Tech Gaming Hardware discussion (& Hardware Sales) thread

[My Spot]

10400 for $150 is a damn good deal. Gaming performance is usually right inbetween the 3600X and 5600X. I love the microcenter near my house, if you refresh the website every morning you can get another 10 to 15% taken off just by buying the same product but open box

Yeah and they have the deal if you buy a CPU and mobo together. It’s only $20 now but it was at least $30 off when I got my i5 8400 and B365M.

 

[KaNesDeath]

10400 for $150 is a damn good deal. Gaming performance is usually right inbetween the 3600X and 5600X.

Yeah and they have the deal if you buy a CPU and mobo together. It’s only $20 now but it was at least $30 off when I got my i5 8400 and B365M.

Pre-Zen3 CPU's are about to see massive cuts in price once Zen3 demand is met; six months or so.

 

[560ti]

Pre-Zen3 CPU's are about to see massive cuts in price once Zen3 demand is met; six months or so.

Since the price gap between the 5600x and 3600/3600x is already so drastic at microcenter I don't think there will be a price cut, there easily going to sell at there current price. Even at regular retail prices the Ryzen 3 prices jump this gen and the cheaper Zen 2 has lowered in price

By the time Zen 3 demand is met + the non X version comes out all of the Zen 2 will probably be gone so no point in signficant price cuts.

 

[Madmick]

https://www.userbenchmark.com/UserRun/35177456

My userbenchmark isn't too bad, GPU and Ram need upgrading.

I'm not sure I would even both with UB, anymore. What was for a brief, shining moment the simplest, most accessible, most useful reference in the history of hardware seems to have fallen completely. To be clear, their problem is not (and never was) their model as a mass user aggregator-- like Passmark-- contrary to what Steve Burke has opined in his unyielding anal-retentiveness; a quality I otherwise almost entirely admire in Steve. No, that only makes them more useful.

The problem is they flat out sold out to Intel. Originally, the aggressive weighting of the "Effective Score" is precisely what made it so great. It did a miraculous job of simplifying the complex nature of CPU performance for casual buyers who don't understand this, and don't care to understand it.

But the revision to their score weighting earlier this year was awful, there's not nearly enough weight for the overall multicore score, 2% is a fucking joke, and now somehow they've managed to go full shill because manipulating the weighting methodology is no longer sufficient (not even possible) to spare Intel's humiliation. You never go full shill.

Final nail in the coffin: Bar-raising AMD Ryzen 9 5950X somehow lags behind four Intel parts including the Core i9-10900K in average bench on UserBenchmark despite higher 1-core and 4-core scores

Earlier this year there was quite a fuss made about UserBenchmark and its questionable scoring method in regard to results produced by Intel Comet Lake chips. The synthetic benchmark also found itself banned from numerous important tech-related subreddits but remains one of the most widely visited processor testing sites. However, the site has frequently been accused of operating with a heavy Intel bias, and current results and ranking for the 16-core, 32-thread Zen 3 AMD Ryzen 9 5950X won’t make those accusations go away.

Looking solely at the results of the Ryzen 9 5950X, which has been described as offering “a new level of consumer grade performance across the board” by the respected review site AnandTech, and it is clear to see the Vermeer CPU has benchmarked well on UserBenchmark. Unsurprisingly, the 16-core part amasses a huge score in the 64-core test (results in screenshots below) but as is well known UserBenchmark doesn’t place too much emphasis, or score, on multi-core benchmarks (“16 cores are only suitable for professional use cases”). However, AMD has been working hard on improving single-core performance with its Zen 3 architecture, and this is demonstrated with the Ryzen 9 5950X’s 1-core and 2-core scores. All of these impressive results deliver the Vermeer chip a decent average benchmark score of 100%.

Apparently, this great performance from the Ryzen 9 5950X (based on a comparatively low sample set of 134 units at the time of writing) is not enough to overtake the Intel Core i9-10900K…or the i9-10850K, i9-10900KF, or the i9-9900KS. If there was some score weighting based on price then it would be understandable as it is possible to pick up Intel’s 10-core i9-10900K for around US$540 at the moment whereas the Vermeer rival is pricey at US$799. But the UserBenchmark weighting is based on performance and core scores (“40% single-core, 58% quad-core, and 2% multi-core” according to Tom’s Hardware). So how does the Ryzen 9 5950X produce average scores of 188/901/2,943 points and still fall behind the Comet Lake chip on 183/878/1,990 points? Even discounting the huge 64-core result difference leaves the AMD CPU ahead – but the average benchmark remains 102% vs. 100% in Intel’s favor.

To UserBenchmark, from the bottom of my heart:

 

[Slobodan]

I'm not sure I would even both with UB, anymore. What was for a brief, shining moment the simplest, most accessible, most useful reference in the history of hardware seems to have fallen completely. To be clear, their problem is not (and never was) their model as a mass user aggregator-- like Passmark-- contrary to what Steve Burke has opined in his unyielding anal-retentiveness; a quality I otherwise almost entirely admire in Steve. No, that only makes them more useful.

The problem is they flat out sold out to Intel. Originally, the aggressive weighting of the "Effective Score" is precisely what made it so great. It did a miraculous job of simplifying the complex nature of CPU performance for casual buyers who don't understand this, and don't care to understand it.

But the revision to their score weighting earlier this year was awful, there's not nearly enough weight for the overall multicore score, 2% is a fucking joke, and now somehow they've managed to go full shill because manipulating the weighting methodology is no longer sufficient (not even possible) to spare Intel's humiliation. You never go full shill.

Final nail in the coffin: Bar-raising AMD Ryzen 9 5950X somehow lags behind four Intel parts including the Core i9-10900K in average bench on UserBenchmark despite higher 1-core and 4-core scores


To UserBenchmark, from the bottom of my heart:

What's funny is that ever since that fiasco earlier this year, they were banned from the Intel subbreddit as well the Nvidia one.

 

[Woldog]

I'm not sure I would even both with UB, anymore. What was for a brief, shining moment the simplest, most accessible, most useful reference in the history of hardware seems to have fallen completely. To be clear, their problem is not (and never was) their model as a mass user aggregator-- like Passmark-- contrary to what Steve Burke has opined in his unyielding anal-retentiveness; a quality I otherwise almost entirely admire in Steve. No, that only makes them more useful.

The problem is they flat out sold out to Intel. Originally, the aggressive weighting of the "Effective Score" is precisely what made it so great. It did a miraculous job of simplifying the complex nature of CPU performance for casual buyers who don't understand this, and don't care to understand it.

But the revision to their score weighting earlier this year was awful, there's not nearly enough weight for the overall multicore score, 2% is a fucking joke, and now somehow they've managed to go full shill because manipulating the weighting methodology is no longer sufficient (not even possible) to spare Intel's humiliation. You never go full shill.

Final nail in the coffin: Bar-raising AMD Ryzen 9 5950X somehow lags behind four Intel parts including the Core i9-10900K in average bench on UserBenchmark despite higher 1-core and 4-core scores


To UserBenchmark, from the bottom of my heart:

Wow I hadn't heard any of that before. I always thought they were pretty fair considering. I did think my score was a little low considering I got nearly the same score with my old 7700k build when I tested it a month back....

 

[jefferz]

Neweggs Black Friday deals have started.

@Madmick That Corsair 4000D is on sale for $75 on Newegg

 

[Woldog]

Can @Madmick or @jefferz recommend 3600mhz ram or faster, I know people say ram is ram but I don't really know much about it and timings when it has to do with the Ryzen series CPU.

 

[Madmick]

Can @Madmick or @jefferz recommend 3600mhz ram or faster, I know people say ram is ram but I don't really know much about it and timings when it has to do with the Ryzen series CPU.

It doesn't look like they've updated your QVL for the new Zen 3 processors, unfortunately:
https://rog.asus.com/motherboards/rog-strix/rog-strix-x570-e-gaming-model/helpdesk_qvl/

Not sure if you're after 16GB or 32GB. The 3600MHz CL16 kits are the best bang-for-your-buck right now. These are the same kits in the different sizes (2x8GB & 2x16GB). Based on timings the Crucial kit is theoretically superior.

16GB
16-18-18-38 (#1 choice)
https://au.pcpartpicker.com/product...6-gb-2-x-8-gb-ddr4-3600-memory-bl2k8g36c16u4b
or
16-19-19-39
https://au.pcpartpicker.com/product...-2-x-8-gb-ddr4-3600-memory-f4-3600c16d-16gvkc

32GB
16-18-18-38
https://au.pcpartpicker.com/product...gb-2-x-16-gb-ddr4-3600-memory-bl2k16g36c16u4b
16-19-19-39
https://au.pcpartpicker.com/product...2-x-16-gb-ddr4-3600-memory-f4-3600c16d-32gvkc


The b-die kits command a huge premium. I know I've seen some early press suggesting 4000MHz will be ideal for the new Zen 3 chips, but I've also seen press indicating the reviewer couldn't even get RAM sticks rated to that speed stable at that frequency. b-die might be your best bet for extreme overclocks like this, and they're also the only kits under $400 AUD (per 32GB), but I just don't think it's worth it.
https://au.pcpartpicker.com/product...T=140140140340,160160160360&F=5625000,9829280

 

[Woldog]

It doesn't look like they've updated your QVL for the new Zen 3 processors, unfortunately:
https://rog.asus.com/motherboards/rog-strix/rog-strix-x570-e-gaming-model/helpdesk_qvl/

Not sure if you're after 16GB or 32GB. The 3600MHz CL16 kits are the best bang-for-your-buck right now. These are the same kits in the different sizes (2x8GB & 2x16GB). Based on timings the Crucial kit is theoretically superior.

16GB
16-18-18-38 (#1 choice)
https://au.pcpartpicker.com/product...6-gb-2-x-8-gb-ddr4-3600-memory-bl2k8g36c16u4b
or
16-19-19-39
https://au.pcpartpicker.com/product...-2-x-8-gb-ddr4-3600-memory-f4-3600c16d-16gvkc

32GB
16-18-18-38
https://au.pcpartpicker.com/product...gb-2-x-16-gb-ddr4-3600-memory-bl2k16g36c16u4b
16-19-19-39
https://au.pcpartpicker.com/product...2-x-16-gb-ddr4-3600-memory-f4-3600c16d-32gvkc


The b-die kits command a huge premium. I know I've seen some early press suggesting 4000MHz will be ideal for the new Zen 3 chips, but I've also seen press indicating the reviewer couldn't even get RAM sticks rated to that speed stable at that frequency. b-die might be your best bet for extreme overclocks like this, and they're also the only kits under $400 AUD (per 32GB), but I just don't think it's worth it.
https://au.pcpartpicker.com/product...T=140140140340,160160160360&F=5625000,9829280

I usually run 32gb. But I'm not in a hurry to do it, so I guess the QVL might be updated by the time I get around to it. Right now the GPU is the next big upgrade it's time to get around to.

I don't think I've even seen 4000mhz kits for sale here yet, unless they were GEIL brand.

Edit: actually looking there are quite a few 4000mhz kits around at 16gb they're quite cheap. there doesn't seem to be much 32gb sets around though.

 

[Madmick]

I usually run 32gb. But I'm not in a hurry to do it, so I guess the QVL might be updated by the time I get around to it. Right now the GPU is the next big upgrade it's time to get around to.

I don't think I've even seen 4000mhz kits for sale here yet, unless they were GEIL brand.

Edit: actually looking there are quite a few 4000mhz kits around at 16gb they're quite cheap. there doesn't seem to be much 32gb sets around though.

The markup isn't insane, but I'm always thinking in terms of opportunity cost, that's all.

If 2x16GB:

• Not b-die CL16 3600MHz ($240 - $250)
• b-die CL14 3200MHz ($315)
• b-die CL16 3600MHz ($380)
• Not b-die CL18 4000MHz ($420)
• b-die CL19 4000MHz ($435)
• b-die CL17 4000MHz ($470)

Don't overlook that b-die RAM is specifically craved because it's famous for overclocking stably well beyond its XMP/AMP ratings.

*Edit*
Oop, saw your edit. Yeah, if you loaded up 4x8GB, this Patriot 4400MHz kit is b-die, and costs barely any more than its 4000MHz brother. Your motherboard supports up to 5100MHz OC's, so it could handle it. Gotta be your best buy. If you could hit this it would be absolutely nuts:

• b-die CL19 4400MHz ($320 = 2 x $160)

 

[Woldog]

The markup isn't insane, but I'm always thinking in terms of opportunity cost, that's all.

If 2x16GB:

• Not b-die CL16 3600MHz ($240 - $250)
• b-die CL14 3200MHz ($315)
• b-die CL16 3600MHz ($380)
• Not b-die CL18 4000MHz ($420)
• b-die CL19 4000MHz ($435)
• b-die CL17 4000MHz ($470)

Don't overlook that b-die RAM is specifically craved because it's famous for overclocking stably well beyond its XMP/AMP ratings.

When did ram start getting so complicated lmao

 

[Madmick]

When did ram start getting so complicated lmao

Check the edit. I think you have a clear winner. Just go with 4x8GB. Use your RAM slots. It's not like adding 2 new sticks to expand years down the road at these RAM speeds is such an economical proposition the way the market moves. That's a strategy that works better for office comps because they usually incorporate sticks near the generation's frequency baseline.

 

[Swing]

I had a weird issue with user benchmark where if I performed a bios flash after a windows install user benchmark wouldn't complete the GPU test. I didn't have any GPU issues outside of that; just a really weird bug.

 

[PEB]

@jefferz check out this thing a Raspberry Pi form factor x86 pc running windows.

 

[dAfTiE]

It doesn't look like they've updated your QVL for the new Zen 3 processors, unfortunately:
https://rog.asus.com/motherboards/rog-strix/rog-strix-x570-e-gaming-model/helpdesk_qvl/

Not sure if you're after 16GB or 32GB. The 3600MHz CL16 kits are the best bang-for-your-buck right now. These are the same kits in the different sizes (2x8GB & 2x16GB). Based on timings the Crucial kit is theoretically superior.

16GB
16-18-18-38 (#1 choice)
https://au.pcpartpicker.com/product...6-gb-2-x-8-gb-ddr4-3600-memory-bl2k8g36c16u4b
or
16-19-19-39
https://au.pcpartpicker.com/product...-2-x-8-gb-ddr4-3600-memory-f4-3600c16d-16gvkc

32GB
16-18-18-38
https://au.pcpartpicker.com/product...gb-2-x-16-gb-ddr4-3600-memory-bl2k16g36c16u4b
16-19-19-39
https://au.pcpartpicker.com/product...2-x-16-gb-ddr4-3600-memory-f4-3600c16d-32gvkc


The b-die kits command a huge premium. I know I've seen some early press suggesting 4000MHz will be ideal for the new Zen 3 chips, but I've also seen press indicating the reviewer couldn't even get RAM sticks rated to that speed stable at that frequency. b-die might be your best bet for extreme overclocks like this, and they're also the only kits under $400 AUD (per 32GB), but I just don't think it's worth it.
https://au.pcpartpicker.com/product...T=140140140340,160160160360&F=5625000,9829280

You want to run your fclk at 1:1:1 if at all possible, so 4k ram is gonna be rough. There's also less mobo support for those kits, and they're expensive. I'd just go for 3600 with as tight timings as you can find.
Best I could do was a set of ballistix 3600 at cl 18, literally the only 2 dimm 64gb kit over 2400 I could find in this entire gods-forsaken country.
Optimumtech did a video testing memory speeds the other day, it's pretty good.

These are cinebench numbers from that video:

 

[JaviBrewski]

seeing all these new ryzen cpu and im still here rocking my 1700x with an rx580 and 8 gb of ram (soon to be 16gb. wife making me wait for xmas to install it)

SADFACE.JPG

 

[jefferz]

seeing all these new ryzen cpu and im still here rocking my 1700x with an rx580 and 8 gb of ram (soon to be 16gb. wife making me wait for xmas to install it)

SADFACE.JPG

Keep an eye out for deals on the 3000 series, even the 2000 series. You might not see a fps bump, but there's a quality of life improvement with the newer generations. It's kind of hard to explain.
For example you take a 9th gen Intel vs a 1rst gen Ryzen cpu, fps is an exact match, but the Intel chip will feel snappier in things like navigating Windows. Newer generations of Ryzen chips have cured this issue.

 

[JaviBrewski]

Keep an eye out for deals on the 3000 series, even the 2000 series. You might not see a fps bump, but there's a quality of life improvement with the newer generations. It's kind of hard to explain.
For example you take a 9th gen Intel vs a 1rst gen Ryzen cpu, fps is an exact match, but the Intel chip will feel snappier in things like navigating Windows. Newer generations of Ryzen chips have cured this issue.

unfortunately, wife isnt gonna let me upgrade already. took me damn near 10 years just to upgrade my pc now from my old as dirt hp pavillion i bought back in 2012. cant see her letting me drop $200 on a new processor. had to convince her to let me upgrade my ram from 8gb to 16gb (she agreed since i got a comfy chair second hand so that was $200+ saved.

shes already looking at me shifty bc i want to upgrade my monitor and my 2012 naga is showing signs of trouble (RB sometimes isnt working)

 

[Madmick]

You want to run your fclk at 1:1:1 if at all possible, so 4k ram is gonna be rough. There's also less mobo support for those kits, and they're expensive. I'd just go for 3600 with as tight timings as you can find.
Best I could do was a set of ballistix 3600 at cl 18, literally the only 2 dimm 64gb kit over 2400 I could find in this entire gods-forsaken country.
Optimumtech did a video testing memory speeds the other day, it's pretty good.

These are cinebench numbers from that video:

I'll take a look at the video later.

Indeed, and this is why my original recommendation focused on CL16 3600MHz sticks. The 16-16-16-36 b-die G. Skill Trident Z and Neo kits are the ones everyone seems to want these days, but when I reviewed Woldog's observation that smaller sticks were much cheaper at 4000MHz+, I saw the Patriot b-die 19-19-19-39 4400MHz sticks. Since they're both b-die, and the cycle time of the Patriot kit is actually lower (8.63ns vs. 8.89ns), in addition to costing $60 less for 32GB total, I figured that since any XMP/AMP profile is technically an overclock, anyway, he could tweak his Patriot kit manually to reach the same frequency/timings or better, and save money. He'll want to downclock to 4000MHz, anyway.

The intrigue of 4000MHz was raised by this PC Gamer article from last month (albeit based on a allegedly leaked slide from within AMD picked up by WCCFTech):
AMD Ryzen 5000 CPUs may run best with faster DDR4-4000 memory (Oct-16)

A leak slide purported to be from AMD hints at faster memory support with AMD Ryzen 5000 processors. It details a new standard for the so-called sweet spot of memory performance, DDR4-4000, which will see AMD Ryzen gaming PCs thirst for even faster memory with the coming generation.

AMD Ryzen CPUs tend to be very responsive to faster memory, much more so than Intel's comparable processors, although there's a limit to how much performance you'll gain by creeping up memory speeds before system latency knocks it down a peg or two. Let me explain: Ryzen 3000 processors have a 'sweet spot', when the Infinity Fabric clock (fclk), memory controller clock (uclk), and memory clock (mclk) are at a 1:1:1 ratio. The best speed to ensure this on existing Ryzen chips is 3,600MHz DDR4, although you'll occasionally see recommendations up to 3,733MHz.

You can see the effect of memory speed on latency over in Chris Szewczyk's recent deep-dive into whether RAM speed matters for gaming. But essentially if you crank up memory speed significantly beyond 3,600MHz on an AMD Ryzen 3000 chip, expect latency to increase with it.

The leaked slides from Technopat (via WCCFTech) suggests that AMD Ryzen 5000 chips will be able to run DDR4-4000 memory while remaining in the sweet spot for high-speed, low-latency performance. It's not a massive bump for compatibility, but with memory manufacturers cranking out kits above 5,000MHz, it's cheaper than ever to get a high speed dual-channel kit.

And faster memory makes for a happier gaming PC, especially with AMD Ryzen.

Other slides from Technopat, and allegedly from AMD, show further evidence of a performance lead for AMD Ryzen 5000 over Intel's Comet Lake in gaming and that support for AMD 400-series boards will arrive in January, 2021 in the form of a BIOS update.

This all remains unconfirmed for now, but we shouldn't have to wait long to find out more. AMD Ryzen 5000 processors launch November 5, starting with the Ryzen 9 5950X, Ryzen 9 5900X, Ryzen 7 5800X, and Ryzen 5 5600X.

PC Gamer followed up the above with coverage of the launch of G. Skill's latest update to those coveted lines:
G.Skill launches 4,000MHz RAM for lucky AMD Ryzen 5000 CPU owners

G.Skill is updating its popular Trident Z Neo memory to 4,000MHz (effective) speeds, which places it right in the so-called 'sweet spot' for AMD Ryzen 5000 processors. With a bit of luck, you'll be running at the optimal 1:1 ratio across memory, Infinity Fabric, and memory controller in no time...

G.Skill's Trident Z Neo memory will now be available in DDR4-4000 CL16 configurations, which will offer a decent blend of speed and latency for AMD Ryzen builds. G.Skill is rolling out in both 16GB and 32GB capacity kits. If you want 64GB, you'll have to settle for lower latencies at CL18.

All kits are built with Samsung B-die chips.

AMD Zen 3 will support up to 4,000MHz memory at a 1:1 ratio. That last bit is the important part. That ratio applies to the memory clock (mclk), the memory controller clock (uclk), and Infinity Fabric clock (fclk), and it's best if you keep all operating at an even beat. Hence the desire to maintain a 1:1 ratio.