I9 9900k + 4070ti same FPS as 3070

[QuantumSingularity]

Well that 9900k is so ancient that it's slower than basically any current gen CPU, so you can't go wrong with the latest gen on both fronts. I think only the 13100F is a bit slower, but only a bit.

[starsmine]

8 minutes ago, QuantumSingularity said:

Well that 9900k is so ancient that it's slower than basically any current gen CPU, so you can't go wrong with the latest gen on both fronts. I think only the 13100F is a bit slower, but only a bit.

I swear yall need to look at the benchmarks sometimes. stop telling people to spend money when it wont improve anything. a hypothetical infinitely fast CPU will only get him from 109 fps avg to 112 fps average at native 1664p with his chosen settings.

[QuantumSingularity]

1 minute ago, starsmine said:

I swear yall need to look at the benchmarks sometimes. 

I don't have to look at any benchmarks, i've upgraded more than 10 of those CPUs in the last 6 months. Yes, at 4k it;s still able to somewhat holdup, but at 1440p and 1080p often it gets smashed by a $130 5600X.

[starsmine]

3 minutes ago, QuantumSingularity said:

I don't have to look at any benchmarks, i've upgraded more than 10 of those CPUs in the last 6 months. Yes, at 4k it;s still able to somewhat holdup, but at 1440p and 1080p often it gets smashed by a $130 5600X.

damn man, 109 fps unplayable, all my homies play at 112fps with that sweet sweet unobtanium branded CPU (at 1664p)

[QuantumSingularity]

2 minutes ago, starsmine said:

damn man, 109 fps unplayable, all my homies play at 112fps. 

109 FPS isn't bad, but 180+ avg FPS sure is more better. Even in slow paced game as COD.

[starsmine]

1 minute ago, QuantumSingularity said:

109 FPS isn't bad, but 180+ avg FPS sure is more better.

fuck dude. the GPU cant do that dude. its not the CPU binding it there. Please just read the whole single page of the thread and look at the benchmarks before just making shit up. 

[QuantumSingularity]

I looked through it and saw that for some reason people still think of the 9900K as a modern CPU. It's from 2018... It's like asking why your ryzen 2700X isn't keeping up well lately. 

[starsmine]

25 minutes ago, QuantumSingularity said:

I looked through it and saw that for some reason people still think of the 9900K as a modern CPU. It's from 2018... It's like asking why your ryzen 2700X isn't keeping up well lately. 

So your reasoning was literally "I made up my performance numbers" 
again dude. 

take out the fact that his 9900k is underperforming vs what a 9900k should



Replace CPU avg, low 5%, and low 1% with ∞ sign, as in unobtanium, infintinyly fast CPU, replace cpu bottleneck with 0%

What do you think the Achieved results would be?

Would it match 1to1 with the GPU?

Oh, right, it would. 
overall avg would be 112, 5% would 79, 1% low would be 63.

18% cpu bottleneck means for 18% of frames generated, CPU finished its frame behind the gpu, it does not say how much behind, It could be a ns, it could be a µs, hell, it could be a ms. does not matter in terms of the number that precedes %

back on topic, its already been established that for some reason, the 9900k is not performing the way it should. that issue has not been diagnosed, but we got to a point where for MW2, it does not matter as it is fast ENOUGH for that resolution to not in any meaningful or significant way bring down fps. OH NO a 1% low of 61 rather then 63, however will I make my shots now. 

Lets go back to 1440p

some magical non existent infinitely fast CPU would only take the fps from 112 to 122. and bring up the 1% lows from 67 to 71. To have the audacity to say, the solution is to spend 500+ dollars on a new mobo, new cpu, and new ram, for that kind of performance uplift because you refuse to look at the benchmarks is wild when we KNOW with certainty that there is a different issue at play here because of those bench marks. 

WHY is overall less then each part you might ask

Lets take two d20s. lets roll them 1000s of times.
Each one has a bottom 10% roll of 2.
Overall you pick the LOWEST of the two rolls
aka, your overall bottom 10% is 1.xsomething, Im not going to do the exact math for you. its a bayesian curve
The bottleneck % is just telling you which dice rolled lower and how often. not by how much.

you could technically have a hypothetical 99% cpu bind, and swap out the CPU and only gain 1% fps because again, that bind % does not tell you by how much. this hypothetical only has the CPU be slower 99% of the time, by nanoseconds.

[igormp]

2 hours ago, Cpt_Poopie_Pants said:

I recently got a ASUS OC edition 4070ti and in COD MW2 my FPS is the exact same as when I had my 3070. 

I play on ultra settings at 1440p

 

 

Current build: I9 9900K 32 GIG RAM @3200, ASUS Z390E, Kraken AIO, corsair 1k watt PS. 

Have you checked for temps on your 9900k?

 

Also, from what I've seen, COD MW2 really likes high frequency ram with low latency, so maybe trying to OC your RAM and tightening your timings may improve your results at lower resolutions.

[starsmine]

18 minutes ago, igormp said:

Also, from what I've seen, COD MW2 really likes high frequency ram with low latency, 

That checks out seeing how massive x3d/3dvcache boosts fps. 

5900 -> 5800x3d looks to boost CPU frame rates from 1% lows from 145 to 187. 

[igormp]

7 minutes ago, starsmine said:

That checks out seeing how massive x3d/3dvcache boosts fps. 

5900 -> 5800x3d looks to boost CPU frame rates from 1% lows from 145 to 187. 

I've heard from people with a 4090 and a 13900k that managed to get over 50 FPS just from going for an extreme ram OC.

[QuantumSingularity]

1 hour ago, starsmine said:

So your reasoning was literally "I made up my performance numbers" 
again dude. 

take out the fact that his 9900k is underperforming vs what a 9900k should



Replace CPU avg, low 5%, and low 1% with ∞ sign, as in unobtanium, infintinyly fast CPU, replace cpu bottleneck with 0%

What do you think the Achieved results would be?

Would it match 1to1 with the GPU?

Oh, right, it would. 
overall avg would be 112, 5% would 79, 1% low would be 63.

18% cpu bottleneck means for 18% of frames generated, CPU finished its frame behind the gpu, it does not say how much behind, It could be a ns, it could be a µs, hell, it could be a ms. does not matter in terms of the number that precedes %

back on topic, its already been established that for some reason, the 9900k is not performing the way it should. that issue has not been diagnosed, but we got to a point where for MW2, it does not matter as it is fast ENOUGH for that resolution to not in any meaningful or significant way bring down fps. OH NO a 1% low of 61 rather then 63, however will I make my shots now. 

Lets go back to 1440p

some magical non existent infinitely fast CPU would only take the fps from 112 to 122. and bring up the 1% lows from 67 to 71. To have the audacity to say, the solution is to spend 500+ dollars on a new mobo, new cpu, and new ram, for that kind of performance uplift because you refuse to look at the benchmarks is wild when we KNOW with certainty that there is a different issue at play here because of those bench marks. 

WHY is overall less then each part you might ask

Lets take two d20s. lets roll them 1000s of times.
Each one has a bottom 10% roll of 2.
Overall you pick the LOWEST of the two rolls
aka, your overall bottom 10% is 1.xsomething, Im not going to do the exact math for you. its a bayesian curve
The bottleneck % is just telling you which dice rolled lower and how often. not by how much.

you could technically have a hypothetical 99% cpu bind, and swap out the CPU and only gain 1% fps because again, that bind % does not tell you by how much. this hypothetical only has the CPU be slower 99% of the time, by nanoseconds.

First of all we are talking about COD MW2 - a game notorious for how much it likes its fast, low latency memory. The only game i can think of which has a more massive boost from the 3D cache of the 5800X3D is ACC. And the 2666MHz DDR4 the 9900k is bound with isn't exactly cutting edge. PCIe 3.0 vs PCIe 4.0 on its own on the exact same system gives 3-12% difference depending on the configuration. CPU improvement isn't only in die size or number of transistors or boost clocks anymore. It's a package of things, depending on the architecture features. Case and point - the 5800X3D. The old GTX 1070 i used was already utilized at 99% all the time and was producing about 50 FPS on average with the 5600X. Simply changing the CPU to 5800X3D basically doubled my FPS in most of the cases, while the 1070 was again at 99% all the time. The interactions between CPU and GPU are becoming more and more complex and it's not as simple as "you do this and i'll do that". Features like Re-Sizable BAR, more PCIe lanes at a higher standard for faster transfer between CPU-storage-GPU, faster memory, more IPC - all stuff that the 9900K lacks compared to any modern CPU.

[Cpt_Poopie_Pants]

https://valid.x86.fr/4nf2ek

 

Link is for CPUID Bench

[igormp]

24 minutes ago, Cpt_Poopie_Pants said:

https://valid.x86.fr/4nf2ek

 

Link is for CPUID Bench

Uh, looks like your ram is running in single channel, did you plug then into the right slots?

[Cpt_Poopie_Pants]

for some reason it wont let me run dual. if i plug in the ram into the slot it wont post. ive tried multiple ram sticks all brand new and i cannot figure it out for the life of me.

[Prometheus935]

11 minutes ago, Cpt_Poopie_Pants said:

for some reason it wont let me run dual. if i plug in the ram into the slot it wont post. ive tried multiple ram sticks all brand new and i cannot figure it out for the life of me.

Do you have XMP enabled?

[starsmine]

1 hour ago, Cpt_Poopie_Pants said:

https://valid.x86.fr/4nf2ek

 

Link is for CPUID Bench

The other thing I noticed is you are not running it at full XMP. those sticks are capable of 3600MT/S and you have them running at 3200MT/s
The memory timings are the XMP timings (and are out of jdec spec for 3200MT/s so XMP is partially enabled)

 

So cool, we likely found our culpret. Lets try to fix the single channel issue first, and if we feeling spicy, fix the xmp clocks back to 3600.

now these sticks are not technically on the QVL for the Mobo, BUT, (that usually can just be ignored because new sticks come out faster then MOBO manufactures can test all the sticks on all the boards and Samsung was making memory dies that could do that speed then. )

I do see you are already on the newest bios revisions

Just as a baseline like the old checking if the computer is plugged into the wall. The ram should go into A2 and B2, so 2nd away and 4th away from CPU, not the ones closest. 

 

If you have tried multiple kits and failed to get duel channel to run, there may be a problem with the traces/connector on the mobo OR the memory controller on the CPU, or simply reseating the CPU to get the pins to all contact right. but usually it should still post and just "not see" the other stick. Clean the connectors with isopropyl and a q tip, and hope and pray. 

[Jerakl]

14 hours ago, QuantumSingularity said:

I don't have to look at any benchmarks, i've upgraded more than 10 of those CPUs in the last 6 months. Yes, at 4k it;s still able to somewhat holdup, but at 1440p and 1080p often it gets smashed by a $130 5600X.

I upgraded from a 9900k to a 13600k and uh... yeah.

I don't wanna say the difference is "night and day", but there was a rather noticeable bump in certain loads.

14 hours ago, starsmine said:

I swear yall need to look at the benchmarks sometimes. stop telling people to spend money when it wont improve anything. a hypothetical infinitely fast CPU will only get him from 109 fps avg to 112 fps average at native 1664p with his chosen settings.

Despite what I said above I do agree. There's no reason to just blindly say "fuck it upgrade" unless there's a serious issue. 

However it would improve things, just at a heavy cost compared to tweaking settings

OP likely just has a tuning issue.

[Cpt_Poopie_Pants]

5 hours ago, starsmine said:

The other thing I noticed is you are not running it at full XMP. those sticks are capable of 3600MT/S and you have them running at 3200MT/s
The memory timings are the XMP timings (and are out of jdec spec for 3200MT/s so XMP is partially enabled)

 

So cool, we likely found our culpret. Lets try to fix the single channel issue first, and if we feeling spicy, fix the xmp clocks back to 3600.

now these sticks are not technically on the QVL for the Mobo, BUT, (that usually can just be ignored because new sticks come out faster then MOBO manufactures can test all the sticks on all the boards and Samsung was making memory dies that could do that speed then. )

I do see you are already on the newest bios revisions

Just as a baseline like the old checking if the computer is plugged into the wall. The ram should go into A2 and B2, so 2nd away and 4th away from CPU, not the ones closest. 

 

If you have tried multiple kits and failed to get duel channel to run, there may be a problem with the traces/connector on the mobo OR the memory controller on the CPU, or simply reseating the CPU to get the pins to all contact right. but usually it should still post and just "not see" the other stick. Clean the connectors with isopropyl and a q tip, and hope and pray. 

Ive made the changes and in game playing normal TDM im at 160 to 185 frames usually in the 170s. playing dmz it crashes to 100fps. 

[starsmine]

13 minutes ago, Cpt_Poopie_Pants said:

Ive made the changes and in game playing normal TDM im at 160 to 185 frames usually in the 170s. playing dmz it crashes to 100fps. 

by changes do you mean it posts in duel channel mode now? or what changes are we referring to?

[Cpt_Poopie_Pants]

https://valid.x86.fr/84taj9

 

New Bench

[Cpt_Poopie_Pants]

1 hour ago, starsmine said:

by changes do you mean it posts in duel channel mode now? or what changes are we referring to?

i made sure memory timings were where they should be. DRAM@3600

[duckargent]

The 9900k is a pretty powerful CPU even in 2023: don't compare it to AMD's older Ryzen CPUs for Intel's older CPUs, with their higher frequency, are still able to hold their own in the newest games. I have the same ASUS TUF OC 4070ti with a stock (4,9GHz) 10900k (which is a 9900k + 2 cores and a little higher frequency) and I get over 200 fps in MWII @ 1080p/360Hz (so a CPU bound scenario) - extreme details, no DLSS/upscaling.

 

Like other users mentioned, the OS version is too old. I recommend to do a full W10 reinstall (clean) with the latest drivers. Most probably it won't help but you will rule out the possibility of having a software-related flaw.

 

The second point is the RAM. If it's Corsair Vengeance 3200, I assume it's the CL16 version (10ns), which is weak: for comparison I am getting this fps with a 3200CL14 kit (8,75ns). Now, of course it is not worth replacing it with another DDR4 kit, this generation is at the end of the lifetime. As others have already said, just make sure to run the memory in dual channel mode and apply the XMP profile (to actually run it at 3200MHz).

 

RAM @3600CL18 won't make any difference: still 10ns true latency. Set the kit at its native 3200MHz and try to tighten the timings below 16.

 

Lastly, maybe there is simply something wrong with the OC variables/voltages: try resetting the BIOS to run the 9900k at stock (auto turbo) and see how it performs. If you insist to keep the OC: have you checked for stability or WHEA errors? WHEA errors are corrupting the files, but it's still worth asking since your CPU is not behaving normal.

[starsmine]

4 hours ago, Cpt_Poopie_Pants said:

i made sure memory timings were where they should be. DRAM@3600

oh no its still single channel. well at least you got it to go that little bit faster. 

Single channel halves bandwidth so if MW2 is a memory bandwidth sensitive game that is probably where the rest of the performance is hiding. Still though, im glad we were able to get the issue mostly resolved, and you are no longer CPU binding, getting all the performance you paid for out of that 4070ti, at 1440p native. 

[Cpt_Poopie_Pants]

6 hours ago, starsmine said:

oh no its still single channel. well at least you got it to go that little bit faster. 

Single channel halves bandwidth so if MW2 is a memory bandwidth sensitive game that is probably where the rest of the performance is hiding. Still though, im glad we were able to get the issue mostly resolved, and you are no longer CPU binding, getting all the performance you paid for out of that 4070ti, at 1440p native. 

I'm glad too. Thanks for the help!