PC Upgrade (Motherboard & CPU)

[Morbad]

Gamers Nexus has just released a comparative review of the CPU retainers:

Source: https://www.youtube.com/watch?v=iYU1OskbY-Q

Just saw that myself.

My earlier recommendation was going almost completely off the price differential, with the knowledge that Thermal Grizzly stuff, while usually nice, is needlessly expensive. This bears out in that review. I was aware of Thermal Grizzly's claims with regard to tight manufacturing tolerances, and the lack of any specs from Thermalright in this area, but Thermalright isn't so sloppy that it would be meaningful (that ~240 micron variance GN measured on Thermalright's frames is already smaller than the tolerances on the socket itself).

With regards to TG's claim regarding potentially superior memory overclocking...there are probably less than one-hundred people on the planet that could reliably quantify a difference as trivial as small variances on LGA pin clamping force (I've been OCing memory for almost 30 years and I'm not one of them). So, if you aren't running carefully hand binned parts in a near laboratory-precise extreme OCing scenario, one is going to be hard pressed to see any difference that couldn't be accounted for simply by adjusting screw torque on either frame. It's such a niche thing that it's a step that would come after trimming off any excess lead length from the DIMM slot through-holes (which can start to act as antennae at the I/O frequencies modern DRAM can reach) and shielding the back of the board with kapton tape and foil to reduce EMI.

One difference that did stand out was the mention that the TG frame doesn't actually contact the board, while the Thermalright frame does. In this regard, Thermalright is actually going to be superior for most people as you get meaningful manual feedback without having to precisely measure torque, with no risk of crushing the socket. The installation method shown is a good one, but honestly, you can be less precise...as long as you get pressure mostly even across the screws on the way down and don't over tighten things, you'll be fine.

You can also see the issue with the stock Intel retention mechanism in the pressure tests. The two tabs pushing the CPU into the socket when it's locked down actually bend the whole middle of the IHS (the hottest part of the CPU) away from the cooler, increasing the gap the TIM has to fill and reducing effective mounting pressure, which is why stock ILM thermal performance sucks, relatively speaking.

 

[Riverside]

Just saw that myself.

My earlier recommendation was going almost completely off the price differential, with the knowledge that Thermal Grizzly stuff, while usually nice, is needlessly expensive. This bears out in that review. I was aware of Thermal Grizzly's claims with regard to tight manufacturing tolerances, and the lack of any specs from Thermalright in this area, but Thermalright isn't so sloppy that it would be meaningful (that ~240 micron variance GN measured on Thermalright's frames is already smaller than the tolerances on the socket itself).

With regards to TG's claim regarding potentially superior memory overclocking...there are probably less than one-hundred people on the planet that could reliably quantify a difference as trivial as small variances on LGA pin clamping force (I've been OCing memory for almost 30 years and I'm not one of them). So, if you aren't running carefully hand binned parts in a near laboratory-precise extreme OCing scenario, one is going to be hard pressed to see any difference that couldn't be accounted for simply by adjusting screw torque on either frame. It's such a niche thing that it's a step that would come after trimming off any excess lead length from the DIMM slot through-holes (which can start to act as antennae at the I/O frequencies modern DRAM can reach) and shielding the back of the board with kapton tape and foil to reduce EMI.

One difference that did stand out was the mention that the TG frame doesn't actually contact the board, while the Thermalright frame does. In this regard, Thermalright is actually going to be superior for most people as you get meaningful manual feedback without having to precisely measure torque, with no risk of crushing the socket. The installation method shown is a good one, but honestly, you can be less precise...as long as you get pressure mostly even across the screws on the way down and don't over tighten things, you'll be fine.

You can also see the issue with the stock Intel retention mechanism in the pressure tests. The two tabs pushing the CPU into the socket when it's locked down actually bend the whole middle of the IHS (the hottest part of the CPU) away from the cooler, increasing the gap the TIM has to fill and reducing effective mounting pressure, which is why stock ILM thermal performance sucks, relatively speaking.

Yeah I agree the TR clamp is definitely easy to install and my initial reluctance to get an aftermarket clamp was based on what I'd seen of the TG one & it's requirement for quite precise & evenly applied torque to each corner bolt. I think the TG one comes with a lapping tool & undoubtedly better quality thermal paste. As you say the TG one seems to be aimed at the very top end niche overclockers where the TR one seems much more mainstream & well within my capabilities to install.

Here's the thing though: With all of these clamping mechanisms, all they need to do is maintain contact with all 1700 tiny sprung levers (or however many for the various different intel ball grid sockets). Enough force that all of them make good electrical contact even with no heatsink at all, but not so much force that the fibreglass motherboard or the plastic socket are distorted.
That's not actually all that much force.

The stock Intel design has a spring that seems way stronger than it needs to be just to basically hold the CPU in place while the HSF/watercooling block is put on. It needs to push the CPU down against the sprung lever contacts but that's it really, and that's all the aftermarket ones do (similar with ZIF sockets). The actual force they apply to the CPU's heatspreader seems much less than the stock pressed steel one.

And for all the precision required to all sides of the heatspreader, the HSF springs apply way, way more force to actually clamp the CPU down into it's socket & make those 1700 little contacts. As far as I could see the main job that TR clap did (apart from not distorting the heatspreader as the stock one does) is to simply hold the CPU in the socket while you tighten down the HSF


As far as actual temps are concerned my enormous aircooler (Noctua D15) maintains about 30c at idle which is colder than my old one, and while playing the game it remains near silent at around 50c, about the same as my old one. However if I run a CPU benchmark (eg stress testing with CPU-Z) the temp spikes up to 75-80c before the fans spool up enough to dump the heat out of it. In that respect I can see maybe a waterblock or AIO might handle that better. In practice (is not benchmarking) it's absolutely fine on air though.

 

[Morbad]

As far as I could see the main job that TR clap did (apart from not distorting the heatspreader as the stock one does) is to simply hold the CPU in the socket while you tighten down the HSF

It imparts a fair bit of extra rigidity to the socket area, which also improves contact. With just the stock ILM, only the backplate is resisting the mounting pressure of the heatsink and most heatsinks will apply enough pressure to cause some degree of warping without extra support.

 

[Riverside]

It imparts a fair bit of extra rigidity to the socket area, which also improves contact. With just the stock ILM, only the backplate is resisting the mounting pressure of the heatsink and most heatsinks will apply enough pressure to cause some degree of warping without extra support.

I remember the stress that was putting a HSF on an Athlon's exposed die The heatspreaders role is to spread that pressure and if the HSF (or coldplate) is touching anything more than just that heatspreader it was installed wrong.

Structurally the part that reduces warping of the board & socket from HSF pressure is the HSF's steel backplate. The CPU clamp is only there to apply slightly more pressure than is required to stop the CPU from falling out

The socket design is a compromise I suppose.
A ZIF PGA socket gives a stable platform that HSF pressure can be applied to, independent of CPU pin connection pressure (which is applied sideways to the CPU pins). But CPU pins are delicate & easily bent.
An LGA socket allows for a more robust CPU package (potentially the more expensive component in many PCs and certainly the one more likely to be handled roughly by a user) but the pressure applied has to be more finely managed by the installer.

I suppose as far as intel are concerned the majority of their CPUs will sit under a stock cooler that's just clipped onto the board, and their solution is that bit less user friendly for the home builder

edit: fixed socket terminology after some googling

ETA done some googling on this, the LGA socket style (with the pins in the socket) allows for higher pin density, which seems like something that will become increasingly important as CPU technology progresses. Hopefully next time I need an upgrade a better solution will have been developed

 

[shadragon]

...with the knowledge that Thermal Grizzly stuff, while usually nice, is needlessly expensive.

The price and performance differences between an Aston Martin Vantage and a Hyundai Santro are quite large. Both will get your kids to school. Both will get you to the grocery store. Both can take you on a weekend trip. However, the Vantage, being ten+ times the cost, offers significantly better performance and ride. You pay for the extra build quality, top grade components and a heck of a lot of research and development. You are not obliged to pay for a Vantage in any way. You are free to buy whatever vehicle fits your budget and needs, but your experience in a Santro on an eight hour road trip will be far inferior to one in a Vantage.

Thermal Grizzly works a treat. It is expensive when you compare it to others, but when you see the benefits of using quality paste and can run your CPU 2-6 Celcius cooler than someone using Ed's Bargain Basement brand from the Pound Shop, it's worth the cost. That's extra overhead my system can use for years. TG names a price and people pay it or not. That's the open market. Some recognize that quality costs a little more, but pays out dividends over time. Others buy the cheap stuff and moan about their "crap PC" overheating when idle.

I've used TG Kryonaut exclusively in my last three PC builds and plan to use it in my new 5975wx Threadripper PRO build. I've never had a second of regret. If you put together a £2,000+ PC, don't nickel and dime yourself with cheap paste.

Some things are worth spending just a little bit more on.

 

[Morbad]

Structurally the part that reduces warping of the board & socket from HSF pressure is the HSF's steel backplate. The CPU clamp is only there to apply slightly more pressure than is required to stop the CPU from falling out

With the frame in contact with the board, it's absolutely going to contribute to the board's ability to resist flexing. The IHS also barely protrudes from the top, so once you reach pressure enough to cause the lid to sit flush, it's very difficult (and mostly futile) to apply more.

Most backplates, including the Intel OEM ones, aren't anywhere near as rigid as I'd like. Intel's are much thicker and stiffer than the ones included with most AMD boards though.

A ZIF PGA socket gives a stable platform that HSF pressure can be applied to, independent of CPU pin connection pressure (which is applied sideways to the CPU pins).

It is perfectly possible to bend an ZIF PGA socket (they are plastic, and not all that thick), like AM4, sufficiently to cause problems, just by mounting some coolers. The boxed cooler that came with my 3900X will warp a cheap 4-layer board with the default backplate by about 3mm when locked down, which is enough to make the corners of the socket visibly higher than the center. I've had this cause issues with CPUs, or memory channels, being detected. I have other coolers, like my Thermalright IB-E Extreme, that have almost twice the mounting pressure of the boxed cooler's clamp. Fortunately, Thermalright provides a much thicker and very stiff backplate, so even with ~200 pounds of mounting pressure, the boards barely flex.

ETA done some googling on this, the LGA socket style (with the pins in the socket) allows for higher pin density, which seems like something that will become increasingly important as CPU technology progresses. Hopefully next time I need an upgrade a better solution will have been developed

Yes, the main reason for switching from PGA or micro PGA to LGA was electrical. There are fans of both packaging types, but I doubt consumer convenience is a major priority for AMD or Intel either way...most CPU sales aren't to end-users.

The price and performance differences between an Aston Martin Vantage and a Hyundai Santro are quite large. Both will get your kids to school. Both will get you to the grocery store. Both can take you on a weekend trip. However, the Vantage, being ten+ times the cost, offers significantly better performance and ride. You pay for the extra build quality, top grade components and a heck of a lot of research and development. You are not obliged to pay for a Vantage in any way. You are free to buy whatever vehicle fits your budget and needs, but your experience in a Santro on an eight hour road trip will be far inferior to one in a Vantage.

Thermal Grizzly works a treat. It is expensive when you compare it to others, but when you see the benefits of using quality paste and can run your CPU 2-6 Celcius cooler than someone using Ed's Bargain Basement brand from the Pound Shop, it's worth the cost. That's extra overhead my system can use for years. TG names a price and people pay it or not. That's the open market. Some recognize that quality costs a little more, but pays out dividends over time. Others buy the cheap stuff and moan about their "crap PC" overheating when idle.

I've used TG Kryonaut exclusively in my last three PC builds and plan to use it in my new 5975wx Threadripper PRO build. I've never had a second of regret. If you put together a £2,000+ PC, don't nickel and dime yourself with cheap paste.

We're not talking about the difference between some generic zinc oxide paste formulated forty years ago as a sunscreen and TG Kyronaut, but the difference between a dozen or so modern pastes with extremely similar properties.

Among modern high-end pastes, I've used TG Kyronaut, Thermalright TFX, Noctua NT-H1/2, Arctic MX-4, Prolimatek PK-3, Cooler Master Mastergel Maker, SYY 157, and a few others. You'd be hard pressed to tell any of them apart in performance, especially the short term.

Raw performance is not the only factor either. NT-H2, for example, performs very well, but tends to flake off into tiny solid chunks when wiped away without using a solvent, which makes it annoying to remove. Likewise, TG Kyronaut and all the Prolimatek pastes use metallic aluminum fillers, and if you strip away the binder (while trying to clean it with solvents), you have ultra fine aluminum dust coating everything it was touching. I once killed one of my favorite X58 boards when I couldn't completely remove some stray PK-1 from the socket area. Also, some TIMs aren't viscous enough to last in the long term; the first incarnation of Thermalright Chill Factor performed extremely well on lapped surfaces, but lasted barely six months of heavy use due to pump out. Conversely, there are a few pastes that are so viscous that heatsink mounting pressure alone cannot achieve optimum bond line, or that require extensive thermal cycling to reach final performance. Some Shin-Etsu pastes were examples of this; they had the viscosity of wax and needed the container to be set in near boiling water or the component it was used on to be run without a fan, to get the TIM thin enough for it to perform right.

I want a TIM that perfoms well, lasts forever, cleans up easily, doesn't mandate special procedures to properly apply, and costs as little as practical.

Kyronaut's viscosity is a little on the low side, and, while it's rarely an issue, I don't like metallic fillers in conventional pastes. They also try to charge 30 dollars a cubic centimeter for it, unless you buy it by the tub.

My current go to paste is SYY-157, which is easily within 1C of Kyronaut, and thus far seems like it will last forever...it's also significantly less expensive than most of it's competitors. No metallic components to SYY 157 either.

For liquid metal TIMs, I've tried all the ones on Tom's chart and settled on Phobya, because it's the cheapest and doesn't really perform any worse than anything else. However, I only use liquid metal between bare silicon dies and nickel plated cold plates, which in practice, means it's currently on one of my video cards and nothing else. Gallium based TIMs alloy with bare copper, making it only last 6-18 months (if you want best performance), unless you carefully treat the surface before hand. Since I lap everything that's convenient to lap, it means everything that has a heat spreader/lid and all my CPU coolers/waterblocks wind up bare copper, even if they didn't come that way. Only GPU coolers keep their nickel plated bases because it's too much work to lap most of them while actually improving flatness.

Anyway, I don't have any regrets about the Kyronaut I've used either, but there are TIMs I consider to be superior, and among those there are at least a half-dozen that are close enough to be fungible.

Some things are worth spending just a little bit more on.

And sometimes you can get something just as good, or better, for a lot less.

Few of TG's products are outright bad (though their Minus Pad 8 came close), but almost all of them have an alternative that is within margin of error for significantly less. About the only thing I'd recommend TG for is their de-lid tools...though those might not even be TG branded, I think they are just Der8auer.

For the record, this is the most recent Tom's Hardware TIM performance chart for high mounting pressure with a cooler with a more or less flat base:

For the whopping 0.08C difference tested (which is way below margin of error) between Kyronaut and SYY 157, I am always going to pick the higher viscosity, completely non-metallic, TIM.

Edit: The newer Kyronaut Extreme (which I haven't tried myself) apparently has no metallic aluminum, only oxides, but I'm pretty sure the original Kyronaut is using metallic aluminum particles.

 

[Morbad]

Igor's Lab recently tested some of the other LGA-1700 contact frames on the market:

Results for the Thermal Grizzly and Thermalright frames essentially mirror GN's tests. Thermally speaking, the TG frame is a bit more finicky, but once dialed in to the correct tension is pretty much the same as the easier to use and cheaper TR frame. However, Igor also didn't do a peak stable memory clock comparison, other than noting the number of retrains required for his DDR4-7000 OC.

The real surprise comes from the other frames tested. The Feng Zao fiberglass frame, which has very poor tolerances, both trained immediately resulted in the best temps. Igor suspects this is due to the flexibility of the material, which does make sense...it wouldn't help increase mounting pressure by resisting board flex, but it would equalize the pressure of the ILM on the CPU.

The big loser is the Jeyi frame. It's both rigid and has poor tolerances and Igor couldn't make his test system stable at the baseline OC while using it.

 

[Riverside]

Igor's Lab recently tested some of the other LGA-1700 contact frames on the market:

Results for the Thermal Grizzly and Thermalright frames essentially mirror GN's tests. Thermally speaking, the TG frame is a bit more finicky, but once dialed in to the correct tension is pretty much the same as the easier to use and cheaper TR frame. However, Igor also didn't do a peak stable memory clock comparison, other than noting the number of retrains required for his DDR4-7000 OC.

The real surprise comes from the other frames tested. The Feng Zao fiberglass frame, which has very poor tolerances, both trained immediately resulted in the best temps. Igor suspects this is due to the flexibility of the material, which does make sense...it wouldn't help increase mounting pressure by resisting board flex, but it would equalize the pressure of the ILM on the CPU.

The big loser is the Jeyi frame. It's both rigid and has poor tolerances and Igor couldn't make his test system stable at the baseline OC while using it.

The most interesting one there is the last one no frame, no ILM)

Fibreglass is difficult to get fine tolerances with, FRP is good though & quite a popular material for moulded parts that need to be stronger than plastic alone.

Clearly there is some benefit to taking the time to get the pressure even & precise, which makes sense.

I removed & reseated by HSF a few days ago, in part to see what the thermal paste spread looked like. It was fine, I had put the recommended (by Noctua) large blob in the middle & four small blobs towards the corners of the IHS, there was a slight X pattern visible in the paste but full coverage. Re-doing it the temps haven't changed which reassured me I'd done it right the first time.

But one interesting thing to note was that all four of the retaining plate screws were slightly loose, suggesting the foam rubber pads under the plate had compressed & permanently deformed. I nipped them up again & checked the four bolts that go through the mobo for the HSF mount & they were similarly slightly loose. Glad I reseated it now

The frames are pretty cheap components, certainly compared to the CPU & mobo. Glad you recommended I get on Morbad, thanks

 

[Morbad]

The most interesting one there is the last one no frame, no ILM)

This was my default before starting to lap CPUs with the aid of a salvaged socket, or when I was running delided parts without replacing the IHS. Before delidding, or the custom tools for it, were common, the safest way to remove the IHS from a soldered part was just to sand through the entire IHS, destroying it, but leaving a convenient support shim around the perimeter. Other methods involved cutting the adhesive holding it to the substrate and either heating the IHS to the reflow temp of the solder (which was considerably higher than the initial solder temp due to the kind of solder used), or using a sharp impact to mechanically separate it from the die (which worked well enough with more conventional TIM, but often fractures or tears off the whole die when it's soldered to the IHS). No IHS is also generally cooler than keeping it, provided the cooler base has sufficiently good contact. Not using any sort of solid retention mechanism does require one to be careful not to shift the CPU when installing the cooler though.

But one interesting thing to note was that all four of the retaining plate screws were slightly loose, suggesting the foam rubber pads under the plate had compressed & permanently deformed.

This is interesting to note, though not surprising. I do wonder if the screws loosened a bit since install as well.

I nipped them up again & checked the four bolts that go through the mobo for the HSF mount & they were similarly slightly loose. Glad I reseated it now

Yeah, they warn about over tightening these, for obvious reasons, but I've always had best results going beyond finger-tight, especially if compressible spacers are used. Noctua, despite being conspicuously higher-end in most areas, still uses plastic standoffs for their cooler mounting systems when a lot of other companies use steel spacers with a fairly stiff nylon washer to protect the board. Screws do go all the way through them, but the plastic still deforms slightly and screws can loosen over time, especially with a lot of thermal cycling (some sort of weak threadlock is often wise on the mounting hardware).

 

[Jumpin' Jack Flash]

Hi All,

I hope nobody minds too much me resurrecting this topic, but after reading through all threads, and going off on a tangent to explore items of interest mentioned in the content, (other web sites etc.), it's took me awhile to post thanks to all that contributed to this topic!
I must say this topic has been quite enlightening, and very interesting, because I'm in the exact position (more or less) that Riverside was in at the start of the topic.
That being, rebuilding one of my computers to a more 'modern' up to date specification.

I'm still not quite certain which cpu & motherboard I'm going to go with, but it's starting to ere on the Intel side. (at the moment).
The one I'm upgrading from is an 'old' AMD Asus Rog Crossfire Formula 3, all I can pretty much salvage from it is the case, some hard drives, case fans and a Logitec G19 key board.
I was hoping to use it's current PSU...A Corsair HX 850 (Silver Edition). That was until I read somewhere (again) that it wasn't a good idea to build a new computer, MB . CPU. and Ram (plus graphics Card) using a PSU that was out of warranty, even though the PSU had been, and is still to all intents and purposes working ok.
Well, the PSU must be well over 10 years old, so that is going to be an extra expense I hadn't bargained for.

My other computer that I currently use is Intel based, runs ED okay, albeit in lower settings, It's got a Corsair AX 850 Gold Psu, but that's halfway through the warranty anyway so I'll leave it where it is.

The aftermarket plate in place of the intel cpu lock down mechanism was a great read, and the video. My previous career before I retired was in light and heavy engineering, I'd had thoughts that the standard mechanism was a bit suspect, but thought well, they must know what they're doing. I'd noticed the sloppy ness of the top plate and always tried to get it dead central, before locking it down. Yep that's going to be another £15 !

Anyway, I Digress.
A thoroughly good read!

Jack

 

[Riverside]

My last few upgrades have alternated between CPU/mobo and GPU/PSU. Getting it all at once has been a bit too much cash to throw at it at once.

Dunno about not using a PSU out of warranty but I'd use a good quality, over-specced PSU rather than a cheap or marginal one I suppose you have to ask yourself how much is piece of mind worth to you?

There is a PSU calculator built into the EVGA GPU management app I use (PrecisionX1), it says I need a 650w PSU with my now current build, and the same build but with a 3080ti in it would need a 750w PSU. I have a 750w PSU now, I'll probably replace it (with maybe an 850w or more) when I eventually get a new GPU rather than potentially run this one to capacity.


Do you have a budget in mind? I spent about £660 on my upgrade but got about £200 back selling the old stuff.

 

[Jumpin' Jack Flash]

Hi Riverside

I'm just out mining (Hawkins Gap) ....and looking in on the forum at the same time.

Do you have a budget in mind? I spent about £660 on my upgrade but got about £200 back selling the old stuff.

Huh! ...did have, but that's gone up in the last week by over £300 quid!
New PSU. a Z690 mb....(Probably Asus), errm, haven't factored in a GPU at the moment, I've (like you) got a retail home edition of Win 10 so should be able to switch to 11 (if I don't leave it too long).

Walked into the kitchen this morning, in the semi darkness, in my socks, suddenly got my feet wet....fridge on the blink due to the hot weather probably....£!50- £300, depending on what my wife picks.....

Budget?....what's one of those?
&*%^$...run out of limpets.

Edit...Okay, jokes aside....well, not about the fridge, my 'budget' was somewhere in the region of £750 at a rough guess. Asus Rog Strix Z690 A Gaming wifi, DDR4. Intel i5 12600 S Alder Lake. 16-32 GB Ram. (not sure what mhz probably 3000?). I'll more than likely need to get a larger SSD drive (Only 120GB on this one).
Plus the PSU, I have been looking at a Seasonic 850 gold full modular one, about £130 ish. So that's going to take me over budget . I've still got to decide on an air cooler for the cpu, have one of three in mind, but at the moment most of the reasonably spec. ones haven't got the LGA 1700 mounting kits supplied (well, in the retail box) except Noctua and a few others. They say on the box or hardware site they have, but when you dig deeper into customers feedback they haven't. Mainly due to old stock still on the shelves etc.
I'm in no immediate rush, I don't think reasonably priced GPU's will be around yet, but I would like to start and finish this build by Christmas!




Jack

 

[Riverside]

Hi Riverside

I'm just out mining (Hawkins Gap) ....and looking in on the forum at the same time.



Huh! ...did have, but that's gone up in the last week by over £300 quid!
New PSU. a Z690 mb....(Probably Asus), errm, haven't factored in a GPU at the moment, I've (like you) got a retail home edition of Win 10 so should be able to switch to 11 (if I don't leave it too long).

Walked into the kitchen this morning, in the semi darkness, in my socks, suddenly got my feet wet....fridge on the blink due to the hot weather probably....£!50- £300, depending on what my wife picks.....

Budget?....what's one of those?
&*%^$...run out of limpets.

Jack

Shame about your fridge, these things can happen any time Our dishwasher died last week, but after I'd already bought my new PC bits (got a new one no problem).

Perhaps an excuse to wait until the end of the year for the new AMD socket?

 

[Jumpin' Jack Flash]

Hi

As you mentioned earlier in this topic, you're more familiar with Intel than AMD. I'm in the same boat, so for the rest of this week I'm going to get stuck in learning what the present AMD Cpu's and Motherboards are like compared to the Intel ones I vaguely have in mind. All this 'investigation' is going to make my brain hurt and my eyes bleed! .

Jack

 

[Ashnak]

Define "present" - and your comparison point. In general, at a comparable price, AMD delivers more cores while Intel pushes harder on the clock speed. Also, very generally speaking, AMD tends to deliver more ops per Watt, i.e. gets a better power efficiency.

I just read this morning that AMD wants to present their next generation AM5 CPUs on the 29 /30 of August (current generation 5 is the last for the AM4 socket).

 

[Riverside]

in my opinion and i may be wrong posting this in a gaming forum ist that the powerconsumption of a gaming pc is not discussed at a level it deserves. at least here in central europe its highly likely that the cost and availability of energy is getting more critical. my current desktop ist an 3700X, 32GB-3200Mhz, 3060TI and only SSD´s running flawless on an 500W psu. i measure the curremt consumption with an precision wattage meter and get 150W average playing Odyssey including an 32" gaming monitor. i can afford a more powerfull rig, but whatfor, i´am happy with 60fps. i´am observing the release of ZEN4 amd the new graphics solutions, but energy consumtion is important to me and must be less than 200W average.

For a given workload newer hardware is probably going to be more efficient but in general if you are concerned about energy consumption of PC gaming your best bet is probably to not do it at all, do it less or turn down the graphics options. These are all options any individual can take proactively depending on their personal circumstances.

 

[Morbad]

My previous career before I retired was in light and heavy engineering, I'd had thoughts that the standard mechanism was a bit suspect, but thought well, they must know what they're doing.

Oh, they know exactly what they're doing. The issue is that the engineers work for the accountants, who work for the shareholders. The best solution for the corporate bottom line may be pretty far off the best solution from a pure engineering perspective.

in my opinion and i may be wrong posting this in a gaming forum ist that the powerconsumption of a gaming pc is not discussed at a level it deserves. at least here in central europe its highly likely that the cost and availability of energy is getting more critical. my current desktop ist an 3700X, 32GB-3200Mhz, 3060TI and only SSD´s running flawless on an 500W psu. i measure the curremt consumption with an precision wattage meter and get 150W average playing Odyssey including an 32" gaming monitor. i can afford a more powerfull rig, but whatfor, i´am happy with 60fps. i´am observing the release of ZEN4 amd the new graphics solutions, but energy consumtion is important to me and must be less than 200W average.

As I've pointed out elsewhere, power consumption on the most power hungry of components is largely arbitrary. Even at stock speeds, we are way past the point of diminishing returns in efficiency, because it's performance that sells.

By and large, the more advanced the hardware, the higher it's fundamental efficiency. You just have to find the optimal settings yourself, because stock settings are going to prioritize performance.

For a given workload newer hardware is probably going to be more efficient but in general if you are concerned about energy consumption of PC gaming your best bet is probably to not do it at all, do it less or turn down the graphics options. These are all options any individual can take proactively depending on their personal circumstances.

The relative efficiency of newer hardware is pretty easy to demonstrate.

If you have any GPU limited benchmark results for your 1080 Ti, and the peak power consumption the part reaches in them, I could see how low some of my newer parts could get on power while targeting a similar level of performance.

 

[Riverside]

i have a pretty good feeling meanwhile what power my computer consumes under certain circumstances. shure i can stop playing and save the energy, but thats not my point. my point is, when is the increased powerconsumption not worth the gaming improvments. or in other words get a good balance. i think the english speaking world has a term for this, "best bang for the buck". my daily surfing computing work i do an an 15" laptop with AMD 4500 CPU no external GPU and its takes 8W average, enough for surfing or doing office stuff.

You can buy pass-through power consumption monitors if you want to find out how much power something is consuming while plugged into mains power. PC gaming (particularly something graphics heavy like ED) is not really a comparable activity to 'surfing or [most] office stuff'. If you tried to run ED on your laptop I imagine it would consume quite a lot more power.

Reducing power consumption is a good thing, but how important it is to the individual is down to their personal circumstances & as I said, well within the individuals control. If you are considering upgrading your PC to be able to play games better, power consumption will inevitably be a lower priority factor in the decision making process than functionality and cost. If power consumption were the #1 factor, not playing would make more of a difference than upgrading to newer (and potentially more efficient) hardware.

 

[Riverside]

If you have any GPU limited benchmark results for your 1080 Ti, and the peak power consumption the part reaches in them, I could see how low some of my newer parts could get on power while targeting a similar level of performance.

GPU-Z shows both board power draw and GPU chip Power Draw. Running Superposition v1.1 at the 4k optimised setting my 1080ti peaks at 291w board power draw (fluctuates between ~265 & ~291), 204w GPU chip power with nothing overclocked. The CPU activity is barely above idle throughout.

 

[Morbad]

GPU-Z shows both board power draw and GPU chip Power Draw. Running Superposition v1.1 at the 4k optimised setting my 1080ti peaks at 291w board power draw (fluctuates between ~265 & ~291), 204w GPU chip power with nothing overclocked. The CPU activity is barely above idle throughout.

Yeah, Superposition is almost purely GPU limited and CPU has a negligible impact on score.

Anyway, you didn't tell me what score you were getting, but if I recall what my own 1080 Ti's got, I would guess something in the ballpark of 9k.

I can crack 9k on my RX 6800 XT at ~130w total board power, and with my 6900 XT (because it's the same thing, but with more shaders, so doesn't need to be clocked as high) at about 120w. My RTX 3080 isn't far off, but has a higher power floor, thanks to relatively hungry GDDR6X.

My next GPU will probably be an RTX 4090, and if I had to hazard a guess based on current leaks, rumors, and speculation, there probably wouldn't be any point in targeting such a low score, but it could be done with sub-100w.