• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***

Soldato
Joined
27 Feb 2015
Posts
12,621
A lot of reviewers tend to benchmark CPU's in games looking at a wall or the sky, where they get the most FPS because somehow they equate that to CPU performance.

In fact it's not, where there's nothing going on in the scene the CPU isn't actually being used, at least not properly, it's how they come up with really idiotic conclusions like this higher clocked 4 core is better than this 8 core, yet if they were to look away from the wall or the sky that 4 core craters as all of its poultry 4 thread's are maxed out while the 6 shoots into the lead.

Digital Foundry did some great 1600 vs 7600K testing where they demonstrated this, with nothing to see in the scene the 7600K was pushing 240 FPS vs 210 on the 1600, literally looking at the sky, looking at the complex vista at ground level the 7600K tanked to 60 FPS while the 1600 was at 110 FPS, nearly twice as fast. so which CPU is actually better?

About 90% of reviewers said the 7600K was better, that tells you everything you need to know about those reviewers, there all idiots.

The two best ways to bench cpu impact on games is pick games that are coded badly that choke because of clear ram/cpu bottlenecks, that would probably highlight the best impact of clock speeds and single thread performance, then games like assasins creed which is a weird game at the other extreme where it seems to soak up whatever cores and threads are thrown at it with constant 90-100% cpu utilisation on all cores. So you testing both scenarios at the extreme. I am also not a fan of testing with vsync disabled, the vast majority of people game at capped 30 or 60fps. I am ok with an additional uncapped test, but it shouldnt be the only test.

I am not a fan of dropping resolution to 720p to try and keep the gpu idle, that kind of benching is not realistic, part of the tests also include the chipset and the speed of data transfer between gpu and cpu, so its important to be testing real world scenarios. Although a few people may play at 720p windowed tho as there is advantages to gaming windowed.

The review industry simply remains flawed all benching the same subset of games as each other, and having flaws in their testing (often deliberate flaws to promote the product they testing as after all they essentially a marketer with their free samples) but it is what it is.
 
Soldato
Joined
1 Mar 2010
Posts
14,373
Location
5 degrees starboard
The internet is a wonderful thing, the CPUs will be available all over the place for MRSP, go to a comparison site and use that to find the cheapest best stock, don't be loyal to companies that only exist to make a profit.

Actually that is what companies do, make a profit. Otherwise everyone is unemployed as it goes bust. I know what you are saying though. Reputations stick and a good rep for service quality and straight dealing is worth its weight in gold.
 
Soldato
Joined
27 Feb 2015
Posts
12,621
Motherboards can't be compared across platforms (naming schemes just follow price points and marketing). Ram should definitely be normalised though.

I agree and disagree.

I understand normalisation is to try and show raw cpu performance only.

However the platforms are a package, the IMC, the chipset, etc. is all a package working together.

So e.g. z370 vs x370 in terms of memory speed, the x370 package supported a higher stock ram speed, but the z370 typically could handle faster ram speeds. In terms of latency timings, these were often disregarded and need higher attention on testing as well given memory latency is more important than memory bandwidth for gaming.

So the best scenario for when reviewing a platform in my opinion is one test in supported configurations which should be advantage to AMD given they support higher ram speeds over intel, then another test in some kind of overclocked configuration, and a third test can perhaps be the one thats normalised.

Ultimately these reviews need to be much more thorough with much more things been tested, but they all only care about rushing out the door for embargo date. A common theme in the reviews is "we ran out of time". There is no time limit, do it properly and publish when its done.

Then after the basic product review is done, a follow up should be done analysing things like the affect of memory timings on games. The best information I am getting now is from hobbyists on reddit, steam etc. rather than professional reviewers, the latter need to up their game. They also need to start testing more mainstream/budget components, even if it means they have to buy the stuff. Their lack of knowledge on lower end is astounding e.g. when they think b450 cannot overclock (they assume is locked like budget intel chipsets) and they all only come with 2 dimm slots.

The issue the hobbyists have is they treated with less respect and not treated as credible, e.g. when I posted my feedback on ram timings I was treated as if my results were not real.
 
Soldato
Joined
19 Feb 2011
Posts
5,849
Actually that is what companies do, make a profit. Otherwise everyone is unemployed as it goes bust. I know what you are saying though. Reputations stick and a good rep for service quality and straight dealing is worth its weight in gold.

I agree while OCUK RMA is renowned, other sites like rainforest are equally as easy to RMA with now.
 
Soldato
Joined
27 Feb 2015
Posts
12,621
Sad to see people cant comprehend why tests ware done on 720p.... But I must Admit they could have added 1440p also. The 720p is showing Raw power thats it.

its because they chasing hits, to only test one situation is lazyness. 6 games is an extremely small sample, and to boot only one resolution.

Some content creators have e.g. said by getting out first you can get like 10x the viewers, and if you publish one week behind everyone else its a massive drop off. So they prefer to rush stuff out, rather than be more detailed and get less eyeballs. Its a sad state of affairs.
 
Soldato
Joined
27 Feb 2015
Posts
12,621
I agree while OCUK RMA is renowned, other sites like rainforest are equally as easy to RMA with now.

rainforest is too easy I wont be buying a cpu or motherboard from them again, because boards are not factory sealed, there is no way to know if your supposedly new board is a previous return and return rates for motherboards are high. Cpu's I expect are not that low either as people want to win the silicon lottery so if they get a loser there will be a temptation to return, some people on reddit admitted they routinely buy 3-5 cpus at once, test all of them and send the worst back to retailer.

When I returned my PSU to OCUK, there was a delay for RMA processing, meaning they actually test whats been returned. Although I dont know if they resell opened stuff as new. My return was still shrink wrapped.
 
Associate
Joined
6 Jul 2019
Posts
15
This sub forum attracts CPU fanatics and we’re on the cusp of a highly anticipated product release. There is no meaningful reason, it’s just people passing the time until reviews hit. At which point it’ll most likely be completely forgotten about.
Yeah I got it... I just hate arrogance, and this forum seems to be quite more argument-driven than others. Anyway, happy waiting everyone! .D
 
Associate
Joined
6 Jul 2019
Posts
15
The sweet spot appears to be 3600-3733MHz. You can run faster memory but you may get lower performance. The results are unknown on say the B450 boards but check out MSI B450 Tomahawk or Carbon Pro. Probably as good as you'll do for £100.
Lower performance is relative. You'd get higher latency - but also higher bandwidth. Anyone who needs the most bandwidth for whatever reason, now can go for 4400+ and be happy (happier) :)
 
Soldato
Joined
13 Jun 2009
Posts
6,847
rainforest is too easy I wont be buying a cpu or motherboard from them again, because boards are not factory sealed, there is no way to know if your supposedly new board is a previous return and return rates for motherboards are high. Cpu's I expect are not that low either as people want to win the silicon lottery so if they get a loser there will be a temptation to return, some people on reddit admitted they routinely buy 3-5 cpus at once, test all of them and send the worst back to retailer.

When I returned my PSU to OCUK, there was a delay for RMA processing, meaning they actually test whats been returned. Although I dont know if they resell opened stuff as new. My return was still shrink wrapped.
My motherboard arrived pre-opened but then how can I expect the retailer to update the BIOS otherwise? Agreed about buying motherboards from the rainforest though. One way to avoid getting a returned CPU is to buy on day one. :D
 

HRL

HRL

Soldato
Joined
22 Nov 2005
Posts
3,028
Location
Devon
According to pcgh the stock 3900x pulls 230w so the vrm is justified. Remember vrm is not built for everyday use exclusively it’s built for overclockers. The 3950x will probably be 260w+ And this is before overclocking

From the wall, right?
 
Back
Top Bottom