3770K bottlenecking Vega64?

[cucker tarlson]

How recent are those benchmark results? I get MUCH higher fps that that with a Ryzen system.. more like 130-140fps most the time at 1440P.

Last 5 mins of a game, skip to 1min in, i didn't intend to show this video, just uploaded to show a mate, but you can see the fps:

Settings:

View attachment 110721

edit- and a 163fps cap in RTSS

R7 1800X @ 4.1ghz, RAM @ 3333mhz CL14, 1080Ti @ 2076/6200mhz (sry, not a vega, but this should still be valid to show platform potential of ryzen)

1. 3333 cl14 ram
2.yes,having a 1080ti is a game changer here, if it's a dx11 game that is single core bound
3.you have bigger fps numbers cause cpu tests are done in places and scenarious where cpu is the limiting factor,you can't just record a minute of gameplay and say there's no bottleneck

 

[Mussels]

I only very recently sold my 3770k, you need to overclock your ram (or add more)
The CPU's themselves have aged well, its stock DDR3 (1333Mhz) that kills them, if you can get above 1866Mhz (I ran DDR3 2400) they can feed the GPU a lot easier


At the time i was running 4K on a GTX1080 with a 3770k at 4.8Ghz, and went from 1600 to 2400 and noticed a huge boost in performance

 

[TheGuruStud]

I only very recently sold my 3770k, you need to overclock your ram (or add more)
The CPU's themselves have aged well, its stock DDR3 (1333Mhz) that kills them, if you can get above 1866Mhz (I ran DDR3 2400) they can feed the GPU a lot easier


At the time i was running 4K on a GTX1080 with a 3770k at 4.8Ghz, and went from 1600 to 2400 and noticed a huge boost in performance

I'll try this if it's possible with 4 dimms (should be at 1,600 currently).

 

[Joss]

At the time i was running 4K on a GTX1080 with a 3770k at 4.8Ghz, and went from 1600 to 2400 and noticed a huge boost in performance

Did you also increase total amount of RAM?

 

[Mussels]

Did you also increase total amount of RAM?

No, 16GB at both speeds. OP only has 8GB, so 16GB of faster ram would definitely be an improvement in many ways.

 

[cucker tarlson]

Frankly, @oli_ramsay just do what mussels suggests for now. 3770K is very capable if paired with fast memory and overclocked. Ryzen 2600x with ddr4 3200 would be faster,but compared to what it would cost to change the whole platform and what you'd get in return in this particular game,I doubt it's worth it. It would be a huge upgrade for multi threaded games, but for single core bound ones - not that much if it stands against an overclocked 3770K with good ddr3. ddr3 can still deliver very good performance against ddr4 due to low latency, look at ddr3 2400 cl11 vs ddr4 3200 cl16 in single thread bound fc4

https://www.purepc.pl/pamieci_ram/t...pamieci_ram_wybrac_do_intel_skylake?page=0,13

Ivy can take 2133MHz easily, so buy a 2x8 2133 CL9/10 kit and find a cheap working mobo to push that CPU anther 700MHz, you'll be more than happy.

 

[Joss]

I find it hard to believe system RAM speed affects game FPS.
The GPU works in real time with the VRAM, not the system's; the latter only feeds the VRAM with data.
Under specific scenarios (APUs that use system RAM, Ryzen chips that have performance tied up with system RAM) or under specific games that may, somehow, be optimized for faster RAM and timings... but for most games out there with dedicated graphics cards I don't think system RAM speed makes a difference.

 

[TheGuruStud]

I find it hard to believe system RAM speed affects game FPS.
The GPU works in real time with the VRAM, not the system's; the latter only feeds the VRAM with data.
Under specific scenarios (APUs that use system RAM, Ryzen chips that have performance tied up with system RAM) or under specific games that may, somehow, be optimized for faster RAM and timings... but for most games out there with dedicated graphics cards I don't think system RAM speed makes a difference.

Feel free to check even new intel CPUs (which have very robust mem controllers already) with varying speeds. It helps a decent amount.

 

[infrared]

1. 3333 cl14 ram
2.yes,having a 1080ti is a game changer here, if it's a dx11 game that is single core bound
3.you have bigger fps numbers cause cpu tests are done in places and scenarious where cpu is the limiting factor,you can't just record a minute of gameplay and say there's no bottleneck

You have to see why I'm sceptical of those results, according to the pic you uploaded I should be seeing 106fps in 1080P Very low, which is horrific, and in reality I'm sometimes sat on the 163fps limit and usually 130-140 at 1440P... It's pretty misleading. That's why I linked to a video so people can see for themselves instead of looking at a bar graph that could be fairly skewed if the tester is biased in any way.
3333-3466 is achievable on nearly every zen+ cpu and a lot of first gens, i don't see why having good ram disqualifies my experience. Also, amusing side note: this was on an Asus prime B350 motherboard worth 80 quid new.

Even with a 1080ti it's still dx11 and as you said, single core limited.. Were you thinking the game runs in dx12 on an nv card or something? I don't understand your 2nd point.

 

[cucker tarlson]

Even with a 1080ti it's still dx11 and as you said, single core limited.. Were you thinking the game runs in dx12 on an nv card or something? I don't understand your 2nd point.

cpu overhead in dx11 on games that are single thread bound, I don't think AMD fixed it.

I find it hard to believe system RAM speed affects game FPS.

Do you believe:
1. faster cpu produces more fps (when the gpu is capable of pushing those extra frames) ?
2.putting faster memory on a graphics card boosts fps ?

then why do you find it hard to believe cpu needs fast ram too ?

 

[kastriot]

Wait for Zen 2 and then decide.

 

[infrared]

cpu overhead in dx11 on games that are single thread bound, I don't think AMD fixed it.

I get that, but dunno how it relates to my earlier posts Seemed like you were saying that single thread limitation didn't apply to me because of the 1080ti or something.

2.yes,having a 1080ti is a game changer here, if it's a dx11 game that is single core bound

.
Doesn't really matter much anyways, I wouldn't try to say AMD is right up there with Intel in this game, there is definitely an fps penalty for AMD (currently anyway, zen2 could be interesting). I just get slightly triggered when I see charts showing absolutely abysmal performance for AMD when that doesn't match up with a typical real world experience.

 

[cucker tarlson]

I get that, but dunno how it relates to my earlier posts Seemed like you were saying that single thread limitation didn't apply to me because of the 1080ti or something.

.
Doesn't really matter much anyways, I wouldn't try to say AMD is right up there with Intel in this game, there is definitely an fps penalty for AMD (currently anyway, zen2 could be interesting). I just get slightly triggered when I see charts showing absolutely abysmal performance for AMD when that doesn't match up with a typical real world experience.

and you blame me ? I did those test apparently ?

even TPU reports stock 8600k beating 2600x by 25% in single thread bound games like Far Cry, and you get triggered about me posting such results ?

 

[EarthDog]

Wouldn't the CPU utilisation need to be 100% before it would "bottleneck" your GPU?

Nope.

And they thermally throttle too, those CPUs!!



Being serious, at 1440p that cpu wont hold vega64 back. At 1080p, with the cpu at stock, it probably shaves a couple-few percent off depending on the title and settings.

 

[infrared]

and you blame me ? I did those test apparently ?

No of course not, not blaming you for anything. All I was wanted was to show that ryzen isn't as useless at gaming as the chart you posted would suggest.
I'm done anyway, not trying to start an argument or derail the topic. I actually agreed with your advice to the OP earlier on.

edit -

even TPU reports stock 8600k beating 2600x by 25% in single thread bound games like Far Cry, and you get triggered about me posting such results ?

I don't play far cry 5, so have no idea how my system would perform in it, I trust the tpu reviews to be unbiased so that result is probably accurate.. I just know from much time spent in the game that pubg performs substantially better on ryzen than the chart you posted earlier.

 

[cucker tarlson]

No of course not, not blaming you for anything. All I was wanted was to show that ryzen isn't as useless at gaming as the chart you posted would suggest.
I'm done anyway, not trying to start an argument or derail the topic. I actually agreed with your advice to the OP earlier on.

look at my previous post. apparently this is going nowhere since you can't understand where I'm getting at. no one said ryzen is useless,neither did I,but it has shortcomings that even your own site points out yet somehow you get triggered when raising the doubt over the very same issue.

 

[TheGuruStud]

No of course not, not blaming you for anything. All I was wanted was to show that ryzen isn't as useless at gaming as the chart you posted would suggest.
I'm done anyway, not trying to start an argument or derail the topic. I actually agreed with your advice to the OP earlier on.

edit -

I don't play far cry 5, so have no idea how my system would perform in it, I trust the tpu reviews to be unbiased so that result is probably accurate.. I just know from much time spent in the game that pubg performs substantially better on ryzen than the chart you posted earlier.

Real world is FAR different from the bench tools included with games. Those are basically worthless, now. Real testing has to be done in game with several similar runs averaged. Tomb raider results, anyone?

 

[infrared]

Real world is FAR different from the bench tools included with games. Those are basically worthless, now. Real testing has to be done in game with several similar runs averaged. Tomb raider results, anyone?

Very true. I need to get afterburner/rivatuner set up to work out the average and 1% lows. I'll record a few full games and see how I get on, I'm pretty curious about this now. I'll do some War mode games (deathmatch), that would be a good worst case scenario test and consistent length games.

 

[cucker tarlson]

Real world is FAR different from the bench tools included with games. Those are basically worthless, now. Real testing has to be done in game with several similar runs averaged. Tomb raider results, anyone?

we need reviewers to do at least 1 hr long runs of each game

 

[EarthDog]

No of course not, not blaming you for anything. All I was wanted was to show that ryzen isn't as useless at gaming as the chart you posted would suggest.
I'm done anyway, not trying to start an argument or derail the topic. I actually agreed with your advice to the OP earlier on.

edit -

I don't play far cry 5, so have no idea how my system would perform in it, I trust the tpu reviews to be unbiased so that result is probably accurate.. I just know from much time spent in the game that pubg performs substantially better on ryzen than the chart you posted earlier.

part of the problem is that is only an online game. Experience and fps can vary every time you test.

As far as in game benchmarks.... they are better than manual run through a. They are repeatable is one huge reason. Most do the exact same thing every time. I dont use them so much as how many fps the game gets, but when compared amongst each other, will translate in game, the pecking order. Typically fps isnt very far off though between the game and bench.

 

[TheGuruStud]

part of the problem is that is only an online game. Experience and fps can vary every time you test.

As far as in game benchmarks.... they are better than manual run through a. They are repeatable is one huge reason. Most do the exact same thing every time. I dont use them so much as how many fps the game gets, but when compared amongst each other, will translate in game, the pecking order. Typically fps isnt very far off though between the game and bench.

They're very simple and scripted. CPU isn't used and often only one scene. I mentioned TR, b/c AMD does very well in game, but it looks bad in the tool compared to nvidia (the one with the snowy mountain scene).

Repeatable isn't valuable when they produce erroneous results. Some good Youtubers have moved to in game to display real world results. Following predefined paths and actions makes it repeatable and averaging runs makes it accurate.

Those always get some fanboy panties in a twist. Remember GTA V with ryzen vs 7700K? Real world had 7700k choking and stuttering in sections.

 

[eidairaman1]

I get that, but dunno how it relates to my earlier posts Seemed like you were saying that single thread limitation didn't apply to me because of the 1080ti or something.

.
Doesn't really matter much anyways, I wouldn't try to say AMD is right up there with Intel in this game, there is definitely an fps penalty for AMD (currently anyway, zen2 could be interesting). I just get slightly triggered when I see charts showing absolutely abysmal performance for AMD when that doesn't match up with a typical real world experience.

Pm me.

 

[EarthDog]

They're very simple and scripted. CPU isn't used and often only one scene. I mentioned TR, b/c AMD does very well in game, but it looks bad in the tool compared to nvidia (the one with the snowy mountain scene).

Repeatable isn't valuable when they produce erroneous results. Some good Youtubers have moved to in game to display real world results. Following predefined paths and actions makes it repeatable and averaging runs makes it accurate.

Those always get some fanboy panties in a twist. Remember GTA V with ryzen vs 7700K? Real world had 7700k choking and stuttering in sections.

in testing we have run, the longer the better for repeatable results when running manually. Again, a number should be taken away from reviews, the experience should also be shared as well... a subjective part of it.

Sadly, not all games can have repeatable sections, try as one may. You can use an on rails section, but then you get I to if that is a realistic part of the game. There is a fine line between an in game bench being valuable and useless... but most give a more than fair representation of results. Better than any manual runs that arent horrifically resource (time) intensive.

 

[TheGuruStud]

in testing we have run, the longer the better for repeatable results when running manually. Again, a number should be taken away from reviews, the experience should also be shared as well... a subjective part of it.

Sadly, not all games can have repeatable sections, try as one may. You can use an on rails section, but then you get I to if that is a realistic part of the game. There is a fine line between an in game bench being valuable and useless... but most give a more than fair representation of results.

True, some are all over the place and I wish that reviewers expounded on the experience (this was a big deal when ryzen was first released, b/c even though avg FPS was lower, they said it was much smoother). Maybe I'm anal, but I find using one metric fairly useless

Lows don't seem to be talked about as much as they were fairly recently. Tweaked Vega is murdering nvidia in value. You can get a vega 56 to produce the same lows as 1080ti/2080. I haven't gamed on either of those two cards, but when I see lows of 60 to highs of 140....it makes me question how smooth that is. I would absolutely notice those drops.

 

[infrared]

True, some are all over the place and I wish that reviewers expounded on the experience (this was a big deal when ryzen was first released, b/c even though avg FPS was lower, they said it was much smoother). Maybe I'm anal, but I find using one metric fairly useless

Lows don't seem to be talked about as much as they were fairly recently. Tweaked Vega is murdering nvidia in value. You can get a vega 56 to produce the same lows as 1080ti/2080. I haven't gamed on either of those two cards, but when I see lows of 60 to highs of 140....it makes me question how smooth that is. I would absolutely notice those drops.

Yeah, can't really argue with the value of the V56s!

I gave up trying to benchmark on war mode games, every time you respawn it was dropping the low fps to almost 0 lol, and while skydiving I was getting 200-300fps so results weren't even remotely reliable.
.
.
I just got a chicken dinner on Sanhok, 22 mins of gameplay and here's the results from rivatuner, started the benchmark when I was almost at the ground:

17-11-2018, 22:13:35 TslGame.exe benchmark completed, 169417 frames rendered in 1348.672 s
Average framerate : 125.6 FPS
Minimum framerate : 46.3 FPS
Maximum framerate : 162.2 FPS
1% low framerate : 44.1 FPS
0.1% low framerate : 10.8 FPS

I'm uploading the corresponding video of the game. I agree about the lows, 46fps does look pretty bad, though they really are just an occasional stutter and it's silky smooth most of the time as you'll be able to see in the video shortly (eta ~1hr). I had a few nasty lag spikes too which may have contributed to the lows. Recorded with geforce experience at 1440P 50MBit so may have reduced performance very slightly. /excuses!



An earlier short game averaged 118fps. I guess it's not miles off from the 106fps ave on the chart @cucker tarlson posted, lower than I expected tbh. (apologies for being a pita )


Vid for validation, it's not a particaularly interesting game, if you decide to watch, fights are at 8:00, 10:40, 19:10, 21:10 and 22:05. The video should have a 1440P option soon so you can see the overlay (apologies, it's quite small, hard to see at 1080P, and quality is pretty garbage)