerplan issue.

Hello!

I'm not sure which video you are refeering to? In my case, SMT is turned on, I have the latest chipset driver and I'm running ryzen powerplan balanced (have tried a few other power plans without any noticable difference).

When monitoring performance I can see;

During cincebench single core the cpu reaches like 4350mhz which I guess is a quite normal boost clock for the 3700x (I've seen people reach 4400+ so might not be the absolute best chip in my case, but still ok)

During cinebench multi core the cpu reaches around 4050-4100mhz on all cores, which I also guess is quite normal.

Not sure if I misunderstood your comment? Either way, thanks for the input!

 

[OllympianGamer]

Hello!

I'm not sure which video you are refeering to? In my case, SMT is turned on, I have the latest chipset driver and I'm running ryzen powerplan balanced (have tried a few other power plans without any noticable difference).

When monitoring performance I can see;

During cincebench single core the cpu reaches like 4350mhz which I guess is a quite normal boost clock for the 3700x (I've seen people reach 4400+ so might not be the absolute best chip in my case, but still ok)

During cinebench multi core the cpu reaches around 4050-4100mhz on all cores, which I also guess is quite normal.

Not sure if I misunderstood your comment? Either way, thanks for the input!

Sorry i thought the video hoturohino posted was your gameplay for some reason hahaha my bad. I do actually have a 3700x in a new computer, I could do a few tests if you want to compare some numbers.

 

[dorsai]

RAM timing doesn't appear to factor much in performance. At least between the standard JEDEC timing and what you can find on the market.

Ram timing plays a major role in the performance you can get from 3000 series Ryzen...especially at the 1% low fps. The Hardwaretimes article is clearly flawed since the lowest timings they ran are actually quite loose at 16-16-16-36 not to mention that they actually manage to lose performance at 3600mhz vs cas20 ram so something is clearly not right in their testing. Feeding 3000 series Ryzen some proper cas14 memory at clock speeds above 3200mhz will wake that chip right up.

View: https://www.youtube.com/watch?v=FB_8zXW46Jw


Almost always low GPU utilization comes from running the screen resolution too low or the in game graphic settings too low. In the case of the 3070 I don't think the card is wasted at all at 1080p...a game like Cyberpunk will even bring a 3080 to it's knees at 1080p. Someone playing competitive games will be much better off at 1080p and 200+ fps vs someone at 1440p getting 120 fps. While a 3070 is more than capable of running at 1440p I don't think it's accurate to say the card isn't being pushed at 1080p...it all depends on the game and the settings and the fps the user is looking for.

 

[Gabriel3331]

@OllympianGamer ; sure! I can message you some numbers when I have a bit more time and we can compare!

@dorsai ; wow, crazy video. That probably means my badly timed ram could be dragging my performance down in some cpu bound games. It also explains why I get a decent score in cinebench where I believe ram timings are less relevant.

If I would see some decent 3200mhz+ cl 16 on sale it could potentially be a fairly cheap upgrade. In that case I gotta figure out what this lenovo MB actually supports, all the listed supported ram on their website are crappy 20cl ram.

 

[cdrkf]

@OllympianGamer ; sure! I can message you some numbers when I have a bit more time and we can compare!

@dorsai ; wow, crazy video. That probably means my badly timed ram could be dragging my performance down in some cpu bound games. It also explains why I get a decent score in cinebench where I believe ram timings are less relevant.

If I would see some decent 3200mhz+ cl 16 on sale it could potentially be a fairly cheap upgrade. In that case I gotta figure out what this lenovo MB actually supports, all the listed supported ram on their website are crappy 20cl ram.

Memory support with Ryzen is complex - the actual memory controller is built onto the CPU so you often find you can run memory outside of the official spec for the motherboard. That said, the motherboard does have an impact on memory performance due to factors like length of traces on the board, number of layers used and such so it can be tricky to know for sure. If you get a faster kit, you can always loosen timings to improve stability if necessary.

 

[Gabriel3331]

Memory support with Ryzen is complex - the actual memory controller is built onto the CPU so you often find you can run memory outside of the official spec for the motherboard. That said, the motherboard does have an impact on memory performance due to factors like length of traces on the board, number of layers used and such so it can be tricky to know for sure. If you get a faster kit, you can always loosen timings to improve stability if necessary.

My fear is mainly what the bios will do. This bios has some weird "memory overclock" option which I'm not completely sure what it does (it seems to only adjust the clock speed to SPD rather than ddr4 default but not adjust the timings), and there are no available manual configuration of your ram in bios. So I had to use ryzen master to overclock my ram which I guess is not ideal but it at least works. The issue I can see is that I can't set ram voltage from ryzen master or the bios. And pretty much all "gaming oriented" ram sticks are assumed to run at 1,35v instead of the 1,2v ddr4 default. My current ram is running at 1,2 and I'm not sure if that is because bios is using the rams SPD (which is 1,2v) to determine that or if it's simply because it's running at ddr4 default.

Long story short, I'm not sure if my bios will let me run for example a 3600mhz cl16 v 1,35v at those advertised specs (because of the limited ram overclock options available). I've asked the question on the lenovo forums, we'll see if I get a decent answer or not. Otherwise if I find a good deal on some memory, I might just buy it and try and return it if it doesn't work.