EXTREMES in testing hardware, normal budget combinations never tested

[agent_x007]

I would test these Intel CPUs of the current gen (no reason to test older, some are just renamed older gen CPUs anyway): 14100 (4 cores, the cheapest CPU), 14500 (14 cores, better value than 14 core 14600K), 14700K (20C - price in the middle between 14500 and 14900K) and 14900K (24C - the best standard CPU Intel makes, 14900KS is overpriced limited edition).

Good. Here's a glance at market (amazon in this case) :

And here's a good performance option for the similar price (at least at the moment, since it's "limited time offer") :


You test for buyers/consumers after all, and not in vacuum (I'm leaving AMD out for this one).

So, would you buy 14500 and a ~250$ GPU (example : 3060 8GB), or 13400F with ~300$ GPU (examples : 3060 12GB/4060 8GB) ?
Both combos are new, with full warranty.

 

[GodisanAtheist]

Ok, not reading through this dumptruck of a thread, but I will say that I broadly agree with the OP's premise, although not entirely in how he suggests things are executed.

I think everyone who is a fan of hardware has a good time when a reviewer does something just for the fun of it, like when a site will do a retro review and test 980Ti's or Vegas or something vs modern 2024 games just to see how the hardware holds up. It can be useful for people holding on to cards 2-3 generations old too (a lot of people out there), so they can get a feel for where their card might fall in the current stack and if it makes sense to upgrade yet.

Sometimes the results can be surprising in a good way, other times cards fall off really quickly, sometimes the drivers stop getting performance patches and the cards you thought your old card was equivalent to actually aren't. I know zWormz on youtube does his retro Sundays, where he grabs some old card for $100 off ebay and just see how it performs against his test suite. Its not very scientific, but it is *fun* and somewhat edifying. I thought my 980ti was roughly equivalent to the 1660 and 3050, but those newer cards outpaced the old 980ti quite a bit in newer games, which was kind of a surprise.

I kind of see the OP's point the same way. I think it would be *cool* (not scientific, just cool from a hobbyist's perspective) to maybe put three systems, a high end spared no expense system ($3000), mid range CPU + Midrange GPU ($1000), and an Entry Level CPU + Entry Level GPU ($500) through say the TPU test suite and see where the performance lands for each of them. We know NV has some overhead with their drivers, so maybe we'd get some surprising results in the mid and entry ranges where AMD actually surprises folks and performs better than expected.

No need to do something like this for every review, but it would be a *fun* mid-gen article to read through just to see roughly what real world performance a $3000/$1000/$500 system will get you.

 

[BoggledBeagle]

...
No need to do something like this for every review, but it would be a *fun* mid-gen article to read through just to see roughly what real world performance a $3000/$1000/$500 system will get you.

That would be interesting for sure.

I was more interested in helping consumers build their mid range or lower end systems which do not contain any of the top of the line components used in the reviews.

An example problem:
I have $800 budget for both CPU and GPU, what should I get, when I do not care about RT?

The consumer will end up with say 7 possible combinations of hardware, not having a clue, which of these systems performs the best, because reviews do not cover this stuff at all.

 

[adilazimdegilx]

I still see no point in this and it's practically impossible to set up an all-satisfactory case anyway. Not to mention, not needed.
At first, it does make sense to test budget builds around 800-1000$ but when you think about it these builds will end up in 100% gpu limit anyway. So you can just refer the GPU reviews of W1zzards. Around those price point, Ryzen 7500f/14500 based builds would give identical results with mid to high range graphics card up to rtx4070 against 7800x3d/14900k. I would be shocked to see a 5% difference in average results. So in that case, just refer GPU charts...
And about the impossibility of optimal build, simply, the variables are too many. It's not just about picking a system. That's actually the easy part.
What resolution you will target? 1080p? 1440p? Surely not 4k, right? But why?
What level of upscaling? Native makes sense but surely at least a quality level upscaling is advised to average gamer in most games to get best picture quality/fps ratio.
What graphic preset you will use? Max? High? Or optimized for each game which is more sensible to average gamer but how do you define optimized? What about RT?
Needlessly to say the list goes on, and you can't find an objective optimal case. Simply because, it depends.

That's all being said there are many channels building and benchmarking PCs in like 1000$ budgets (easy cash grab video to make). And after watching many of them just 'cause I like pc building like many here, I can comfortably say:

Refer to the GPU reviews.

 

[AusWolf]

GPUs are tested with the best CPU available.
CPUs are tested with the best GPU available.

Why are sensible budget combinations never tested? As a customer you simply cannot find information anywhere about how these combinations perform.

For example graphic cards with budget friendly Intel 14500 CPU. (What is the AMD price equivalent to this CPU?)

Or CPUs with RX 7800 XT or RTX 4070 (super?) ?

I look at reviews as purely scientific information these days. You'd want to know how well a certain piece of hardware performs in ideal conditions, and not in certain hardware combinations, of which there's countless variations. It's a review, not buyer's guide.

There's two things that I do find disturbing in reviews, though:

1. Commentary. A lot of times I see a chart with data that I personally find acceptable, but then the reviewer goes "what an utter turd this CPU/GPU is", and I'm left scratching my head what the hell is going on. Your turd is someone else's gold.

Edit: It's equally disturbing when a GPU performs 5% better than another one, which is absolutely nothing in my books, but the reviewer goes "what an amazing piece of tech, it's so much better". Seriously, what the heck?

2. No mainstream CPUs, like Intel's non-K chips. TPU did a review on the R7 7700 non-X, big kudos for that, but a lot of reviewers don't bother, and I don't understand why. Personally, I'm much more interested in these sensible, budget-friendly options than the top of the top.

 

[biorpg]

@Bill_Bright
That was implied, obviously. I am not suggesting taking the results from tests that aren’t made with as many possible bottlenecks removed as possible. You will never be able to remove all of them completely, that’s implausible, and variance in test runs also obviously exists, but aforementioned extrapolation doesn’t need perfect baseline to go off on, “reasonably established as valid” is good enough in this case.

Example: W1zz could THEORETICALLY lessen the bottleneck on CPU review by using an overclocked to the gills 4090 Matrix. Or he could THEORETICALLY reduce the bottleneck on the GPU reviews by, again, OCing the hell out of the 14900K under an industrial chiller and using a golden sample chip with perfectly tuned memory. Both will be theoretically better, but in practice, what he has is enough.

Oh, @BoggledBeagle, what I described above is what ACTUAL Extreme testing would be for reviews. What W1zz does is just sane and reasonable best practices.

(Says the man running a GTX 1070 alongside a 5900X.)
BTW: The performance bottleneck is a state of relativity, not a space that can be filled.

 

[AleXXX666]

You can find a lot on youtube.

yeah and fake ones too

 

[biorpg]

what? I dont see how thats the case at all. You get a definition of what could be the best in any given scenario. You can extrapolate performance from that. There are too many combinations.

Its unrealistic and testing low to mid end hardware wont fix it.

Further tweaks can be made per game, or engine if you want to be that pandantic.

If you dont have the best GPU or CPU then you are simply slower. Thats the point. You can go look at old tests on old CPUs and how they compare to new so you can make educated guesses on where you might end up landing.

There are hundreds of CPU and GPU comparisons to each other.

I think buyers are just lazy.

There are many newer systems that run older games (like Half-Life 1/2) worse than the systems in their hay-day did. NVIDIA's GeForce drivers are tailored around the popular games of the time, and it can be difficult to reproduce an entirely backdated software system to play on that is compatible with your newer hardware. Also, half of new hardware is always slower than half of the old hardware.

 

[DaemonForce]

There are many newer systems that run older games (like Half-Life 1/2) worse than the systems in their hay-day did.

I have experiences with this.
Intel Pentium MMX 233MHz + ATi Rage XL 8MB AGP
Seems to run things okay on XP and Longhorn.

AMD K6-2/300 ↗400 + ATi Rage Magnum 16MB AGP
Runs everything great.

Pentium 4 3.00E ↗3300 + Radeon 9200LE 128MB AGP
Kind of drags in DX9 stuff but all the old stuff still runs great.

Pentium 4 3.00E + AH3450 512MB AGP
CONDUCTOR WE HAVE A PROBLEM

DX9 stuff finally starts to run right but older stuff breaks without patches. Half the ATi demos that tested technologies of an earlier time are broken. No idea about the nVidia side but as time goes on with each AMD card picked up, this only gets worse with infini-ghosting, red/rainbow seams in some areas that were NEVER on my radar, broken shaders and the improper loading/caching of textures.

On the DX10/11 side I haven't run into that yet but I'm guessing it's only a matter of time. This may be a valid criticism to hold onto older hardware when gaming/streaming old titles that don't look right on new/virtualized systems. It might be time to pick up a capture card too.

 

[BoggledBeagle]

I found a test that includes the CPUs I mentioned.

Test • Intel Core i5-14500 / Core i5-14400 / Core i3-14100

Après avoir tester les versions haut de gamme (K) de la 14ème génération de processeurs Core d'Intel, place cette fois à l'entrée et au moyen de gamme.

hardwareand.co

You can see, that in FHD with RTX 4090 the 14500 performs notably worse than 14600K. But how would it perform with 4070 at 1440p? Would not these CPUs perform nearly identical, with the 14500 being cheaper and more energy efficient???

A consumer may spare money by getting the cheaper CPU, may pay less for electricity, for a cooler, but according to the test above he would buy 14600K, which is simply a wrong decision for his use.

 

[masterdeejay]

I have experiences with this.
Intel Pentium MMX 233MHz + ATi Rage XL 8MB AGP
Seems to run things okay on XP and Longhorn.

AMD K6-2/300 ↗400 + ATi Rage Magnum 16MB AGP
Runs everything great.

Pentium 4 3.00E ↗3300 + Radeon 9200LE 128MB AGP
Kind of drags in DX9 stuff but all the old stuff still runs great.

Pentium 4 3.00E + AH3450 512MB AGP
CONDUCTOR WE HAVE A PROBLEM

DX9 stuff finally starts to run right but older stuff breaks without patches. Half the ATi demos that tested technologies of an earlier time are broken. No idea about the nVidia side but as time goes on with each AMD card picked up, this only gets worse with infini-ghosting, red/rainbow seams in some areas that were NEVER on my radar, broken shaders and the improper loading/caching of textures.

On the DX10/11 side I haven't run into that yet but I'm guessing it's only a matter of time. This may be a valid criticism to hold onto older hardware when gaming/streaming old titles that don't look right on new/virtualized systems. It might be time to pick up a capture card too.

HD3xxx and HD4xxx series with AGP has some issues with a few chipsets. (try to disable AGP fast write, and update chipset drivers if possible)
Back then Ati drivers was awfull. I only remember the black screens and vpu recovery warnings, opengl games was much slower...etc.
For AGP i recommend the Nvidia 6-7 series. 6600GT-7600Gt was a great deal back then.

My builds for AGP:

Abit IC7, 3Ghz nortwood 800fsb, 4x256 ram, Ti4200 or 6600GT or 7600GT AGP

For the C2Q (Q9550) i use Quadro Fx5500 (7900GTX) or FX5600 (8800 GTX)


I have a lot of old ati cards but i dont want to use them because of the drivers.


So for DX8 games the best i think the Geforce4 Ti series.
DX9 7xxx series. Windows 98/XP
And for DX10/DX11 i think all modern card works

But for Win98 7950GT is the fastest that has driver.

 

[Onasi]

(Says the man running a GTX 1070 alongside a 5900X.)
BTW: The performance bottleneck is a state of relativity, not a space that can be filled.

Yeah, considering that I don’t play AAA games at all and the GPU market being a mess currently I am mostly content with what I’ve got. Guess it’s that relativity kicking in.

 

[DemonicRyzen666]

GPUs are tested with the best CPU available.
CPUs are tested with the best GPU available.

Why are sensible budget combinations never tested? As a customer you simply cannot find information anywhere about how these combinations perform.

For example graphic cards with budget friendly Intel 14500 CPU. (What is the AMD price equivalent to this CPU?)

Or CPUs with RX 7800 XT or RTX 4070 (super?) ?

Basically, you're asking why not benchmark with the Steams survey's "most used CPU" & most "used graphic cards" together.

Maybe you can find someone who did a budget build & you can see the benchmark scores they get with such system I'm sure if search long enough you'll find it.

 

[DaemonForce]

HD3xxx and HD4xxx series with AGP has some issues with a few chipsets. (try to disable AGP fast write, and update chipset drivers if possible)

Abit IC7, 3Ghz nortwood 800fsb, 4x256 ram, Ti4200 or 6600GT or 7600GT AGP

That's why you know this.
ABit IS7, 3GHz Prescott, 2x256MB, Radeon 9200LE
That was my initial build. By the end it was 2x1GB and the Asus AH3450.
JFC that's when the chipset still had its OE fan.

Yeah I'm not in any need to power that thing on anymore. Last straw was moving large sums of data from a dying HDD. The chipset would overheat, cut out connections and bluescreen, taking the data with it. The entire reason I got into a habit of adding CRC checksums to movie names was because of that. Dealt with it 7 years ago, no more issues with my data. Still not getting those old game experiences the way they were meant to be, so a lot was sacrificed.

 

[Wirko]

So many scientists here are assuming that there's exactly one bottleneck in any given system. So (to get the analogy closer to real bottle necks), if you're watering your garden with 10 metres of 1/2" hose, then extend it with another 10 metres of 3/4" hose, the flow rate will stay the same, right?

 

[Bill_Bright]

You are missing the point. Testing is what happens before it goes to the track to make record attempts or gets pulled up in front of the cameras for publicity.

No, I didn't miss the point. This thread is about "review" or "evaluation" testing by 3rd party reviewers to see, among other test criteria, (1) if the product meets published specs and (2) how well the product compares to its competition in real-world and simulated scenarios. It is not about the product makers trying to obtain marketing fodder for their photoshopped, AI generated, publicity shots.

So many scientists here are assuming that there's exactly one bottleneck in any given system.

Huh? Please show us one, just one scientist, one recognized expert, one self-proclaimed expert, or one person in this thread who made such a claim or assumption. I just read through the thread and I don't see one.

Giving an example of one bottleneck does not in anyway suggest that is the only bottleneck in the system. It is just illustrating one - potentially one of many.

So (to get the analogy closer to real bottle necks), if you're watering your garden with 10 metres of 1/2" hose, then extend it with another 10 metres of 3/4" hose, the flow rate will stay the same, right?

I don't understand your point. This is just saying the same thing already said starting in post #3 above, and reiterated several times by others since, and that is NOT in dispute. So not sure your point for repeating what has already be said multiple times, and is fully understood.

But to your comment, setting aside the fact the flow rate will decrease anytime you add length to a hose due to the additional friction in the new hose section (regardless the fact that section in your example has a larger diameter and thus increases "pressure" in that section - see Bernoulli's Principle), you totally ignore any possible bottleneck that may be caused by whatever is on the end of the hose - like a coupler or nozzle - that might be restricting (creating a bottleneck) the free flow of that water out the end. Not to mention, gravity (if not on totally level ground), curves or bends in the hose, even how sharp a bend or curve in the hose can impose a bottleneck. In other words, there are "several" potential bottlenecks, even in your example - not just one.

So while bottleneck is a "fluid" dynamics term, probably it better to stick with other motive forces in analogies, like electrons flowing through conductors. I'm just saying... .

 

[ThrashZone]

Hi,
Frankly I don't see why this thread is in general hardware instead of site feedback about product reviews.

Nobody "Not even the OP" is offering up any tests/ reviews of hardware combinations.

So is everyone off topic lol

 

[AusWolf]

I found a test that includes the CPUs I mentioned.

Test • Intel Core i5-14500 / Core i5-14400 / Core i3-14100

Après avoir tester les versions haut de gamme (K) de la 14ème génération de processeurs Core d'Intel, place cette fois à l'entrée et au moyen de gamme.

hardwareand.co

View attachment 340785

You can see, that in FHD with RTX 4090 the 14500 performs notably worse than 14600K. But how would it perform with 4070 at 1440p? Would not these CPUs perform nearly identical, with the 14500 being cheaper and more energy efficient???

A consumer may spare money by getting the cheaper CPU, may pay less for electricity, for a cooler, but according to the test above he would buy 14600K, which is simply a wrong decision for his use.

I get your point, but we can't expect reviewers to do such tests with every single CPU+GPU combination in existence. This is where applying some common sense becomes useful. For example, most of the results in the above chart are above 140 FPS minimum. Does an average gamer need more than that? Not likely.

 

[arbiter]

I found a test that includes the CPUs I mentioned.

Test • Intel Core i5-14500 / Core i5-14400 / Core i3-14100

Après avoir tester les versions haut de gamme (K) de la 14ème génération de processeurs Core d'Intel, place cette fois à l'entrée et au moyen de gamme.

hardwareand.co

View attachment 340785

You can see, that in FHD with RTX 4090 the 14500 performs notably worse than 14600K. But how would it perform with 4070 at 1440p? Would not these CPUs perform nearly identical, with the 14500 being cheaper and more energy efficient???

A consumer may spare money by getting the cheaper CPU, may pay less for electricity, for a cooler, but according to the test above he would buy 14600K, which is simply a wrong decision for his use.

Thing is to do testing they did for all that is probably took them over a week as setting up doing games testing isn't a fast process as gotta do mutliple runs of each test. adding 1 gpu in addition to it doubles the testing work.

 

[BoggledBeagle]

I do not think that the demand would be that high. I would start with analysis of the games - you really do not need to test dozens of games each time. You can analyze how the games perform, and then test just one of the group of the games that perform very close to each other.

You do not need to test every CPU available from each manufacturer, you may limit the number to 3 or 4 CPUS from both manufacturers to cover the price spectrum.

The same with GPUs. 3 or 4 from each manufacturer.

That is max 64 tests. Less than that, because few people will want to use the most expensive GPU with the cheapest CPU and vice versa.

Let us say 40 tests per game, with 5 games it makes 200 tests, which can be automated somehow I believe.

And the consumers would get a lot of useful information and would buy just what they really need.

I get your point, but we can't expect reviewers to do such tests with every single CPU+GPU combination in existence.

I already wrote what I quoted above. I proposed significant reduction of the number of games tested, and testing with just a few CPUs/GPUs.

This is where applying some common sense becomes useful. For example, most of the results in the above chart are above 140 FPS minimum. Does an average gamer need more than that? Not likely.

I do not think that your measure of something "being good for an average gamer" is reflected in any reviews available anywhere today.

BTW if you wanted to find a best combination (cheapest?) of hardware which would provide that according to you "satisfactory performance of 140 fps", you would also need to test a lot of hardware combinations.

 

[AusWolf]

I already wrote what I quoted above. I proposed significant reduction of the number of games tested, and testing with just a few CPUs/GPUs.

What games would you reduce it to? Some are more CPU sensitive, others are more GPU sensitive. Some need more VRAM, some need more CPU cache. You can't test with 5 games and call it an in-depth review.

I do not think that your measure of something "being good for an average gamer" is reflected in any reviews available anywhere today.

BTW if you wanted to find a best combination (cheapest?) of hardware which would provide that according to you "satisfactory performance of 140 fps", you would also need to test a lot of hardware combinations.

What I mean is, you know what you want from your PC and you should look at charts with that regard.

I built a rig for someone not long ago who wasn't happy with his last setup because it didn't max out his 360 Hz monitor in CS2 all the time. I, on the other hand, am happy anywhere over 40 FPS. This is why I was vexed at some 6500 XT reviews when the card did 55 FPS in X game, and the reviewer commented "what an utter piece of garbage, it can't even do 60". Fun fact that I actually bought a 6500 XT later just for shits'n'giggles, and never sold it because it ended up being the perfect emergency replacement card, being able to play most of my games with modest graphical settings. The guy I mentioned earlier would probably join the reviewer in calling it a turd.

All in all, read every review with a grain of salt and make up your own conclusion. Just because the 4090 can do a bazillion FPS when you pair it with a high-end CPU, it doesn't automatically mean you need one in your setup. It also doesn't mean that you need a bazillion FPS in the first place. No single review can teach how to be an informed buyer.

 

[Gica]

You can see, that in FHD with RTX 4090 the 14500 performs notably worse than 14600K. But how would it perform with 4070 at 1440p? Would not these CPUs perform nearly identical, with the 14500 being cheaper and more energy efficient???

I used 13500 from February to November together with 3070Ti. From November it was replaced with 14700KF. I haven't seen any difference between the two in AAA games. With the others, where the video card generates hundreds of frames per second, I didn't even bother to compare. I can survive if instead of 280 FPS I have "only" 260 FPS.

P.S. There are differences in 3DMark in general (CPU Test and combined) but the Graphics Score is identical.

 

[KedsDead]

I simply look at the CPU review, where they use the fastest GPU to see how far it can be pushed. Then I look at graphics card review, the models I am looking into, then check to see which card is around the speed that the CPU I was looking can push the fastest graphics card. Then I make my choice on price/performance in the frame rate range.

 

[BoggledBeagle]

What games would you reduce it to? Some are more CPU sensitive, others are more GPU sensitive. Some need more VRAM, some need more CPU cache. You can't test with 5 games and call it an in-depth review.

I proposed analyzing the games and dividing them into groups of similarly performing games and testing just one game from this group. By that you would get information about all the games in this group.

Say you have 30 games, and you manage to divide them in 8 groups - you then test just 8 games, but get information about performance of all 30.

What I mean is, you know what you want from your PC and you should look at charts with that regard.

It does not help when the charts DO NOT CONTAIN the information you need.

 

[AusWolf]

It does not help when the charts DO NOT CONTAIN the information you need.

One single chart does not contain the information you need. Several of them do, however.

1. You look at a 4090 review. You see that it pushes the boundaries of any CPU in the right games with the right settings.
2. You look at CPU reviews. You see that the majority of them do absolutely fine with the 4090. Whether you have 280 instead of 300 FPS is inconsequential. With that info, you have your CPU choice.
3. You check GPU reviews, and see which one of them approaches the level of performance your CPU is comfortable with.
4. If you want a graphics card weaker than the 4090, then you needn't worry, your chosen CPU won't bottleneck it much, if at all (unless you went for a Celeron which everybody knows you shouldn't).
+1. It doesn't matter anyway because no graphics card below a 7900 XTX or 4080 Super is bottlenecked by any moderately capable new CPU (unless you call 280 instead of 300 FPS a bottleneck).

TL,DR: Bottlenecking is overrated anyway.