***OFFICIAL*** Ryzen 5000 / Zen 3 Launch Thread REVIEWS BEGIN PAGE 39

[shady28]

The gaming performance looks underwhelming to me. It's a bit faster than CML-S with a RAM speed advantage. With the same RAM the gaming peformance seems to be on par:

AMD Ryzen 9 5900X Review

The Ryzen 9 5900X dominates Intel's Core i9-10900K in our testing because of AMD's massive IPC improvements. At $550, this processor is certainly not cheap, but it offers so much more performance, especially single-threaded, that AMD has a clear winner on their hands.

www.techpowerup.com

Considering that CML-S is based on the Skylake architecture this isn't great.

Zen 3 just isn't all that great with anything lower than DDR4-3600. I'm a bit disappointed that none of these 'professional review' sites seem to have caught onto it. TPU has always used DDR4-3200 CL14 on both platforms, which normally favors Zen 2 a bit. However, it wasn't a flattering setup for Zen 3.

 

[yeshua]

If you call the 5600X overpriced junk, what have you been calling the 10600K? I swear I don't wanna search for your comments, so I'm sorry if you already stated your opinion before. I'm really curious.

In my opinion most Intel uArchs released after Sandy Bridge were hot junk, except Sky Lake and Broadwell. The company innovated at a snail's pace and did everything in their power not to allow AMD to compete. As for this particular CPU - no opinion at all. I now have a Ryzen 7 3700X and had an Intel Core i5 2500 before it. I upgrade my CPU only if the new one offers a very perceivable performance boost - upgrading yearly or biennially? It's not that easy to sell your old stuff at a good price here, so that's not for me. Cheers!

Lastly the 5600X itself is a great CPU but its price makes it bad. Again, the 5600X is the 50% more expensive replacement of the 3600. It's ugly.

 

[lobz]

No worries. You can find a lot of examples on PCMark site right now. Pull up time spy, single GPU, dual channel memory (to eliminate HEDT systems), sort by CPU score, limit to 5900X and then 5950X. The highest scoring 5950X and 5900X are running at DDR4-3800. Do the same to the Intel 10900K and set max core speed to 4900 (the max all core turbo Intel specs on the 10900K). The Intel rigs in this range are running DDR4-4400 to DDR4-4600.

You missed my point. As I said and I really don't know the answer to: what realistic advantage do Intel CPUs get from being able to run memory at 4400-4600 instead of 3900-4000?

 

[Hitman928]

Btw, yeshua, your previous nickname on AT forum was birdie?

He's been copying and pasting his anti-AMD posts on other forums under the nickname birdie, so most likely, yes.

 

[lobz]

In my opinion most Intel uArchs released after Sandy Bridge were hot junk, except Sky Lake and Broadwell. The company innovated at a snail's pace and did everything in their power not to allow AMD to compete. As for this particular CPU - no opinion at all. I now have a Ryzen 7 3700X and had an Intel Core i5 2500 before it. I upgrade my CPU only if the new one offers a very perceivable performance boost - upgrading yearly or biennially? It's not that easy to sell your old stuff at a good price here, so that's not for me. Cheers!

Lastly the 5600X itself is a great CPU but its price makes it bad. Again, the 5600X is the 50% more expensive replacement of the 3600. It's ugly.

Thanks and I can totally respect your opinion. I'm not happy about the pricing, for example the 5800X does not even exist to me, just like the whole XT refresh lineup.
However, I can also put this into a bit of perspective say this is still much better than what the market would have becone without Zen. This is still acceptable for me (that doesn't have to mean every single SKU) and if AMD doesn't go down the way they did 15 years ago, I'm content with they way things unfolding. Here's hoping Intel can keep them on their toes, and also that Intel will never succeed ever again in making AMD irrelevant.

 

[yeshua]

Thanks and I can totally respect your opinion. I'm not happy about the pricing, for example the 5800X does not even exist to me, just like the whole XT refresh lineup.
However, I can also put this into a bit of perspective say this is still much better than what the market would have becone without Zen. This is still acceptable for me (that doesn't have to mean every single SKU) and if AMD doesn't go down the way they did 15 years ago, I'm content with they way things unfolding. Here's hoping Intel can keep them on their toes, and also that Intel will never succeed ever again in making AMD irrelevant.

To be frank duopolies are not much better than monopolies (I've got a very strong suspicion AMD and NVIDIA have been cooperating their GPU pricing strategies for the past at least four years) and I'm really sad we have just two x86 vendors and basically two GPU vendors - it's way too easy to collude without leaving any trail.

 

[Ranulf]

The 5800X runs pretty hot in all-core workloads due to 142W PPT being spread across 8 cores. I'm not sure if the NH-U12A is going to be enough.

Having run one of the older U12S heatsinks with a FX8350, I doubt the new U12A will have problems.

 

[shady28]

You missed my point. As I said and I really don't know the answer to: what realistic advantage do Intel CPUs get from being able to run memory at 4400-4600 instead of 3900-4000?

TimeSpy results, sorted by CPU, 10900K with max CPU frequency 4900 (this is max per Intel for all core boost). This is probably unlocked on power, but not overclocked.

The #1 10900K with that setup is running DDR4-4600 and scores 15503 on 3d Mark Time Spy CPU.

The #1 Ryzen 5900X is running DDR4-3600 and scores 15357. This is a 12 core processor vs 10 core.

The #1 Ryzen 5800X is running DDR4-3800 and scores 13191

So basically that power unlocked but not CPU overclocked 10900K with DDR4-4600 is beating both the 5900X and the 5800X. So yes faster memory matters. It's worth noting that at least the 5800X there appears to be overclocked.

Edit: Heres a better example. 10900 *Non-K*. The most you can do here is power unlock and do a +3% bclock overclock. Scores 14699 running DDR4-4400. So the 12 core 5900X is a whopping 4.5% faster in this test, and has 2 more cores.

I haven't found anything much on Zen 3 on those charts running at DDR4-4000 so far. If they are there, they aren't at the top.

 

[yeshua]

Speaking of the 5800X temps under load, here's an extended version (source: reddit - many people confirm it):

Spoiler

TechPowerUp : https://www.techpowerup.com/review/amd-ryzen-7-5800x/20.html (NH-U14s)
5900X: 67 ºC
5800X: 75 ºC
5600X: 59 ºC

Linus Tech Tips:

Spoiler: video

(NH-D15)
5950X and 5900X: ~70 ºC
5800X: ~80 ºC
5600X: ~61 ºC

Guru3D https://www.guru3d.com/articles_pages/amd_ryzen_7_5800x_review,6.html (240 mm LCS)
5950X and 5900X: 70 ºC
5800X: 73 ºC
5600X: 60 ºC

Anandtech https://www.anandtech.com/show/1621...e-review-5950x-5900x-5800x-and-5700x-tested/7 (cooler not specified)
5950X: 64 ºC
5800X: 90 ºC
5600X: 73 ºC

Just got mine today and temps have been driving me crazy too. I’m getting 90c within seconds at 100% load when stress testing with cinebench and aida64. I get around 70c when gaming (30-40% cpu): AC Odyssey, TW Warhammer 2, Apex Legends.

I have an X570 Tomahawk + 280mm aio (celsius+ s28) and I’ve disabled PBO, reapplied thermal paste multiple times now. I talked to another guy on the sub discord having the exact same problems and we suspect our batch has a bad interface b/w ihs and die.

 

[shady28]

To be frank duopolies are not much better than monopolies (I've got a very strong suspicion AMD and NVIDIA have been cooperating their GPU pricing strategies for the past at least four years) and I'm really sad we have just two x86 vendors and basically two GPU vendors - it's way too easy to collude without leaving any trail.

I'm not sure the same hasn't happened with Intel and AMD. The 10850K was a rather conveniently timed release, and very odd one that befuddled a lot of folks at the time. I think Intel knew everything there was to know about Zen 3 long before it was released. Intel's pricing also conveniently re-arranged itself before the Zen 3 release, such that they are competitive. The fact that happened before the official release seems like foreknowledge of both performance and pricing of Zen 3.

 

[Hitman928]

TimeSpy results, sorted by CPU, 10900K with max CPU frequency 4900 (this is max per Intel for all core boost). This is probably unlocked on power, but not overclocked.

The #1 10900K with that setup is running DDR4-4600 and scores 15503 on 3d Mark Time Spy CPU.

The #1 Ryzen 5900X is running DDR4-3600 and scores 15357. This is a 12 core processor vs 10 core.

The #1 Ryzen 5800X is running DDR4-3800 and scores 13191

So basically that power unlocked but not CPU overclocked 10900K with DDR4-4600 is beating both the 5900X and the 5800X. So yes faster memory matters. It's worth noting that at least the 5800X there appears to be overclocked.

Edit: Heres a better example. 10900 *Non-K*. The most you can do here is power unlock and do a +3% bclock overclock. Scores 14699 running DDR4-4400. So the 12 core 5900X is a whopping 4.5% faster in this test, and has 2 more cores.

I haven't found anything much on Zen 3 on those charts running at DDR4-4000 so far. If they are there, they aren't at the top.

Where are these coming from? Those all look like overclocked results. I would also be more interested in Timespy Extreme cpu results, although it is really more of a physics compute test and not really representative of actual game performance.

 

[shady28]

Edit - these links don't seem to pull up sorted by CPU score. Use the pulldown on the left to select sort by CPU score.

Edit2: you can do the same thing for TimeSpy Extreme. The 5800X seems to outpace the 10700 Non-K by ~9% here. This still to me is not impressive for a $450 CPU vs a $300-330 CPU.

Here is the link for the search for the 10900:

3DMark.com search

3DMark.com search

www.3dmark.com

The top score is running DDR4-4400. Note the 10900 can't be overclocked significantly.

Here's a link to the 5800X results, no filters, sorted by CPU score :

3DMark.com search

3DMark.com search

www.3dmark.com

For comparison 8C to 8C, here's a link to the 10700 Non-K. I'm sure power unlocked at the top, but again you can't OC a non-K CPU much, just the RAM.

3DMark.com search

3DMark.com search

www.3dmark.com

 

[Hitman928]

Edit - these links don't seem to pull up sorted by CPU score. Use the pulldown on the left to select sort by CPU score.

Edit2: you can do the same thing for TimeSpy Extreme. The 5800X seems to outpace the 10700 Non-K by ~9% here. This still to me is not impressive for a $450 CPU vs a $300-330 CPU.

Here is the link for the search for the 10900:

3DMark.com search

3DMark.com search

www.3dmark.com

The top score is running DDR4-4400. Note the 10900 can't be overclocked significantly.

Here's a link to the 5800X results, no filters, sorted by CPU score :

3DMark.com search

3DMark.com search

www.3dmark.com

For comparison 8C to 8C, here's a link to the 10700 Non-K. I'm sure power unlocked at the top, but again you can't OC a non-K CPU much, just the RAM.

3DMark.com search

3DMark.com search

www.3dmark.com

Looking at the scores, memory speed doesn't seem to have much effect. The fastest 10700 running 3822 MHz DDR4 is only 2% faster than a 10700 running 3000 MHz DDR4 and the one with the faster memory has a 2.2% faster reported clock speed to boot which is probably what makes for the main difference in scores.

I scored 8 124 in Time Spy Extreme

Intel Core i7-10700 Processor, NVIDIA GeForce RTX 3080 x 1, 32768 MB, 64-bit Windows 10}

www.3dmark.com

I scored 6 282 in Time Spy Extreme

Intel Core i7-10700 Processor, NVIDIA GeForce RTX 2080 Ti x 1, 65536 MB, 64-bit Windows 10}

www.3dmark.com

Again, Timespy (and all 3dmark CPU tests) are just physics compute tests. They are valuable tests, but shouldn't be used as representative of CPU gaming performance.

 

[shady28]

I'm not sure if you found some anomalous reading, but it's fairly well known that Gen 9 and Gen 10 scale up terrifically with RAM speeds.

See below. The 9900K going from JEDEC 2666 (the standard for 9th gen) to 3600 CL 16 got +15% FPS.

When they both OC'd the 9900K and went to 4133 CL17 they got +31%.

Edit: Stock was all core 4.7Ghz. OC was all core 5.0Ghz, so only 300mhz which is 6.4%. The rest of that extra 16% came from the RAM OC.

Core i9-9900K vs. Ryzen 9 3900X mit RAM-OC in Spielen: Testergebnisse und Fazit

CPU-Spieleleistung im Test: Testergebnisse und Fazit / Testsysteme und Testmethodik / Getestete Spiele / So wurde getestet / Testergebnisse

www.computerbase.de

 

[IntelUser2000]

I'm not sure if you found some anomalous reading, but it's fairly well known that Gen 9 and Gen 10 scale up terrifically with RAM speeds.

Core i9-9900K vs. Ryzen 9 3900X mit RAM-OC in Spielen: Testergebnisse und Fazit

CPU-Spieleleistung im Test: Testergebnisse und Fazit / Testsysteme und Testmethodik / Getestete Spiele / So wurde getestet / Testergebnisse

www.computerbase.de

That test also tells me that Zen 2 scales fantastically with RAM speeds. They should have tested DDR4-2666 as well. To see if the gap is larger/smaller than the 9900K with DDR4-2666.

People can also buy two systems and test them out. Trust me, you'll be very appreciated by both camps and forever have our gratitude.

You can also wait for RAM scaling reviews on Zen 3. They are sure to have some.

Zen 3 just isn't all that great with anything lower than DDR4-3600. I'm a bit disappointed that none of these 'professional review' sites seem to have caught onto it. TPU has always used DDR4-3200 CL14 on both platforms, which normally favors Zen 2 a bit. However, it wasn't a flattering setup for Zen 3.

Kinda sounds like AT's review for Skylake. All other reviews showed much better gains especially in games. Slow RAM was also the potential issue.

 

[Hitman928]

I'm not sure if you found some anomalous reading, but it's fairly well known that Gen 9 and Gen 10 scale up terrifically with RAM speeds.

See below. The 9900K going from JEDEC 2666 (the standard for 9th gen) to 3600 CL 16 got +15% FPS.

When they both OC'd the 9900K and went to 4133 CL17 they got +31%.

Edit: Stock was all core 4.7Ghz. OC was all core 5.0Ghz, so only 300mhz which is 6.4%. The rest of that extra 16% came from the RAM OC.

Core i9-9900K vs. Ryzen 9 3900X mit RAM-OC in Spielen: Testergebnisse und Fazit

CPU-Spieleleistung im Test: Testergebnisse und Fazit / Testsysteme und Testmethodik / Getestete Spiele / So wurde getestet / Testergebnisse

www.computerbase.de

This link doesn't even have Timespy tested and also doesn't have the memory frequency range in question.

I'll try to make this easy. The question was, do modern Intel CPUs even benefit from having memory speeds around 4500 MHz compared to around 4000 MHz. Your first response was to point to Timespy. All I was saying in my response is pointing to Timespy is not very helpful because Timespy CPU is just a physics compute test and will not tell you how a CPU will perform in actual games.

It's well known that Intel and AMD CPUs scale decently well in games with faster memory speeds, especially from 2100 MHz up to 3200 or 3400 MHz. After that they still scale well but you get diminishing returns. The question is, since Intel can handle memory frequencies above 4000 MHz, does it even matter? What is the scaling once you start to get above 4000 MHz in gaming (again, don't point to Timespy, it's not a gaming benchmark and does not scale with memory speeds).

 

[Hitman928]

After a very quick search, it looks like Intel CPUs don't benefit from RAM speeds above 4000 MHz in any significant way in gaming.

Memory Overclocking on Z390 Coffee Lake: What RAM Speed Do You Need?

We evaluate the impact of memory speeds from DDR4-2133 to DDR4-4400 on the Coffee Lake Refresh platform.

www.tomshardware.com

 

[tamz_msc]

Having run one of the older U12S heatsinks with a FX8350, I doubt the new U12A will have problems.

The heat density of these 7nm dies is significantly higher than that of the older dies built on 32nm. There are at least two published reviews(Anandtech and Computerbase) stating that the 5800X runs significantly hotter than the other chips, and now user reports from reddit are also coming in. All this indicates that heat is going to be a problem with the 5800X.

 

[maddie]

The heat density of these 7nm dies is significantly higher than that of the older dies built on 32nm. There are at least two published reviews(Anandtech and Computerbase) stating that the 5800X runs significantly hotter than the other chips, and now user reports from reddit are also coming in. All this indicates that heat is going to be a problem with the 5800X.

How exactly? Failure rates? Sensibilities?

 

[Dave3000]

The heat density of these 7nm dies is significantly higher than that of the older dies built on 32nm. There are at least two published reviews(Anandtech and Computerbase) stating that the 5800X runs significantly hotter than the other chips, and now user reports from reddit are also coming in. All this indicates that heat is going to be a problem with the 5800X.

If I can get a 5900x next week I think I'm going to go for that instead of the 5800x. I already purchased a Noctua U12A and I don't want to go for a high-end AIO (I prefer air cooling) or a bulkier air cooler. At that point I might as well spend the $100 more and go for the 5900x for the 4 more cores, full memory write bandwidth, and cooler temperatures, besides reviews I read so far are showing that gaming overall is slightly faster on the 5900x than the 5800x, despite the 8-core CCX on the 5800x, but still nothing that would be noticeable. However, I'd prefer to build my system sometime next week and if the 5900x is not in stock by then, I think I will purchase the 10700k, as I have one more week for the return period for my brand new power supply that I purchased 3 weeks ago.

 

[shady28]

After a very quick search, it looks like Intel CPUs don't benefit from RAM speeds above 4000 MHz in any significant way in gaming.

https://cdn.mos.cms.futurecdn.net/sDzEBANKRWUCkYTekPpt5X-970-80.jpg.webp

Do you have a link to the article? Those are just images saves somewhere, not even from their site, and for all I know that could be an 8700K or even an AMD chip.

Anyway, I'm not going to argue. Below is yet another link, and another set of images. There is actually more nuance here than simply faster RAM, the RAM has to fit the platform and frequency you're running at. That said, people who want to willfully remain ignorant, are welcome to do so.

Core i5 10600K and i9 10900K: Memory bandwidth analysis

The official Digital Foundry review of the Intel Core i9 10900K and Core i5 10600K, hosted by Eurogamer, including synthetic and gaming performance tests.

www.eurogamer.net

3000 Mhz Ram :


3600Mhz

 

[tamz_msc]

How exactly? Failure rates? Sensibilities?

If you're going to run heavy AVX2 loads(not power virus but something like Blender), then 80-90 degrees sustained temperatures is something you'd wish you didn't have to deal with. It's a problem the same way high temperatures under heavy workloads is a problem with the i9-10900K.

It would be hypocritical if AMD gets a pass while Intel is criticized heavily for their heat output.

 

[Hitman928]

Do you have a link to the article? Those are just images saves somewhere, not even from their site, and for all I know that could be an 8700K or even an AMD chip.

Apologies, I thought I put it in the post but I pasted a link to one of the images by accident.

Memory Overclocking on Z390 Coffee Lake: What RAM Speed Do You Need?

We evaluate the impact of memory speeds from DDR4-2133 to DDR4-4400 on the Coffee Lake Refresh platform.

www.tomshardware.com

Anyway, I'm not going to argue. Below is yet another link, and another set of images. There is actually more nuance here than simply faster RAM, the RAM has to fit the platform and frequency you're running at. That said, people who want to willfully remain ignorant, are welcome to do so.

Core i5 10600K and i9 10900K: Memory bandwidth analysis

The official Digital Foundry review of the Intel Core i9 10900K and Core i5 10600K, hosted by Eurogamer, including synthetic and gaming performance tests.

www.eurogamer.net

3000 Mhz Ram :

View attachment 33280
3600Mhz

View attachment 33281

I'm not trying to be rude, but I honestly don't even know what your argument is. You keep posting data that has nothing to do with the question in the post you originally replied to.

 

[lobz]

TimeSpy results, sorted by CPU, 10900K with max CPU frequency 4900 (this is max per Intel for all core boost). This is probably unlocked on power, but not overclocked.

The #1 10900K with that setup is running DDR4-4600 and scores 15503 on 3d Mark Time Spy CPU.

The #1 Ryzen 5900X is running DDR4-3600 and scores 15357. This is a 12 core processor vs 10 core.

The #1 Ryzen 5800X is running DDR4-3800 and scores 13191

So basically that power unlocked but not CPU overclocked 10900K with DDR4-4600 is beating both the 5900X and the 5800X. So yes faster memory matters. It's worth noting that at least the 5800X there appears to be overclocked.

Edit: Heres a better example. 10900 *Non-K*. The most you can do here is power unlock and do a +3% bclock overclock. Scores 14699 running DDR4-4400. So the 12 core 5900X is a whopping 4.5% faster in this test, and has 2 more cores.

I haven't found anything much on Zen 3 on those charts running at DDR4-4000 so far. If they are there, they aren't at the top.

Sigh........... I could also start citing Cinebench scores. None of the above mean a realistic advantage that may swing a purchase decision.

 

[lobz]

To be frank duopolies are not much better than monopolies (I've got a very strong suspicion AMD and NVIDIA have been cooperating their GPU pricing strategies for the past at least four years) and I'm really sad we have just two x86 vendors and basically two GPU vendors - it's way too easy to collude without leaving any trail.

Conspiracy theories are not my expertise for sure. I can't comment on this.