Some games are seriously lagging on Xeon E5-2690v2

[David is Back]

It seems to happen on map segments preloading. Can it bee because of my cheap SATA SSD, PCI-E 3.0 or because AMD GPUs (mine is RX 6660 XT) are screwed up for such games? Should I change my platform, if CS2 which I play most of the time is running fine, and my PC is more than OK in working tasks?

[David is Back]

引用自 xSOSxHawkens：

Please quit wasting peoples time here arguing that your chip is the best

Of course it's not the best. I never said that. AMD Epic CPUs are better, newer Xeons are better. Intel 12-13th generations are better in general, 10900 and 10900K, 11900 and 11900K may be better in terms of computing power. And maybe some high end Ryzens CPUs of the last 2 years too.

[r.linder]

引用自 David Is Back：

引用自 尺．し工几ᗪヨ尺：

Completely different SKUs and cores don’t scale the same by power consumption. It’s not 1+1 mathematics.

You still don't get the point. Each new generation is cooler that the previous ones because of the renewed techprocess. So IvyBridge EP can't be hotter that SandyBridge EP, which showed 130W under overclock.

So not true, CPUs are getting hotter under refined processes, not cooler, because the more you refine the process and fit more transistors, the more heat can be produced in that area as a byproduct, it’s not just about wattage.

The latest flagships run hot on many cooling solutions at stock config.

引用自 David Is Back：

引用自 xSOSxHawkens：

83w idle and 426w load.

Please quit wasting peoples time here arguing that your chip is the best when you are blatantly ignoring publicly validated information for multiple posts, and arguing about it.

It can't be 426w load. What software was used in the measurment? Are you sure it's correct? Again, E5-1680 overclocked can't consume 4 times less power.

It is, actually read the article instead of arguing. It draws a lot more power than desktop chips because it’s a SERVER CPU. Those machines are designed to run heavier loads so they naturally draw a lot more power than the desktop segment of its generation.

Threadripper 7000 series draws even more power, you can draw an insane amount of power from the flagships but we’re talking well over 32 cores.

[_I_]

the intel refresh (2nd gen in each new socket)
is slightly faster and/or more efficient

the first gen with a new socket is generally more power and less efficient than previous socket cpus

[David is Back]

引用自 尺．し工几ᗪヨ尺：

It is, actually read the article instead of arguing. It draws a lot more power than desktop chips because it’s a SERVER CPU.

And so is E5-1680v2.

I already skimmed through the text. They were using some dubious software, not the phyusical multimeter - so this numbers are just "taken from the sky, random. I trust more to the numbers that I saw on YouTube (though it really was another model). I can test and write my own ones if you want to

[David is Back]

引用自 尺．し工几ᗪヨ尺：

It is, actually read the article instead of arguing. It draws a lot more power than desktop chips because it’s a SERVER CPU. Those machines are designed to run heavier loads so they naturally draw a lot more power than the desktop segment of its generation.

Threadripper 7000 series draws even more power, you can draw an insane amount of power from the flagships but we’re talking well over 32 cores.

So I ran the stress test, here you are.

https://drive.google.com/file/d/1USiI0XiRwTp9UyiBZRsH9ewgyHdfLTpO/view?usp=sharing

92-93 watts according to AIDA. Unfortunately I cannot run FPS Monitor in game now, since I don't have a valid license. But this number somehow is smaller than TDP from specs, not greater.

[David is Back]

I did read to the end - seems they were running Linpack. I've heard it is heavy (everyday load will never put so much stress on the CPU), though these numbers still seem strange because they are like 4.5 times higher. So I'd rather not take them just by belief

[r.linder]

It can vary by settings or by throttling, whether that be by thermals or electrical limits at the board, just because it’s supported in BIOS doesn’t mean the board can actually supply enough current to drive it, you can install a 5950X on an A320 board and it’ll run, but at a pretty significant cost in performance because those boards can barely even drive a 6 core CPU, let alone a 16 core, and there have been issues running Ryzen 9 on such low end boards because it’s unstable.

Intel describes their TDP has the maximum power consumption at sustained loads at their base frequency, doesn’t account for boost or overclocking. It also means all cores have to be running at their base clock of 3 GHz for the v2 to run at 130W. But if it’s running above that spec, the result can be much higher if the process isn’t that efficient, which is how there can be a large difference, the last 10% of the 4090’s performance is garnered by almost HALF of its TDP, you can run it at 270W TGP compared to its default 450W spec and lose around 8%, meaning 40% of its power usage out of the factory is allocated to such a small portion of what it actually offers, and it was originally going to be 600W, but the last 150W made such a negligible difference that they changed it, Linus Tech Tips tested it at that power level and found that they weren’t getting much and it was still limited in terms of power. Overclocking makes a huge difference to power consumption when you’re wringing silicon for everything it has, and there’s diminishing returns in performance the higher you go, with the requirements only increasing with every percent.

So if it’s not even reaching it’s TDP in a stress test then it’s not being pushed to its limit or it’s being held back by something. If it’s running below the base clock during the tests, that’s why it’s lower.

[David is Back]

引用自 尺．し工几ᗪヨ尺：

whether that be by thermals or electrical limits at the board

Sure, the motherboard should limit power consumption to save CPU and prevent damage and even fire, it is monitoring both temps and powers. And the guys who did the study used some customized test board from Intel designed specifically for tests, they are stating that in the text. It could have those limits turned off of course.

[David is Back]

引用自 尺．し工几ᗪヨ尺：

Intel describes their TDP has the maximum power consumption at sustained loads at their base frequency, doesn’t account for boost or overclocking. It also means all cores have to be running at their base clock of 3 GHz for the v2 to run at 130W.

Yeap, but AIDA64 stress test puts my CPU into Turbo Boost, which is 3.6 GHz (all cores load), and still it shows 92-93W. My power unit is 750W as I wrote somewhere before in this thread.

[David is Back]

引用自 尺．し工几ᗪヨ尺：

So if it’s not even reaching it’s TDP in a stress test then it’s not being pushed to its limit or it’s being held back by something. If it’s running below the base clock during the tests, that’s why it’s lower.

https://drive.google.com/file/d/1PNlTYfOwn4R5AyhD7MMKeDZYeRbsJcro/view?usp=sharing
OK, I was wrong, it's not 3.6 GHz, it seems to be equal to base clock with FSB oveclock of 13% applied to it. But still power is less than TDP.

[Χάρης]

I don't think using a xeon for gaming is a good idea ...

[r.linder]

引用自 David Is Back：

引用自 尺．し工几ᗪヨ尺：

whether that be by thermals or electrical limits at the board

Sure, the motherboard should limit power consumption to save CPU and prevent damage and even fire, it is monitoring both temps and powers. And the guys who did the study used some customized test board from Intel designed specifically for tests, they are stating that in the text. It could have those limits turned off of course.

Yeah, they tested it in a server system the CPU was actually designed for and not some run of the mill desktop "gaming" PC. The motherboard you have could easily be insufficient for actually powering that CPU.

引用自 David Is Back：

引用自 尺．し工几ᗪヨ尺：

So if it’s not even reaching it’s TDP in a stress test then it’s not being pushed to its limit or it’s being held back by something. If it’s running below the base clock during the tests, that’s why it’s lower.

https://drive.google.com/file/d/1PNlTYfOwn4R5AyhD7MMKeDZYeRbsJcro/view?usp=sharing
OK, I was wrong, it's not 3.6 GHz, it seems to be equal to base clock with FSB oveclock of 13% applied to it. But still power is less than TDP.

That... isn't how turbo boost is applied but OK.



None of this changes the fact that your performance will only improve massively by upgrading, it's a 10+ year old Xeon with single-core performance similar to an i7-2600K but at a lower frequency. Boxing day sales will likely bring some pricing down a lot, there's also second hand i9s on Z490/590 and X299 as an option if you prefer Intel.
Or just accept the fact that the Xeon will only give you performance issues because of its age, it doesn't matter how many cores it has, any CPU will show its age after a decade.

[David is Back]

引用自 尺．し工几ᗪヨ尺：

Yeah, they tested it in a server system the CPU was actually designed for

It's not a server system, read carefully. "Intel dev board". Though the second motherboard was indeed a server one from Supermicro.

引用自 尺．し工几ᗪヨ尺：

it doesn't matter how many cores it has

In fact, the frequency after BCLK overclock is pretty close to 3.4 GHz of i7 2700K, and my number of cores is literally 2.5 times larger. Again, my previous CPU was i5 2500K, which has 4 cores and does not even have HT, so only 4 threads. It's a massive improvement, and it feels so.

Dead By Daylight, Fortnite, DayZ and many other games are running perfectly on this PC.

[David is Back]

引用自 尺．し工几ᗪヨ尺：

Boxing day sales will likely bring some pricing down a lot, there's also second hand i9s on Z490/590 and X299 as an option if you prefer Intel.

Of course I can do another upgrade, but why, upgrade for an upgrade? The PC is working fine. I just have several games that are lagging and trying to find out why.

[r.linder]

引用自 David Is Back：

引用自 尺．し工几ᗪヨ尺：

Yeah, they tested it in a server system the CPU was actually designed for

It's notr a server system, read carefully. "Intel dev board". Though the second motherboard was indeed a server one from Supermicro.

引用自 尺．し工几ᗪヨ尺：

it doesn't matter how many cores it has

In fact, the frequency after BCLK overclock is pretty close to 3.4 GHz of i7 2700K, and my number of cores is literally 2.5 times larger. Again, my previous CPU was i5 2500K, which has 4 cores and does not even have HT, so only 4 threads. It's a massive improvement, and it feels so.

Dead By Daylight, Fortnite, DayZ and many other games are running perfectly on this PC.

They're still server systems, I don't care about the semantics around what the motherboard is referred as and what they're used for. It's still using a lot of power on Supermicro servers.

I'll explain again since you're not listening. Core count never made a difference in gaming performance prior to around 2018, 4 cores was more than enough, extra cores were there for multi-core performance which is less important of a spec for gaming than single core performance. Games weren't developed for higher core counts until after Ryzen released, so the Xeon will hold up better than regular Sandy Bridge CPUs these days but it's still slow by today's standards.

i5s were notably slower than i7s due to the lack of HyperThreading but they were still good enough back then, nowadays they obviously struggle due to having 4 threads so anything feels like an improvement, FX used to struggle against i3s and i5s but nowadays they're able to keep up and win because even though the cores share FPUs, it's more than enough to outperform 4 thread CPUs now.

But regardless, no matter how many cores a CPU has, it doesn't mean it's going to keep up in performance forever, your 10 core CPU from 10 years ago is nowhere near as fast as my 10 core CPU from 2020. (i9-10850K)
What actually matters is how those cores perform, that Xeon was an incredible CPU 10 years ago, but nowadays, it's bested by several i5s and R5s (most of which are 6 core) because their core performance is higher. My 10850K can easily be beaten by newer CPUs with fewer cores as well.

https://www.cpubenchmark.net/compare/2057vs3824/Intel-Xeon-E5-2690-v2-vs-Intel-i9-10850K
https://www.cpubenchmark.net/compare/2057vs3735/Intel-Xeon-E5-2690-v2-vs-Intel-i5-10600K
https://www.cpubenchmark.net/compare/2057vs4233/Intel-Xeon-E5-2690-v2-vs-Intel-i5-11400
https://www.cpubenchmark.net/compare/2057vs4677/Intel-Xeon-E5-2690-v2-vs-Intel-i5-12400
https://www.cpubenchmark.net/compare/2057vs4994/Intel-Xeon-E5-2690-v2-vs-Intel-i5-13400
https://www.cpubenchmark.net/compare/2057vs3481/Intel-Xeon-E5-2690-v2-vs-AMD-Ryzen-5-3600
https://www.cpubenchmark.net/compare/2057vs4811/Intel-Xeon-E5-2690-v2-vs-AMD-Ryzen-5-5600
https://www.cpubenchmark.net/compare/2057vs5172/Intel-Xeon-E5-2690-v2-vs-AMD-Ryzen-5-7600

引用自 David Is Back：

引用自 尺．し工几ᗪヨ尺：

Boxing day sales will likely bring some pricing down a lot, there's also second hand i9s on Z490/590 and X299 as an option if you prefer Intel.

Of course I can do another upgrade, but why, upgrade for an upgrade? The PC is working fine. I just have several games that are lagging and trying to find out why.

You've already been told the answer by several people but you seemingly refuse to accept it. You just argue with people and try to defend your processor when people say it's slow.

It's lagging because it's old and slow, it's a freaking decade old. If the performance is not satisfactory to the point that you're unhappy, then it's time to upgrade. If you've convinced yourself that it's fine, then keep using it, but don't expect performance issues in some games to just go away, because that won't happen.