B580 tanks performance with low end CPUs

[eidairaman1]

Not sure to be honest. It's being worked on. Solaris17 mentioned it on the first page of this thread and that was in January.

Those with core i 10/11 can only hope so, also what about the Ryzens of that Era too

 

[DAPUNISHER]

I have a Gigabyte B450 AORUS ELITE - Ryzen 3700X - 4x8GB CL16 3200 combo not in use. I'll get it setup with latest 11 pro and the Odyssey B580 and see how it goes.

 

[BlueBee]

I have a Gigabyte B450 AORUS ELITE - Ryzen 3700X - 4x8GB CL16 3200 combo not in use. I'll get it setup with latest 11 pro and the Odyssey B580 and see how it goes.

Good luck bud, very similar to one of my systems, I was going to get the Intel B580 for it but grabbed a 16GB Pulse 9060xt for £320 quid on launch day instead. The B580`s seem to have really dropped in price recently, at least at Overclockers where they`re down at £240

 

[chrcoluk]

I think I remember some time ago some people were requesting that when hardware is tested, that the entire build should match the expected component mix, but most reviewers rejected the idea as too time consuming and they they felt they should always use the best hardware possible in rest of system as they feel its wrong to gimp what they reviewing.

Well these revelations kind of show the merits of using an expected build with the product that is being reviewed.

 

[DemonicRyzen666]

I think I remember some time ago some people were requesting that when hardware is tested, that the entire build should match the expected component mix, but most reviewers rejected the idea as too time consuming and they they felt they should always use the best hardware possible in rest of system as they feel its wrong to gimp what they reviewing.

Well these revelations kind of show the merits of using an expected build with the product that is being reviewed.

There should be an something like the average steam build consisting of a some x thread cpu & the most used graphics card as the baseline for performance. the "100%. point."

 

[iameatingjam]

There should be an something like the average steam build consisting of a some x thread cpu & the most used graphics card as the baseline for performance. the "100%. point."

I think both have value honestly, I mean if you want to really get the idea of what a gpu can do, not just what it can do in its most likely configuration, you can't have other components holding it back, it kinda defeats the purpose, because what if you decide to upgrade your whole system except the gpu, right?

At the same time, if you ignore the fact that lower end GPUs are going to be run on lower end systems you end up missing things like b580s CPU overhead issue or the 8 lane 8gb cards (like the 4060/ti/5060/ti and 7600) weakness on pcie 3/ddr4 systems.

 

[DAPUNISHER]

At the same time, if you ignore the fact that lower end GPUs are going to be run on lower end systems you end up missing things like b580s CPU overhead issue

Having the highest CPU overhead was something those of us using Alchemist already knew about and documented. I think it was us ARC owners asking if the issue was fixed on Gandalf/Battle Mage that led to Hardware Canucks testing it.

Good luck bud, very similar to one of my systems, I was going to get the Intel B580 for it but grabbed a 16GB Pulse 9060xt for £320 quid on launch day instead. The B580`s seem to have really dropped in price recently, at least at Overclockers where they`re down at £240

I would have made the same call. The Pulse is a great pick at that price.

I am going to start setting up the test bed now, and download Spiderman Miles Morales to see how bad it bottlenecks. It ran smooth with a 5600X3D+B550i. Though it hit as high as 98% CPU usage for a second here and there swinging fast near the street on Broadway with crowds and traffic maxed.

 

[mechtech]

Poor W1zz will have to test with the oldest 4 core supported by win11 and down clocked at 2GHz now to see what’s the best upgrade for potato PCs.

 

[lexluthermiester]

as they feel its wrong to gimp what they reviewing.

This. Absolutely correct.

 

[oxrufiioxo]

Poor W1zz will have to test with the oldest 4 core supported by win11 and down clocked at 2GHz now to see what’s the best upgrade for potato PCs.

Just a disclaimer that the hardware may not perform the same on older hardware is fine especially in scenarios where gpu makers really skimp on the pcie lanes.

There's also testing by other outlets already negating how necessary it is for tpu to do the same testing.

At the same time testing different pcie bandwidth with the fastest available gaming cpu and latest memory isn't representative of people actually using platforms locked to older standards making it mostly useless testing.

 

[GodisanAtheist]

Poor W1zz will have to test with the oldest 4 core supported by win11 and down clocked at 2GHz now to see what’s the best upgrade for potato PCs.

- Feels like a fun article actually.

It's gotta get boring running cards through their paces on the 9800X3d beast, let's see where the lower bounds of playability are!

 

[DAPUNISHER]

I used 1080p high first in Miles Morales, which is the settings Aussie Steve uses in Spiderman Remastered. I was seeing as low as 72% GPU usage and it rarely got past mid 80s. With RT reflections on GPU went down to 50-65%. XeSS does not help because it is all CPU bottleneck. PCIe 3.0 may be playing a part too. Not like it isn't playable as FPS was in the 70s-110s, but the B580 is capable of much more than that.

To clarify: Those fps were for 1080 high.

I made things harder yet and recorded almost 15 minutes using Xbox game bar. Then turned on RT reflections, then added XeSS quality at 1440.

Swinging fast near the streets in the part of the game where it is snowing heavily it could go into the 40s fps. I will check the video for the exact time, but it froze for a good 10 seconds while I could see assets loading in as the 3700X struggled to keep up.

For CPU heavy games + RT + recording, it is on the struggle bus for certain. I'll nuke the ARC drivers and test with the 3060 12GB under the same conditions but with ShadowPlay, to see if there is a significant difference in GPU utilization. I could use game bar again, but apples to apples isn't the goal here. Because if recording on ARC with 3rd party is more CPU demanding than the competition with their in house solutions, that's arguably another reason to go with either green or red for older builds.

 

[chrcoluk]

This. Absolutely correct.

I think both views are correct.

There is a merit to showcasing the product to run in the best way possible, although thats more what I would expect a marketing department to do. But there is also merit showcasing the product in a realistic operational environment.

This is why I am not a fan of things like 100% fan speeds on CPU reviews, disabling temperature cap on reviews, unstable overclock on reviews, and finally poor hardware configuration on reviews.

The reviewers job isnt to showcase the product, but try to guide users to what is a good product for their systems. When they only doing low to mid end GPUs on 9950x CPU's and only doing low to mid end CPU reviews with 5900 GPU's. They not serving the people with low to mid hardware.

Do we need 20 reviewers reviewing everything the same way similar builds, or are we better served with a variety?

 

[lexluthermiester]

I think both views are correct.

This is not a subject where one can fence-sit. You have to choose which side you're on.

One can't properly test the limits of a piece of hardware if the adjoining hardware and/or platform it's being tested on is itself the limiting factor. Sure, there's an argument to be made for testing on more "consumer realistic" hardware, but then the top tier sectors are left out of the equation. That's just not wise nor smart.

It's always best practice to test a given piece of tech on the best hardware available at the time of testing. This is common sense and has been common practice for decades. It does not need to change and should not change.

 

[chrcoluk]

This is not a subject where one can fence-sit. You have to choose which side you're on.

One can't properly test the limits of a piece of hardware if the adjoining hardware and/or platform it's being tested on is itself the limiting factor. Sure, there's an argument to be made for testing on more "consumer realistic" hardware, but then the top tier sectors are left out of the equation. That's just not wise nor smart.

It's always best practice to test a given piece of tech on the best hardware available at the time of testing. This is common sense and has been common practice for decades. It does not need to change and should not change.

If a hardware only works well with high end hardware because it has higher overheads yet its designed for the low to mid range market, then by only using high end hardware to test it means you are hiding the flaw with that hardware. You are taking the view that a reviewers job is to make the product they are reviewing look as good as possible. Which is fine, everyone has their own opinion, but just because this is way things have been tested for eons, it doesnt mean its the right way forward.

Thanks for ignoring the question, why we should have a dozen reviewers doing the same thing instead of variety. You dont lose anything from that.

- Feels like a fun article actually.

It's gotta get boring running cards through their paces on the 9800X3d beast, let's see where the lower bounds of playability are!

I would reward him for doing it by donating, I find those kind of reviews far more valuable than just sticking to the status quo. I like things that are out of the box.

 

[lexluthermiester]

If a hardware only works well with high end hardware because it has higher overheads yet its designed for the low to mid range market

New flash: High end hardware ALWAYS performs worse on lower end platforms. That's how it's ALWAYS worked. This is not earth shattering information. It's common sense and is commonly known.

An RTX 5090 is NOT going to perform as well on a Core i3 8100T as it would on a Core i9 14900K. Likewise a B580 is the exact same dynamic, even though it's not in the same tier as a 5090, the effects are the same. Test a B580 in a 14900K based system at it will perform at it's peak. Test it in a Core i3 8100T and it will NOT perform at it's peak.

This is not a flaw, nor a glitch, nor a bug. It's the law of IC physics. Plain and simple.

Thanks for ignoring the question

I didn't ignore anything. I saw right through it and argued completely passed it. Why? Because it's a nonsensical, illogical and meritless perspective.

The whole premise of the argument this thread invokes is beyond credibility.

 

[TumbleGeorge]

Reviewers who point out the flaws of a product too blatantly usually no longer receive free test samples and, if they want to continue reviewing, have to purchase the hardware from the store with their own funds.

 

[chrcoluk]

I didn't ignore anything. I saw right through it and argued completely passed it. Why? Because it's a nonsensical, illogical and meritless perspective.

The whole premise of the argument this thread invokes is beyond credibility.

So a product that performs worse against the competition when used with lower end hardware is something that shouldnt be factored into a review, the idea is nonsensical?

Reviewers who point out the flaws of a product too blatantly usually no longer receive free test samples and, if they want to continue reviewing, have to purchase the hardware from the store with their own funds.

This I feel is a factor in the idea, that the product must be shown in the best way possible.

 

[lexluthermiester]

So a product that performs worse against the competition when used with lower end hardware is something that shouldnt be factored into a review, the idea is nonsensical?

How are you failing to understand this dynamic?

 

[iameatingjam]

So a product that performs worse against the competition when used with lower end hardware is something that shouldnt be factored into a review, the idea is nonsensical?

I agree. Its only fair to show the card in its best possible configuration, and in configurations where it is likely to be used and where it where it might have additional issues. We can have both.

And we kinda do. I can think of some reviewers that use more modest hardware setups, or that point out the weaknesses in situation x and y..... Though to be fair, they are less plentiful than the same best case scenario reviews, of which there is a lot.

Large support base and financial security can give you more freedom to do this, unfortunately..... money makes the world go around.

 

[chrcoluk]

How are you failing to understand this dynamic?

I am struggling to make sense of your opinion. I dont think I ever will, so will leave it at a agree to disagree.

 

[iameatingjam]

An RTX 5090 is NOT going to perform as well on a Core i3 8100T as it would on a Core i9 14900K. Likewise a B580 is the exact same dynamic

Right, I think the thing is there's a range, or spectrum, for how much the 8100T would be holding back whatever video card its using. Like if the card only has 8 lanes and 6/8gb vram, thats going to have a bigger performance deficit on a lower end cpus, during periods where the buffer is exceeded. Similar situation with the b580. It seems to hold back performance more, relative to other cards, while running on lower end cpus.

I don't know how many people are upgrading 8100Ts with news GPUs, but lets put out some more common examples like say a 3600/x or 10600/k. Both these could be gaming cpus, depending on what games are played, blah blah all the other variables, and what gpu is chosen. Couldn't it be the case with a cpu like one of these, that GPU x is ahead, while in a modern system gpu y is ahead in certain gaming conditions? Perhaps its by a significant margin. I would want to know that, if I had one of those CPUs.

You acknowledge there's variability there, right?

This is not a subject where one can fence-sit. You have to choose which side you're on.

Why? Can I not appreciate the value of a review that looks at the best case scenario and also a review that looks at components that may show additional weakness ontop of expected cpu limits? I mean I can understand why there would legititimate reasoning for both kinds of analysis.

 

[MihaiPopaTech]

An RTX 5090 is NOT going to perform as well on a Core i3 8100T as it would on a Core i9 14900K. Likewise a B580 is the exact same dynamic, even though it's not in the same tier as a 5090, the effects are the same. Test a B580 in a 14900K based system at it will perform at it's peak. Test it in a Core i3 8100T and it will NOT perform at it's peak.

Another example: a RTX 4090 or 5090 will be not as good on a Pentium 4 as on a Ryzen 7 9800X3D or i9 14900K.

 

[lexluthermiester]

I am struggling to make sense of your opinion.

That's not my problem..

I agree. Its only fair to show the card in its best possible configuration, and in configurations where it is likely to be used and where it where it might have additional issues. We can have both.

Tell THAT to the reviewers who already spend mountains of time creating reviews. See what they tell you. They simply don't have the time to spend reviewing a single piece of hardware on multiple system configs. I've written a few reviews myself and I know what I would say to the above suggestion: Get stuffed! You want to review it on hardware like what you have? Buy a card yourself, test it yourself and make your own conclusions.

Right, I think the thing is there's a range, or spectrum

Maybe, but it's not needed. In the case of the B580, the reviewed performance metrics show what it'll do in optimal situations. Lower end systems with lower end CPU's will experience lower performance. Tada.. This is NOT rocket science. It's common sense.

Why? Can I not appreciate the value of a review that looks at the best case scenario and also a review that looks at components that may show additional weakness ontop of expected cpu limits?

So what you're saying is, you are willing to accept a review that fails to show complete performance potential? That's what reviewing hardware on lesser platforms is. If you a review a GPU on a system that has a CPU that is completely bottlenecking said GPU, you will not see the GPU's fullest potential.

Another example would be reviewing a PCIe 5.0 4X NVMe drive by testing it on an old PCIe 3.0 2X connection. The drive would be artificially bottlenecked by the system it's plugged into, which would invalidate the results and conclusion.

That is why testing new hardware on older/slower platforms is at minimum, narrow flawed thinking. At worst, beyond foolish and, dare I say, stupid. No reviewer that has any modicum of competence will create a review for a piece of hardware while testing on a platform ill-equipped to handle it. It's just not done and for excellent reasons.

@ Everyone
So folks, can we start using our brains for something other than a seat cushion and STOP listening to daft twaddle coming from the likes of Steven Walton? See sense people... Or not, it's up to you..

Another example: a RTX 4090 or 5090 will be not as good on a Pentium 4 as on a Ryzen 7 9800X3D or i9 14900K.

Exactly! No one is going to have to struggle to understand that an RTX card(any of them) will be severely bottlenecked by such a CPU. It's not surprising and it doesn't take a ton of testing to explain.

 

[iameatingjam]

So what you're saying is, you are willing to accept a review that fails to show complete performance potential?

Well the devils in the details. If its trying to convey the message like "this is all it can do" while running it on a system that is bottlenecking it, then no I wouldn't. But if its put into proper context, with the purpose of the analysis made clear, then yes, I would.

I just think there's room, and demand, for both.

I understand why there's best case scenario reviews, I mean, it is a gpu review, not a cpu/ram/mobo/whatever review.

And I understand why there's explorations into how the components may work in older systems, as we seem to agree, that they don't always cleanly line up with the best case scenario's results.


EDIT:

Tell THAT to the reviewers who already spend mountains of time creating reviews. See what they tell you. They simply don't have the time to spend reviewing a single piece of hardware on multiple system configs. I've written a few reviews myself and I know what I would say to the above suggestion: Get stuffed! You want to review it on hardware like what you have? Buy a card yourself, test it yourself and make your own conclusions.

Geeze man I never even said I expect the same people to be doing both things. I'm not attacking people who do best case scenario reviews or saying they are doing it wrong. I'm just saying there's room for other kinds of reviews. And no its not even about what I have. I already told you about two examples in particular that concern me and neither apply to me.