AMD Ryzen 7 2700x - Verified 15% performance decrease

iphorde

Honorable
Oct 1, 2018
15
0
10,510
AMD has introduced a performance penalty in their latest update to their AGESA 1.0.0.4 Microcode. This update decreases the multithreaded performance of the CPU by more than 15%.

I have been in contact with Gigabyte and they acknowledged the problem and were able to reproduce the results as well. The performance decrease can be noticed using real world applications and synthetic benchmarks such as Passmark's PerformanceTest.

Gigabyte said they would be contacting AMD about the issue.

I have since tested this on two other motherboard manufactures and found the same result.

Could someone test the Ryzen 5 2600x and see if this penalty exists there as well?

My guess at the moment is a variant of Spectre flaw was patched and has critically hindered the performance and AMD just silently didn't tell anyone.


open

 
Solution
There are some claims that newest W10, 1809 lowered performance by 10% but I'm running Skip ahead 1809 and couldn't verify that either. all runs as usual or at least same benchmark scores. Linux Mint doesn't show any different.

iphorde

Honorable
Oct 1, 2018
15
0
10,510


Memory profile settings don't impact this one way or another. But yes, all BIOS settings have be thoroughly tested. This is an AMD problem.


 

xeon_fan

Reputable
Feb 3, 2016
105
0
4,710


Any news regarding this one ?

 

Which exact AGESA ? Improvement comparing to what ?
Agesa AGESA 1.0.0.2a (PinaclePi) for Asus x470 got security update with one BIOS version (4011) with SMU 43.18 and since than (even with newest BIOS (4024) peek performance is down by 5 -7% but average performance and stability is not negatively affected.

 

iphorde

Honorable
Oct 1, 2018
15
0
10,510


If you read the above it is version 1.0.0.4. It is more than 15% of applications such as Blender, compiling with Visual Studio, CMake/g++, …


The real problem is that Intel and AMD are public companies and they need to notify their customers that their processors are not performing as they advertised. You buy something expecting it not to have 15% of the performance cut off the top after you buy it. This is becoming a real problem and they lying through omission.

There has been no response from AMD. None, they're staying silent it appears.

I would start shorting the stock. When the market realizes the processors are not the "hail marry" pass everyone thought they were, Intel will start looking attractive again. If this problem exists in the Ryzen series, it will exist in the Thread Ripper series.



 
hogwash! intel already suffer all or MORE Spectre and Meltdown pain. just look at the thousands of posts of intel boxes stuttering in most games... lol

intel won't ever look attractive. always bloated and overpriced.

what they--all cpu plaers--need are better processor designs that can't be hacked
 
OK, had to find it for myself. 1004 is also PinaclePI version. As I said before, it's most probably that security update at fault.
As far as companies keeping mum, that's nothing new. Intel knew about security holes months before it was discovered by 3rd party, never did anything about it until outcry and than just turned off performance enhancing feature giving them advantage over comparable AMD processors.
Apparently, Ryzen security flaw is much more beguine but after patch it did lower performance.
Abhorrent PR but nothing new. Remebner Pentium 90 math processor flaw, Nvidia 1/2GB unusable memory, AMD Rx 560 1/2 number of shaders with some manufacturers..... etc.
I don't have 1004 in my BIOS so can't comment on that but only what I found after the patch. It was offset with better OC and memory handling though.
 

iphorde

Honorable
Oct 1, 2018
15
0
10,510
Other than the fact Gigabyte verified the problem, and I have degrees in Computer Science and with an emphasis in Computer Architecture and was a game developer for years and years. I guess my street cred isn't legit enough anymore.

Run the benchmarks for yourself and find out.

dmL5oAKIeYeNtm3Jvdgqa5Luvmdqbez51BejRp0XCZXAfzjqN6os0uaYqz3X7emVeVc6DW20_Su65UWENyUBQFCqX1iSFIioFv1vkjecgBPgAIt8we8SNnPy5qypd-I3JmKq4QXRx09H4MRSb3dI9D3fX6qi40NsTYnMCt8kANHL2z9fX3fFbmP5N5wOX1ys7HN6oQpvvpUEOinUcoNvNzk9kIwM6Q59GwkqpuALO6_KcVKzBciwRuWFer8TgGr_EAVUBpsNzr9_i5SKs6xMRsOA3Vd4DQwLZTG3f6c4mM2NPwjRctKEFBbZMbhoniE1bDYcIdmADTSbx1gCBblxoyU9y7QF32W0yzQ6m6CJ0QIm50hPOpQf5vCJlo7ta5JK5D-q3FynV8sqvOv0Jmb4SF7kMqQwGMpO5GYfRkTnxApMiJ9RIn7LesIfcL1HCUiQSPMid2_jMc_L2vVyany1rz9_AHysKSZu8ccrr-M0Khg73lZK5a8mICA8NGZNporCJnA-JKAq0hvd69FH6FJxQ8n0Uoo4OxMuX2Pb-_-XgWnm8u0C3Md_iyqeUL8QiZaaLR3TeqX3uHtQuuNglhUEIZ5XG1Qphupa8SMMhRzsKSG8NWwDPm6PIlotcW97g6y8=w854-h216-no


GjGTsl6DxLsl7rqSVPoo6NKRB-R5n4JAp15gZ06Pkq6iLtFWcGb6Mr00V5IZ7M3Lnm83t_zEjSMbPBj7NCEn1UiJhx_XsetKmrUU8b9WGesLAIuqUucL0t_n6u-S0UMAJbFoyZSTdEJ4abTCYnS7YBFfaJkBcErOwYqiBYEEEvcAorCCQt1DxC3URaD77IfdTJ62q4wpUM7CAEQ_0EagFMf_bhE3dKnzgmTbeVFr-TqWHof0dgdiYhm1N_U2xaI2E1q7-iZYviFx5-NNz_8wfmL9AQNpONObToNW757GBfOGdAn1WFlrvJbXr58CycQrGhMIuT49ZMp9bBqCQCJb8-Hxy4-uuhqoctzdEzCk7o8O_6HhbMhEtMNQRJ2TItMPNVsaCou6cUuxt8stj98S_4yl2-3UujU3pFzxRZWWv95S30EEDaynBmnDVom2pglra9wcNGvR_TzhXh9dze6clprPwJN-ryQii7QyJLnoSu9upV9A3EG1umiIYsJ7-ghwrT47DA6_7FLQrJf3RyPsAd_4EmrhJQWPrwv7kaIupzv5IZF7dwMj031G7GFaie8WEupH41svIvhtCRujSDCHkrWweoS9drr34cxPwYM76P-xi9ePYi0sLy4Q05brwqZu=w923-h269-no




 
Oct 2, 2018
1
0
10


But your post doesn't show Gigabyte verifying the 15% performance loss you claimed, only that the only changes they from F2 to F3 was the AGESA code. Do you have any other sources verifying the performance loss?

"I would start shorting the stock" - That's not sketchy at all.
 

iphorde

Honorable
Oct 1, 2018
15
0
10,510


Run the benchmarks for yourself. You can do as you wish. Do the work, then run the benchmark, and get back to us. Tell us if you find something different. Make sure you create a baseline configuration for the BIOS and then test each version.

What other source do you want? Do you want AMD to announce they have a big screw up? That would be great.

I want AMD to fix the problem, but they're not responsive. Hopefully this will start getting their attention.

And yes, they did acknowledge the problem. They said use version F2 to get the best performance.

 

iphorde

Honorable
Oct 1, 2018
15
0
10,510


I've tested it both ways. It didn't have an impact. That is interesting to know that it isn't impacting the 1600x. How have you verified that it hasn't impacted 1600x?
 

iphorde

Honorable
Oct 1, 2018
15
0
10,510


Your 1600x is out performing the 2700x's we have. Pull up Z-CPU and record a video of the benchmark with it overlaid.
 
Other than the fact Gigabyte verified the problem, and I have degrees in Computer Science and with an emphasis in Computer Architecture and was a game developer for years and years. I guess my street cred isn't legit enough anymore.

Yeah, your "cred" is not legit enough here. This is an internet forum and no one knows who you are. There are a lot of people on this forum who's credibility is good. But you made an account on this site 2 days ago and you go on with unsubstantiated claims and recommend short selling stock. Don't be surprised if people ask a few questions.

 

iphorde

Honorable
Oct 1, 2018
15
0
10,510


Also, I found this article, which shows for mainly single threaded games, they also so a performance drop on the 1800x with 1.0.0.4.

https://thetechaltar.com/amd-ryzen-agesa-1-0-0-4-testing/