AMD Ryzen 5 5600X Takes the Crown of the Fastest CPU in Passmark Single-Thread Results

[BoboOOZ]

You guys are being way too patient trying to reason with fanboys. Anyway, if you want a more historical perspective on how intel lies, here you go:

Rounding a few numbers on a presentation is really on the nice side.

 

[RandallFlagg]

I never said anything about OC, you keep bringing it up.
Here we are talking about the 5000 series CPUs and you have gone back 2 gens for both Intel and AMD talking about OC... to proof what exactly? How superior Intel Is?

Well have it...

Raja Koduri to Present at Samsung Foundry Forum amid Intel's Outsourcing Efforts

Intel's chief architect and senior vice president of discrete graphics division, Mr. Raja Koduri, is said to be scheduled to present at Samsung Electronics Event day. With a presentation titled "1000X More Compute for AI by 2025", the event is called Samsung Foundry SAFE Forum. It is a global...

www.techpowerup.com

If you go back and read what I was responding to - nevermind, i'll do that for you too.

It was a rebuttal to the veracity of AMDs performance claims. In relation to Zen 3, that chart I posted about Zen 2 vs Gen 9, is pretty close to their current claims. I seriously doubt you are going to see a 5600X unseat say a 10600K in gaming in the real world. Match it - maybe. Beat it in any significant way, not likely.

 

[Deleted member 24505]

Funny how only intel users are fanboys to amd users.

Sick of all this crap on TPU.

If this 5600x is as good as it seems, then i will probably switch to it. I have no care for either Intel/Amd or Nvidia/Radeon. I just buy what is best for what i can afford.

 

[thesmokingman]

How can you talk about the performance of a 3600 then turn around and give power numbers for a 2700x, come on man...

You have to ignore the resident shill.

 

[Makaveli]

Funny how only intel users are fanboys to amd users.

Sick of all this crap on TPU.

Nothing funny about it. There are fan boys on both sides and its also not a TPU only issue.

this crap goes on all over the internet.

At the end of the day your allegiance should be to your wallet alone and buy what you like.

 

[Zach_01]

If you go back and read what I was responding to - nevermind, i'll do that for you too.

It was a rebuttal to the veracity of AMDs performance claims. In relation to Zen 3, that chart I posted about Zen 2 vs Gen 9, is pretty close to their current claims. I seriously doubt you are going to see a 5600X unseat say a 10600K in gaming in the real world. Match it - maybe. Beat it in any significant way, not likely.

This is a different matter than OC capabilities. I understand, you have no faith to AMDs claims. Its your right. But bringing a slide that claims parity in 8 games and then a TPU chart that proof this wrong by 2.7% in 15 games is not actually a discovery nor that disproofs AMD's integrity. If we start this about marketing claims and perspectives that corporations try to present we will end by 2022. Intel has worst claims in the recent years. AMD seems to getting better their sh1ts together the last couple of years if not more. And in last ZEN3 anouncment they seemed more professional. They even showed games that traditionally lose.

I know and many know that synthetics do not tell the truth about actual real life performance. But tell me (or show) this... How much faster is the 10600k in gaming (1080p) than a 3600X?

 

[RandallFlagg]

...

I know and many know that synthetics do not tell the truth about actual real life performance. But tell me (or show) this... How much faster is the 10600k in gaming (1080p) than a 3600X?

It's in the chart I posted, 7.6% on average FPS at stock speed.

To note, these differences become even more pronounced with new faster GPUs. The 9900K had a 4.8% advantage over the 3900XT in TPUs comparison using the 2080 Ti.

When TPU did the 3900XT vs 9900K vs 10900K comparison with a 3080, that difference went to a full 10%.

It's clear Zen 2 was already limiting gaming performance, but Skylake+++ had more breathing room. It will be interesting to see what happens when they put a Zen 3 up against Gen10 with one of these new GPUs.

 

[Zach_01]

It's in the chart I posted, 7.6% on average FPS at stock speed.

To note, these differences become even more pronounced with new faster GPUs. The 9900K had a 4.8% advantage over the 3900XT in TPUs comparison using the 2080 Ti.

When TPU did the 3900XT vs 9900K vs 10900K comparison with a 3080, that difference went to a full 10%.

It's clear Zen 2 was already limiting gaming performance, but Skylake+++ had more breathing room. It will be interesting to see what happens when they put a Zen 3 up against Gen10 with one of these new GPUs.

+7.6% difference for the 10600k. And if we apply the same 3080 rule (3900XT vs 10900k) that difference will go were? to 15%? maybe 16%?
And the 5600X with +20% IPC and and another 4.5% speed uplift on top of that wont make it past that 16%?
I know its not 1:1 for IPC+MHz against gaming % perf but still, will see.
And dont we start about the multi threaded perf... or the power draw... let it go.

 

[RandallFlagg]

+7.6% difference for the 10600k. And if we apply the same 3080 rule (3900XT vs 10900k) that difference will go were? to 15%? maybe 16%?
And the 5600X with +20% IPC and and another 4.5% speed uplift on top of that wont make it past that 16%?
I know its not 1:1 for IPC+MHz against gaming % perf but still, will see.
And dont we start about the multi threaded perf... or the power draw... let it go.

IPC is really almost meaningless, I've not seen any site effectively measure it as it is very use case specific. For example, you choose to use Cinebench to measure IPC. And what if I choose to user SuperPi, where a 4.3Ghz 10400 whacks a 4.5Ghz 3600XT to the tune of 8% better performance with 4% lower clocks? That would mean the 10400 has 10-12% better IPC, now wouldn't it?

Basically in the end, it all comes down to the benchmarks in the things you use. Every time someone starts talking nm / IPC / Ghz - that's just meaningless drivel in the end.

 

[Valantar]

It's 179W on torture test. Note that is (far) less than a 2700X stock on the same test:

View attachment 173421

Again, you are very much misrepresenting the truth here. Are you even reading the graphs you are posting? 179W is a lot more than the 125W of a stock 2700X in the same test. It is less than the 223W of an overclocked 2700X in the same test, and the stock 9600K at 119W is slightly lower than the 2700X at stock.

If you go back and read what I was responding to - nevermind, i'll do that for you too.

It was a rebuttal to the veracity of AMDs performance claims. In relation to Zen 3, that chart I posted about Zen 2 vs Gen 9, is pretty close to their current claims. I seriously doubt you are going to see a 5600X unseat say a 10600K in gaming in the real world. Match it - maybe. Beat it in any significant way, not likely.

A) AMD's performance claims this go around are not only about a general IPC uplift, but also specific improvements in latency-sensitive applications like games. With the 3600X they made no such claims, only an overall IPC improvement. In other words, Ryzen 5000 is specifically made to improve on this one point where Intel was still beating them, and it has a major architectural change (the unified L3 cache) as well as several smaller ones done with the express intent of achieving this.
B) With Zen 2 vs. Gen 9, AMD was going from a ~8% IPC deficit to a ~7% IPC advantage. This time they're going from a ~7% IPC advantage to a ((107*1.19)-100=) 27.33% IPC advantage. That is certainly a much bigger change than last time around in relation to Intel's current desktop chips. (Ice Lake and Tiger lake have ~18% higher IPC than Skylake, but are only available for mobile.)

 

[Zach_01]

And what if I choose to user SuperPi, where a 4.3Ghz 10400 whacks a 4.5Ghz 3600XT to the tune of 8% better performance with 4% lower clocks?

Does it? I really dont know... Im asking.

And indeed how a SuperPi can determine CPU gaming performance... but nor a Passmark can.

And I didnt choose CB test for IPC determination uplift. AMD claimed that for an avg result over many different workloads. Not just Cinebench... So...

 

[RandallFlagg]

Does it? I really dont know... Im asking.

And indeed how a SuperPi can determine CPU gaming performance... but nor a Passmark can.

And I didnt choose CB test for IPC determination uplift. AMD claimed that for an avg result over many different workloads. Not just Cinebench... So...

I'm pointing out that it is trivial for me to choose a workload that favors one cpu over the other, so how much more trivial is it for AMD to choose a workload - or mix of workloads as you say - that puts their product in the best light. SuperPi isn't the only one where Intel excels in such a way that it would refute IPC claims, it also excels in OCR applications and 2D -> 3D image conversion as well as AI benchmarks. What say I make my workload mix made of those benchmarks?

Remember, you're the one that brought up 20% increase in "IPC", I simply said that IPC is meaningless.

So I think the real difference here, with Zen 3, is going to look like the difference between the 4.0Ghz 3100 and the 3300X on the chart in the link below, plus a few % for the clock increases. You'll just get it across the board at all SKUs. That would come out to more like +6-7%. Which would bring them roughly to parity with Comet Lake.

AMD Ryzen 3 3300X Review - The Magic of One CCX

AMD Ryzen 3 3300X crams all its cores into a single CCX. We tested the CCX impact in our review with impressive results, especially in games, where the new CPU design achieves great numbers that are close enough to more expensive Ryzen 5 and Ryzen 7 models, especially if you consider the cost...

www.techpowerup.com

 

[Zach_01]

I'm pointing out that it is trivial for me to choose a workload that favors one cpu over the other, so how much more trivial is it for AMD to choose a workload - or mix of workloads as you say - that puts their product in the best light. SuperPi isn't the only one where Intel excels in such a way that it would refute IPC claims, it also excels in OCR applications and 2D -> 3D image conversion as well as AI benchmarks. What say I make my workload mix made of those benchmarks?

Remember, you're the one that brought up 20% increase in "IPC", I simply said that IPC is meaningless.

So I think the real difference here, with Zen 3, is going to look like the difference between the 4.0Ghz 3100 and the 3300X on the chart in the link below, plus a few % for the clock increases. You'll just get it across the board at all SKUs. That would come out to more like +6-7%. Which would bring them roughly to parity with Comet Lake.

AMD Ryzen 3 3300X Review - The Magic of One CCX

AMD Ryzen 3 3300X crams all its cores into a single CCX. We tested the CCX impact in our review with impressive results, especially in games, where the new CPU design achieves great numbers that are close enough to more expensive Ryzen 5 and Ryzen 7 models, especially if you consider the cost...

www.techpowerup.com

What exactly the statement of "+19% IPC uplift in geomean of 25 workloads" could mean?

By the way, the example of 3100 vs 3300X is irrelevant. The claimed +19 IPC uplift that AMD has stated is not only from the unified (and larger now) cache but from a lot more architectural improvements



I really dont think that AMD will risk to embarace its self with these kinds of slides and in the end will be on par with the competition while they increase prices. Even the rocks will laugh and they know it. This time I think its serious, and some are doing everything they can to minimize the gravity of it.





Those 2 CPUs have only 100MHz difference. Do they dare to show this and be on par with Intel? How much of a difference would that be between 3600X, 5600X and 10600k? I wonder.

 

[Valantar]

I'm pointing out that it is trivial for me to choose a workload that favors one cpu over the other, so how much more trivial is it for AMD to choose a workload - or mix of workloads as you say - that puts their product in the best light. SuperPi isn't the only one where Intel excels in such a way that it would refute IPC claims, it also excels in OCR applications and 2D -> 3D image conversion as well as AI benchmarks. What say I make my workload mix made of those benchmarks?

Remember, you're the one that brought up 20% increase in "IPC", I simply said that IPC is meaningless.

So I think the real difference here, with Zen 3, is going to look like the difference between the 4.0Ghz 3100 and the 3300X on the chart in the link below, plus a few % for the clock increases. You'll just get it across the board at all SKUs. That would come out to more like +6-7%. Which would bring them roughly to parity with Comet Lake.

AMD Ryzen 3 3300X Review - The Magic of One CCX

AMD Ryzen 3 3300X crams all its cores into a single CCX. We tested the CCX impact in our review with impressive results, especially in games, where the new CPU design achieves great numbers that are close enough to more expensive Ryzen 5 and Ryzen 7 models, especially if you consider the cost...

www.techpowerup.com

IPC will always be a simplification, as you can never test all the workloads. But there are still industry standard benchmarks that are widely seen as reliable, such as SPECcpu. Anandtech uses that for their IPC testing, and found Zen 2 to both match AMD's claims (despite them not being based on testing with SPEC) and overall beat Skylake++++ by ~7%. This clearly doesn't cover gaming performance, but it doesn't cover SuperPi either, given that it's not actually a useful workload. There is as such little reason to expect AMD's numbers for Zen 3 to not be representative, though there will as alwyd be outlier tests.

As for gaming performance, they are claiming much, much more than 19% performance improvement: their wording is that the 5900X when compared to the 3900X is "from 5% to 50% improved" with the overall improvements in their testing at 26%. You're presenting it as if they were saying "IPC is better, so games are better", but that is very explicitly not what they are saying; the claims about gaming performance are separate from the IPC claims, and are about specific SKUs compared (in other words including clock speed increases). Again, there will obviously be examples where Intel still wins. But there's little reason to not expect AMD to be on the money when they claim to beat the 10900k overall in gaming.

 

[RandallFlagg]

Here are the full results from that Passmark, comparison to 5950X and 10900 (nonK).

Some of these scores don't quite add up to the hype. Keep in mind the i9-10900 here is limited on the graphics tests due to the 2070 Super VS 3080s in the two Ryzen boxes.

If these scores are real the performance on the 5600X is somewhat of a mixed bag.

What's funny here is, the one everyone posted is CPU Single Threaded MOps/Sec, ignoring everything else.

CPU	AMD Ryzen 5 5600X	AMD Ryzen 9 5950X	Intel Core i9 10900
GPU	RTX 3080	RTX 3080	RTX 2070 Super
Memory	CMW16GX4M2C3200C16	F4-3600C16-8GTRG	F4-3466C16-8GTZR
Drive	CT500P1SSD8	T-Force TM8FP500	Samsung 970 EVO Plus 1TB
Integer Math (MOps./Sec.)	77936	202089	84927
Floating Point Math (MOps./Sec.)	40368	106087	53041
Prime Numbers (Million Primes/Sec.)	136	226	82
Extended Instructions (SSE) (Mill. Matrices/Sec.)	17362	38147	25627
Compression (KBytes/Sec.)	266462	616165	376500
Encryption (MBytes/Sec.)	16740	40316	9003
Physics (Frames/Sec.)	1188	1366	1302
Sorting (Thousand Strings/Sec.)	27851	67871	47355
CPU Single Threaded (MOps./Sec.)	3495	3693	3336
Cross-platform Mark (Composite average)	44089	102760	45984
Simple Vectors (Thousand Vectors/Sec.)	17	15	17
Fonts and Text (Ops./Sec.)	281	290	328
Windows Interface (Ops./Sec.)	58	68	84
Image Filters (Filters/Sec)	2946	2653	2673
Image Rendering (Thousand Images/Sec)	375	318	401
Direct 2D (Frames/Sec.)	75	79	56
PDF Rendering (Ops./Sec.)	57	87	70
Direct 2D - SVG (Frames/Sec.)	100	106	75
DirectX 9 (Frames/Sec.)	275	143	264
DirectX 10 (Frames/Sec.)	373	60	293
DirectX 11 (Frames/Sec.)	537	141	435
DirectX 12 (Frames/Sec.)	71	90	61
GPU Compute (Ops./Sec.)	16466	15696	9106
Database Operations (KOps./Sec.)	4545	8199	8444
Memory Read Cached (MBytes/Sec.)	35022	37384	36442
Memory Read Uncached (MBytes/Sec.)	20791	25431	20105
Memory Write (MBytes/Sec.)	12067	17981	18359
Available RAM (Megabytes)	13869	10317	27029
Memory Latency (ns (lower is better))	56	41	24
Memory Threaded (MBytes/Sec.)	30379	53945	43982
Disk Sequential Read (MBytes/Sec.)	1817	434	2890
Disk Sequential Write (MBytes/Sec.)	909	400	2676
IOPS 32KQD20 (MBytes/Sec.)	370	299	1393
IOPS 4KQD1 (MBytes/Sec.)	85	34	76
CPU Mark (Composite average)	22824	45564	24682
2D Graphics Mark (Composite average)	1123	1182	1141
Memory Mark (Composite average)	2692	3438	4271
Disk Mark (Composite average)	11305	4147	24995
3D Graphics Mark (Composite average)	25258	10850	20444
PassMark Rating (Composite average)	7454	7483	9018

 

[Makaveli]

Here are the full results from that Passmark, comparison to 5950X and 10900 (nonK).

Some of these scores don't quite add up to the hype. Keep in mind the i9-10900 here is limited on the graphics tests due to the 2070 Super VS 3080s in the two Ryzen boxes.

If these scores are real the performance on the 5600X is somewhat of a mixed bag.

What's funn here is, the one everyone posted is CPU Single Threaded MOps/Sec. According to this, the 5950X is faster at single thread than the 5600X.......

CPU	AMD Ryzen 5 5600X	AMD Ryzen 9 5950X	Intel Core i9 10900
GPU	RTX 3080	RTX 3080	RTX 2070 Super
Memory	CMW16GX4M2C3200C16	F4-3600C16-8GTRG	F4-3466C16-8GTZR
Drive	CT500P1SSD8	T-Force TM8FP500	Samsung 970 EVO Plus 1TB
Integer Math (MOps./Sec.)	77936	202089	84927
Floating Point Math (MOps./Sec.)	40368	106087	53041
Prime Numbers (Million Primes/Sec.)	136	226	82
Extended Instructions (SSE) (Mill. Matrices/Sec.)	17362	38147	25627
Compression (KBytes/Sec.)	266462	616165	376500
Encryption (MBytes/Sec.)	16740	40316	9003
Physics (Frames/Sec.)	1188	1366	1302
Sorting (Thousand Strings/Sec.)	27851	67871	47355
CPU Single Threaded (MOps./Sec.)	3495	3693	3336
Cross-platform Mark (Composite average)	44089	102760	45984
Simple Vectors (Thousand Vectors/Sec.)	17	15	17
Fonts and Text (Ops./Sec.)	281	290	328
Windows Interface (Ops./Sec.)	58	68	84
Image Filters (Filters/Sec)	2946	2653	2673
Image Rendering (Thousand Images/Sec)	375	318	401
Direct 2D (Frames/Sec.)	75	79	56
PDF Rendering (Ops./Sec.)	57	87	70
Direct 2D - SVG (Frames/Sec.)	100	106	75
DirectX 9 (Frames/Sec.)	275	143	264
DirectX 10 (Frames/Sec.)	373	60	293
DirectX 11 (Frames/Sec.)	537	141	435
DirectX 12 (Frames/Sec.)	71	90	61
GPU Compute (Ops./Sec.)	16466	15696	9106
Database Operations (KOps./Sec.)	4545	8199	8444
Memory Read Cached (MBytes/Sec.)	35022	37384	36442
Memory Read Uncached (MBytes/Sec.)	20791	25431	20105
Memory Write (MBytes/Sec.)	12067	17981	18359
Available RAM (Megabytes)	13869	10317	27029
Memory Latency (ns (lower is better))	56	41	24
Memory Threaded (MBytes/Sec.)	30379	53945	43982
Disk Sequential Read (MBytes/Sec.)	1817	434	2890
Disk Sequential Write (MBytes/Sec.)	909	400	2676
IOPS 32KQD20 (MBytes/Sec.)	370	299	1393
IOPS 4KQD1 (MBytes/Sec.)	85	34	76
CPU Mark (Composite average)	22824	45564	24682
2D Graphics Mark (Composite average)	1123	1182	1141
Memory Mark (Composite average)	2692	3438	4271
Disk Mark (Composite average)	11305	4147	24995
3D Graphics Mark (Composite average)	25258	10850	20444
PassMark Rating (Composite average)	7454	7483	9018

You still going on with the hard reaching? You are not going to stop anyone from going Zen 3 with these post nor are you some kinda guru of hidden truth. Give it up!

The 5950X is suppose to be faster then 5600X in single thread because of clock speed.

The later has a 4.9Ghz Turbo and the other is 4.6Ghz.

 

[Valantar]

You still going on with the hard reaching? You are not going to stop anyone from going Zen 3 with these post nor are you some kinda guru of hidden truth. Give it up!

The 5950X is suppose to be faster then 5600X in single thread because of clock speed.

The later has a 4.9Ghz Turbo and the other is 4.6Ghz.

Not to mention that in a lot of those tests a $300 6-core AMD CPU is beating a $500 10-core Intel CPU ... There are obviously use cases where the 10900K is faster still, but the more time passes, the fewer there seem to be.