Discussion GPUs should be reviewed on an AMD 3900X AND intel 9900K system

[Kedas]

There are too many competitive variables at play that you cannot really say anymore THAT is the best platform for a GPU test.
So I would like to see GPU reviews on two systems 3900X and 9900K

Certainly when intel joins the GPU party.

 

[ZeroRift]

I'm surprised that this didn't get more attention, so I'm throwing my hat in here

I agree with the OP on this. In the Bulldozer era, this obviously wasn't a concern, but with threads like this one cropping up and showing potential differences in GPU performance between platforms, we should at least investigate to ensure that GPUs are actually platform agnostic, rather than assuming such.

 

[pandemonium]

I've wanted this for a while, but now it's even more relevant and I wholeheartedly agree. I can imagine the work load increase for benching being annoying on new release reviews though. Of course, that's entirely up to the review site to properly manage.

 

[Shmee]

Time permitting, this sounds like a good idea to me.

 

[Guru]

It would be great, but it is time consuming. I'd rather review sites actually focus on retesting GPU's and CPU's when doing a new review, so they have updated numbers, rather than using 3-4 months old numbers, while we have new game versions, new drivers and even new windows, where they performance might be completely different.

 

[loki1944]

There are too many competitive variables at play that you cannot really say anymore THAT is the best platform for a GPU test.
So I would like to see GPU reviews on two systems 3900X and 9900K

Certainly when intel joins the GPU party.

Actually I'd prefer it much more if they were reviewed on majority platforms by function; i.e. if gaming performance review do it on the most common gaming CPU. More than likely that's definitely not going to be a 3900X or 9900K in that case.

 

[Markfw]

Time permitting, this sounds like a good idea to me.

I agree

 

[Shmee]

Even better, do a 3800XT and 10700k, for a more modern review set, plus both are 8c16t so it is more apples to apples. Both are still typically cheaper than a 9900k and 3900X.

 

[Arkaign]

Even better, do a 3800XT and 10700k, for a more modern review set, plus both are 8c16t so it is more apples to apples. Both are still typically cheaper than a 9900k and 3900X.

Yeah. The XT refresh was slightly disappointing (mainly the price hike, the thing was slightly improved, but not worth $100 extra for 1-2% benchmark gain), but it does reflect the improved Zen2 silicon available whether or not you buy an X or XT model.

The 10700K is more or less a rebadged 9900K, and both the 10700K and 3800XT represent the best available 8C/16T parts on offer.

3600X and 10600K would be good as well I think with mid-range GPU reviews

 

[Kedas]

More than a year later I would prefer to see new GPU tests on an 3950X.
Simply because I have one.
Testers will soon change to an 4950X setup I assume.
But obviously also make a gaming intel setup including Marvel super heroes

 

[NTMBK]

Reviewer time is a limited resource. I would rather see them focus on more rigorous testing (i.e. frametime distribution, instead of just average framerates), and a wider range of games tested, rather than duplicating all their work on two different CPUs.

 

[Shmee]

This is a good point, too many reviews have so few games played.

 

[DAPUNISHER]

Hammer on box has switched to the 3950x for benching. He recaps why in the first couple of minutes of this vid -

It will be nice to see games tested by a pro on a non de facto Platform.

 

[aigomorla]

Time permitting, this sounds like a good idea to me.

no...
because people wont stop nag and demand for more when they have no idea how real methodology for tests work..
next its going to be asking for on a X570 vs B550 vs B450 vs Z270 vs Z170 vs (ect...)

its absolutely pointless.... reason being your holding the same core constant, so the differences will be seen in the card itself.
You start pulling cross platforms, and your asking for a can of worms not to be leashed, but thrown at you after someone took a dookie in the can because there can be so many different floating variances.

This is also why you keep testing on the same test bed, and if your test bed dies in the middle, then your SOL and need to restart the test all over on a new test bed, and can't just start again where you left off.

 

[Kenmitch]

Hammer on box has switched to the 3950x for benching. He recaps why in the first couple of minutes of this vid -

It will be nice to see games tested by a pro on a non de facto Platform.

RTX 3090 PCI 4.0 vs 3.0 testing should be interesting.

Besides he'll want to switch over once Zen 3 drops!

 

[DrMrLordX]

RTX 3090 PCI 4.0 vs 3.0 testing should be interesting.

Besides he'll want to switch over once Zen 3 drops!

It certainly was easier to decide on a test bed when Intel produced the fastest non-HEDT desktop CPU available and also wasn't lacking for PCIe bandwidth. Now you've got the 10900k which is probably the fastest gaming CPU Intel has "out of the box" (not gonna say "stock" since motherboard can determine behavior in some cases), while for AMD their fastest gaming CPU is probably the 3800XT if tweaked or the 3950X if not.

 

[aigomorla]

Again you can not cross test in the middle of an ongoing test bed, because it will invalidate all your previous tests.
If you start changing to an AMD bed, then you will need to retest all the prior GPU's to hold a decent graph, as it will invalidate all the previous cards you tested on the old bed.

In car terms, its like your testing that Honda, and your testing mods like an intake, cams, ect...
You can't pull a toyota on new tests and run the same comparison numbers for the honda, you would need to retest the air intake, cams ect on the toyota all over again.

Its a massive time consuming pit hole to change beds like this.

 

[Kenmitch]

,

Again you can not cross test in the middle of an ongoing test bed, because it will invalidate all your previous tests.
If you start changing to an AMD bed, then you will need to retest all the prior GPU's to hold a decent graph, as it will invalidate all the previous cards you tested on the old bed.

In car terms, its like your testing that Honda, and your testing mods like an intake, cams, ect...
You can't pull a toyota on new tests and run the same comparison numbers for the honda, you would need to retest the air intake, cams ect on the toyota all over again.

Its a massive time consuming pit hole to change beds like this.

I don't really see your argument being valid.

They switch to the latest and greatest Intel gaming offering each time.

You hate AMD?

New GPU's are going PCI 4.0 and should be tested on a supported platform. I doubt I'm the only curious one.

It's not like they can't still test on the old rig also.

 

[DAPUNISHER]

New GPU's are going PCI 4.0 and should be tested on a supported platform. I doubt I'm the only curious one.

Hardware Unboxed took a poll. They had over 60,000 votes and of those, 83 percent wanted the 3950x. You are far from alone in wanting to see an all PCI-E 4.0 test bed.

 

[biostud]

Techspot.com did this piece

AMD or Intel for GeForce RTX 3080 Benchmarking: Is PCIe 4.0 a Factor?

Nvidia is set to introduce faster next-gen Ampere GPUs next month. That's exciting, but we've got a tough choice to make. The new graphics cards will support...

www.techspot.com

 

[Markfw]

I said this in another thread, but it bears repeating

WOW, between RTX 3000 cards and AMD 4000 cpus, the gaming world could change drasticly in the next 6 months or less (once both are available)

 

[DAPUNISHER]

Techspot.com did this piece

AMD or Intel for GeForce RTX 3080 Benchmarking: Is PCIe 4.0 a Factor?

Nvidia is set to introduce faster next-gen Ampere GPUs next month. That's exciting, but we've got a tough choice to make. The new graphics cards will support...

www.techspot.com

hammer on box aka hardware unboxed is their youtube channel.

 

[aigomorla]

,


I don't really see your argument being valid.

They switch to the latest and greatest Intel gaming offering each time.

You hate AMD?

New GPU's are going PCI 4.0 and should be tested on a supported platform. I doubt I'm the only curious one.

It's not like they can't still test on the old rig also.

They retest all the cards on the new bed.
Why would any reviewer want to torcher himself and have to repeat the test all over again on an AMD bed when it will show about the same thing?

The only valid point i see in having 2 beds, is if the user wanted to test PCI-E 3.0 vs 4.0
Then yeah, the tester would need to test all generations of video card on the 3.0 using the SAME CPU, and then test it again on the 4.0.
This would lead to a nightmare if s/he had to do it a third time on an Intel platform.

And why are you asking if i hate AMD? what does this have anything to do with AMD when im saying the tester would be rotting his time repeating all the tests over and over again.

Maybe i should ask you the same question in reverse what is wrong with the current tests? are you salty they are not done on AMD?

I can pretty much tell you if a gpu shows 10% performance boost over GPU 2, then on BOTH platforms AMD / Intel it will probably be the same if you keep the same bed, unless again, your changing the bed to pci-e 3.0 vs pci-e 4.0.

Maybe your not aware but I was an independant tester for a lot of hardware like watercooling and older Intel chips so i am fully aware of of methodologies and where people complain that your testing is flawed, and once people start pointing it out to you, and picks up motion, it will ruin your reputation as a tester.

This is why Hardware Unboxed also always sticks to ONE bed and they want feedback before they waste time reduing the GPU data on the link Biostud gave.
Its just a waste of time unless your doing again, that single gpu comparison on pci-e 3.0 vs pci-e 4.0 trying to keep to as close as the same core hardware as possible.... meaning same ram same cpu only different board, or better yet if the board has an option to go pci-e 3.0 to 4.0 enabled.

 

[Kenmitch]

Maybe i should ask you the same question in reverse what is wrong with the current tests? are you salty they are not done on AMD?

No.

How else can you find out if PCI 4.0 cards benefit from PCI 4.0? There is no need to go back generations on lower performing gpus anyways.

If you don't agree that's fine.

 

[biostud]

Since the 1440p and 4k results are so close on both systems it doesn't matter that much. But once zen3 is released they have to redo all the tests anyway