GPUs & CPUs & Enthusiast hardware: Questions, Discussion and fanboy slap-fights - Nvidia & AMD & Intel - Separe but Equal. Intel rides in the back of the bus.

  • 🐕 I am attempting to get the site runnning as fast as possible. If you are experiencing slow page load times, please report it.
It's appearing that to get full benefit from a 3080 I will have to get a new mobo and cpu as I have a 7700k.

Ryzen here I come!
AMD is having a CPU event on the 8th of October, current gen might be a bit cheaper on the used market in a month or so if you can wait for the new ones to release.
 
  • Agree
Reactions: Allakazam223
AMD is having a CPU event on the 8th of October, current gen might be a bit cheaper on the used market in a month or so if you can wait for the new ones to release.
Yeah I'm just in the planning/saving stage since I've got some other expenses to take care of first. I have plenty of time for new generations and for prices to drop so it's all smooth sailing for me.
 
Ah, I miss those JohnnyGuru "gutless wonder" reviews.

But yeah, any decent power supply nowadays is just fine. Shitty components or overrated gpu clocks are gonna fuck up regardless.

Just amazing you can spend $700+tax minimum on a gpu and still get shit. Then again Space Invaders RTX Edition shows that even spending well over a grand doesn't buy you properly tested components. What a shame.

Frankly, though, there's no reason not to get a platinum rated power supply IMO. Efficiency and build quality tend to correlate in the world of power supplies because something haphazardly put together will by dint of the laws of physics be inefficient. Even in small form factor land both Silverstone and Corsair do platinum rated bricks and there's even a Silverstone titanium rated SFX-L power supply out there.
 
  • Agree
Reactions: Allakazam223
Frankly, though, there's no reason not to get a platinum rated power supply IMO. Efficiency and build quality tend to correlate in the world of power supplies because something haphazardly put together will by dint of the laws of physics be inefficient. Even in small form factor land both Silverstone and Corsair do platinum rated bricks and there's even a Silverstone titanium rated SFX-L power supply out there.
I agree when the budget allows for it. Some of those high end psus have equally high prices. I got a good deal on a supernova g2 and I fully trust the Superflower unit inside of that. So for now, I stick with my gold.
 
  • Agree
Reactions: Aberforth
AMD is having a CPU event on the 8th of October, current gen might be a bit cheaper on the used market in a month or so if you can wait for the new ones to release.

It's really cool how Ryzen turned the i7 7700 into a cheap i3. The current gen i3 10300 have the same amount of cores/threads as the 7700 and slightly faster base and boost speeds.
@DanteAlighieri think about that when buying something new, we are in a golden age where really good is really cheap so don't get stressed out by looking at charts showing that a $500 CPU is 54% faster than a $250 CPU because the 250 CPU is still fantastic for the price.
 
Ok so technically this thread is about GPUs, but I was looking at the 80 Plus Wikipedia article and found a link to Plug Load Solutions' database of power supplies that meet different certification criteria (I think this is site is maintained by an industry group). Using that database, you can view the actual test results for specific PSUs instead of just certification tiers. (Aside: the industry should really change the certifications' branding so that it's clear which tier is best. It's needlessly difficult to remember if "Platinum" or "Titanium" is the best certification, and since the criteria for Titanium already specifies 90%+ efficiency at 10%, 20%, 50%, and 100% output, an easy solution might be to rebrand Titanium as "90 Plus". No need to pay me.)

It's unfortunate that 10% efficiency is only accounted for by the Titanium certification. This makes sense if you're buying a PSU for some industrial application where it'll be under load all or most of the time. But it can be misleading for consumers whose machines spend lots of time idle or running low-wattage activities like web browsing.

People end up buying PSUs that appear efficient based on their 80 Plus certifications but are actually quite inefficient most of the time they're in use. Check it out:

80 Plus CertificationMake/Model10% load20% load50% load100% load
BronzeEVGA 850 B5
74.90
83.48
87.52
84.70
SilverThermaltake BT 50078.6285.1989.1288.57
GoldEVGA SuperNOVA 750 G280.8887.0190.5690.12
PlatinumRosewill Quark 65078.9890.3692.8991.1
 
Ok so technically this thread is about GPUs, but I was looking at the 80 Plus Wikipedia article and found a link to Plug Load Solutions' database of power supplies that meet different certification criteria (I think this is site is maintained by an industry group). Using that database, you can view the actual test results for specific PSUs instead of just certification tiers. (Aside: the industry should really change the certifications' branding so that it's clear which tier is best. It's needlessly difficult to remember if "Platinum" or "Titanium" is the best certification, and since the criteria for Titanium already specifies 90%+ efficiency at 10%, 20%, 50%, and 100% output, an easy solution might be to rebrand Titanium as "90 Plus". No need to pay me.)

It's unfortunate that 10% efficiency is only accounted for by the Titanium certification. This makes sense if you're buying a PSU for some industrial application where it'll be under load all or most of the time. But it can be misleading for consumers whose machines spend lots of time idle or running low-wattage activities like web browsing.

People end up buying PSUs that appear efficient based on their 80 Plus certifications but are actually quite inefficient most of the time they're in use. Check it out:

80 Plus CertificationMake/Model10% load20% load50% load100% load
BronzeEVGA 850 B5
74.90
83.48
87.52
84.70
SilverThermaltake BT 50078.6285.1989.1288.57
GoldEVGA SuperNOVA 750 G280.8887.0190.5690.12
PlatinumRosewill Quark 65078.9890.3692.8991.1
Obviously there's a connection to quality here, but is efficiency even really the best metric? In the absence of all other criteria, minimizing inefficiency across all load ranges is better, but is it possible to do so without extra circuitry with extra components, and thus more components that can fail? This sounds like it could account for some of the extra price out at the expensive end. Not just more components, but having to specify better ones to make up for an inherently more vulnerable design.

That said, if there's a correlation with delivering more constant output voltages across typical loads, then obviously that's a good thing. Is there?
 
Obviously there's a connection to quality here, but is efficiency even really the best metric? In the absence of all other criteria, minimizing inefficiency across all load ranges is better, but is it possible to do so without extra circuitry with extra components, and thus more components that can fail? This sounds like it could account for some of the extra price out at the expensive end. Not just more components, but having to specify better ones to make up for an inherently more vulnerable design.

That said, if there's a correlation with delivering more constant output voltages across typical loads, then obviously that's a good thing. Is there?

So I actually made that post with only efficiency in mind - not efficiency as a proxy for quality (somebody else made that claim, though I agree with it). "80 Plus" is a rating of efficiency, i.e., how much energy the PSU wastes as heat. The less efficient a PSU is for a given load %, the more electricity you have to pay for and the hotter it'll make the room your computer's in.

Let's say you buy a Rosewill Quark 650 because you see that on the box it's marked as having achieved the 80 Plus Platinum certification. You might do that thinking that in the long run, you'll pay for less electricity and your room won't get as hot. However, in reality, that means that when your computer is consuming 65w (a realistic load for idle or some undemanding activity), the PSU is pulling 82.3w from the wall. If your PSU met the Titanium certification, it would need to be >= 90% efficient at 10% load, meaning it would need to pull 72.2w from the wall to deliver 65w to the system. Over time, that'll add up.
 
It's really cool how Ryzen turned the i7 7700 into a cheap i3. The current gen i3 10300 have the same amount of cores/threads as the 7700 and slightly faster base and boost speeds.
@DanteAlighieri think about that when buying something new, we are in a golden age where really good is really cheap so don't get stressed out by looking at charts showing that a $500 CPU is 54% faster than a $250 CPU because the 250 CPU is still fantastic for the price.
I just want to know where my current build falls on logical increments so I know what parts are worth looking at. I think a 3070 will be my best bet so whatever CPU I get will need to be good enough to not be bottlenecked. Honestly, I have no idea what to look for to get that.
 
It's really cool how Ryzen turned the i7 7700 into a cheap i3. The current gen i3 10300 have the same amount of cores/threads as the 7700 and slightly faster base and boost speeds.
@DanteAlighieri think about that when buying something new, we are in a golden age where really good is really cheap so don't get stressed out by looking at charts showing that a $500 CPU is 54% faster than a $250 CPU because the 250 CPU is still fantastic for the price.
This holiday season should be a really great time to build a PC, probably better than last year since GPU prices still kinda sucked with AMD/NVidia refusing to offer better value than the old RX580s. For the majority of even enthusiasts today, even "mid-range" $150-250 CPU/GPUs are most than enough to build a very capable PC that can play everything you want at the highest settings for some time.
 
I just want to know where my current build falls on logical increments so I know what parts are worth looking at. I think a 3070 will be my best bet so whatever CPU I get will need to be good enough to not be bottlenecked. Honestly, I have no idea what to look for to get that.

Almost any current CPU will be good enough to run at 60+ fps with plenty of margin not only for know but for some time going forward. Same with a 3070. If the GPU is capable of rendering many more frames than the CPU can feed it then you can always up the resolution and pile on effects/higher settings until they're running at parity. The same can't be done on the CPU, you can't drag a slider that makes the enemy AI consume more cycles per tick, so balance wise I think it's always wise to have more power on the GPU side. Unless it's a 486 and a Voodoo 2 like one guy I knew, that was far too lopsided.

With PSUs I've heard a pretty sensible argument for buying way more than you need, like using a 850w [number]+[metal] PSU for something that runs under 500-600w at max load, and the reason is that during normal operation the PSU won't break a sweat, the fans won't even turn on. It just doesn't generate enough heat for that at a ~250w load.
 
  • Informative
Reactions: DanteAlighieri
From Igor's Lab:

"Owners of cards, which still run really stable despite six solids, owe everything to the very good chip quality. Owners of cards with MLCC, where errors still occur, may be annoyed about a GPU that does not even consistently cope with the stored voltage/frequency curve. This is exactly the point where the board partners couldn’t test anything at all with the first cards due to the lack of suitable drivers. There are certainly many cards in circulation here that would not have been suitable as OC cards.

The fact that NVIDIA has split the GPU’s power supply voltages between NVVDD and MSVDD also shows that they are well aware of the problem. I noticed that the MSVDD has much less changes and is generated independently from the NVVDD, so you should be able to get along with a well equipped MLCC group. So more MLCC does no harm if the rest of the layout allows this interpretation. Without it, however, becomes slow and sluggish."

Oh hey, a rushed and sloppy launch.
 
With PSUs I've heard a pretty sensible argument for buying way more than you need, like using a 850w [number]+[metal] PSU for something that runs under 500-600w at max load, and the reason is that during normal operation the PSU won't break a sweat, the fans won't even turn on. It just doesn't generate enough heat for that at a ~250w load.
Yes, peak efficiency for PSUs are generally in the 50-70% range. But, take into account how often you're pushing to a near a max load, which is very rarely even if you're a heavy gamer. Recommended PSU levels suggested by GPUs are very conservative and safely high.

Thus, it's generally best to just buy some decent quality ~600w PSU and you're good. Unless you're really into overclocking, dual GPUs, etc...
 
  • Like
Reactions: Allakazam223
Yes, peak efficiency for PSUs are generally in the 50-70% range. But, take into account how often you're pushing to a near a max load, which is very rarely even if you're a heavy gamer. Recommended PSU levels suggested by GPUs are very conservative and safely high.

Thus, it's generally best to just buy some decent quality ~600w PSU and you're good. Unless you're really into overclocking, dual GPUs, etc...
I say whatever you can get a good deal on. Got my 1300w G2 for $100.

Just make sure it's a good build platform and the rest is downhill from there.
 
From Igor's Lab:

"Owners of cards, which still run really stable despite six solids, owe everything to the very good chip quality. Owners of cards with MLCC, where errors still occur, may be annoyed about a GPU that does not even consistently cope with the stored voltage/frequency curve. This is exactly the point where the board partners couldn’t test anything at all with the first cards due to the lack of suitable drivers. There are certainly many cards in circulation here that would not have been suitable as OC cards.

The fact that NVIDIA has split the GPU’s power supply voltages between NVVDD and MSVDD also shows that they are well aware of the problem. I noticed that the MSVDD has much less changes and is generated independently from the NVVDD, so you should be able to get along with a well equipped MLCC group. So more MLCC does no harm if the rest of the layout allows this interpretation. Without it, however, becomes slow and sluggish."

Oh hey, a rushed and sloppy launch.
Never buy electronics at launch. I don't care if the shiny new 3080 TI Super Black Edition FE comes with a lifetime supply of enchanted frozen burritos. Don't do it.
 
More AMD stuff, this time from Newegg Insider whatever that is.
AMD-Radeon-RX-6000-Series-Newegg.png


What catches my eyes is the 5700XT/6700XT specs. Surely, if true, the 6700XT won't be worse than the 5700XT, and if so potential performance gains can be guesstimated. It looks to be doing more with less(clock speed, membus, bandwidth) if we assume that it is also faster than its predecessor. That would be a big thing for AMD.
 
  • Informative
Reactions: Allakazam223
More AMD stuff, this time from Newegg Insider whatever that is.
View attachment 1625275

What catches my eyes is the 5700XT/6700XT specs. Surely, if true, the 6700XT won't be worse than the 5700XT, and if so potential performance gains can be guesstimated. It looks to be doing more with less(clock speed, membus, bandwidth) if we assume that it is also faster than its predecessor. That would be a big thing for AMD.

Hmmm. Big RAM good, but fast RAM better. Why no HBM?
 
Never buy electronics at launch. I don't care if the shiny new 3080 TI Super Black Edition FE comes with a lifetime supply of enchanted frozen burritos. Don't do it.
That goes along with my rule which is never buy the "flagship" model of anything.
The price premium compared to the second best iteration is often not at all worth it.
 
Hmmm. Big RAM good, but fast RAM better. Why no HBM?
HBM is too expensive. It’s also only useful in Datacenters that require that bandwidth, not gaming. Just because RAM can be fast doesn’t mean it will be useful for games, which prefer clock speeds
 
Back