Are bar graphs good enough?

[specfreq]

Price vs. Performance Link

These days, all the major tech sites use bar-graphs to compare hardware against each other, but that only tells one side of the story. Ultimately, this method of reviewing leaves it up to the consumers to figure out if they are actually getting a good price on the latest and greatest next best thing. Remember when there were price/performance charts? (Pepperidge Farm remembers) In short, we all need to start seeing them come back again.

---

I've slapped together a list with 3 generations of GPUs to show you what shopping for hardware **could** look like again. My price data is the from both AMD and Nvidia's websites most affordable sellers per part, if they didn't have a direct link to a seller I did not include it in this chart. 

Performance data is simply SPGFLOPS I pulled from Wikipedia for AMD and Nvidia as an example. GFLOPS is *not* indicative of real world performance in an application like a video game, **please do not mistake it for FPS.** RX 480, RX 470, RX 460 data is rumors and speculation at this time.

---

The hardest part for newcomers to the PC world is wading through the endless sea of hardware from dozens of OEMs. A picture speaks a thousand words, a tech site will reach a hundred thousand more.

[Enderman]

if you bother to watch LTT videos you will see they actually made graphs on performance per dollar

[specfreq]

I just bothered to watch a LTT video. Yeah they made a graph, a dollars per frame bar-graph which you need then to compare to another bar-graph to get a sense of value.

 

[Bananasplit_00]

there is usiually a fps/$ chart somewhere in the bechmark lists if your watching any good reviewer tbh those old graphs are hard to read and are sometimes good but not needed, what you usiually want to know is the fps/$ spent and then comparing that to the price of the card you planned on and cards around the same price

[specfreq]

5 minutes ago, Bananasplit_00 said:

there is usiually a fps/$ chart somewhere in the bechmark lists if your watching any good reviewer tbh those old graphs are hard to read and are sometimes good but not needed, what you usiually want to know is the fps/$ spent and then comparing that to the price of the card you planned on and cards around the same price

I've haven't seen a price/perf scatterplot in years, maybe you can link one?

They give you both X and Y for price and performance but then separate them into 2 different graphs. What do you mean that it's harder to read?

[Bananasplit_00]

1 minute ago, specfreq said:

I've haven't seen a price/perf scatterplot in years, maybe you can link one?

They give you both X and Y for price and performance but then separate them into 2 different graphs. What do you mean that it's harder to read?

i find them harder to read then a bar graph where either hifger or lower depending on what its about is better, you get to see what is best preformance per dollar, i dont quite see the apeal in one of those big graphs with all the different GPUs but sure some people might i just dont think its needed

[Clanscorpia]

Those graphs can be difficult to read for some people though

[specfreq]

21 minutes ago, Bananasplit_00 said:

i find them harder to read then a bar graph where either hifger or lower depending on what its about is better

I swapped X and Y so now the ones at the top are better. The further right you go, the more you pay for it.

[Enderman]

The reason price performance graphs are stupid is because they do not include the total cost of the system, only that of the GPU

 

as linus has said before, that is the wrong way to measure value

 

You could use your iGPU instead of a dedicated gpu, so you spend $0 on a GPU for horrible performance

but since you're dividing that performance by 0, you get infinite price performance

 

and then basically you end up buying the cheapest GPU possible because it has great price performance

which is completely wrong if your PC costs $1000

then you have a great PC with a horrible GPU all because of "performance per dollar"

 

there are many better ways to choose a GPU, such as picking the best performing one that fits your budget, or picking the cheapest one taht can play your games at high settings, or stuff like that

 

price performance is a meaningless number unless you include the total system cost in it, which is pretty much impossible since everyone's system costs a different amount

so to get an idea of performance/dollar you need to calculate it yourself with the cost of your system

not have some reviewer do it for you

[specfreq]

16 minutes ago, Enderman said:

You could use your iGPU instead of a dedicated gpu, so you spend $0 on a GPU for horrible performance

but since you're dividing that performance by 0, you get infinite price performance

That's not how this works, because you are not comparing integrated GPUs against discrete GPUS when shopping for discrete GPUs.

Price vs Performance is a meaningful number when when testing against a constant, when reviewer gets a GPU they use the same CPU, motherboard, RAM, resolution, clockspeeds, etc. and only swap out the GPU. You don't compare the rest of everyone elses system's to the overall cost when you're ONLY testing the GPU.

[Enderman]

4 minutes ago, specfreq said:

That's not how this works, because you are not comparing integrated GPUs against discrete GPUS when shopping for discrete GPUs.

Price vs Performance is a meaningful number when when testing against a constant, when reviewer gets a GPU they use the same CPU, motherboard, RAM, resolution, clockspeeds, etc. and only swap out the GPU. You don't compare the rest of everyone elses system's to the overall cost when you're ONLY testing the GPU.

ok, so you get a cheap GT730 for $40, and it gets 20fps in a game

then you compare that to a $700 1080 that gets 200fps in a game

 

OMG!

THAT CHEAP GPU HAS SO MUCH BETTER PRICE PERFORMANCE!

YOU SHOULD TOTALLY GET THAT FOR YOUR $2000 PC!

 

do you see how your logic is flawed?

[specfreq]

That's right! We're talking about a ratio. But people don't buy the most efficient card for the framerates, they buy the most efficient card for the their budget.

[Prysin]

1 hour ago, Enderman said:

if you bother to watch LTT videos you will see they actually made graphs on performance per dollar

If you believe in LTT videos and how they test things you should re-evaluate your sanity before using said info to buy anything