Core 9000 series

[sidthesexist]

One example is that he revealed in GN testing on ryzen2 chips he pushed fan's to 100% instead of AUTO,

Believe he said in yesterdays video that they do that for all chips they test?

 

[IT Troll]

I'll be back in an office this week so I'll have all the details. Off the top of my head the b350 mortar and Taichi boards are very good. If it's quality sounds you're after then go with an external DAC and some half decent speakers.

There isn't mATX AM4 Taichi (there is a nice Threadripper board though). The B350/B450 Mortar is one that only has heatsinks on half of the VRM. For audio I only want a nice clean headphone out, but with good driver support for surround sound etc. Quality on-board works better for me, or failing that a dedicated soundcard, rather than more boxes on the desk.

There are many at the ryzen discussion who have used the mATX B450 boards. Pop for a visit there.

I've spent a lot of time in the Motherboards, but no one was singing any praises.
https://forums.overclockers.co.uk/threads/about-to-give-up-on-am4-matx-any-news.18822747/

I will see if I can find some happy (but not deluded) mATX owners in some of the Ryzen threads.

 

[Panos]

There isn't mATX AM4 Taichi (there is a nice Threadripper board though). The B350/B450 Mortar is one that only has heatsinks on half of the VRM. For audio I only want a nice clean headphone out, but with good driver support for surround sound etc. Quality on-board works better for me, or failing that a dedicated soundcard, rather than more boxes on the desk.

I've spent a lot of time in the Motherboards, but no one was singing any praises.
https://forums.overclockers.co.uk/threads/about-to-give-up-on-am4-matx-any-news.18822747/

I will see if I can find some happy (but not deluded) mATX owners in some of the Ryzen threads.

Look, my question is why you need mATX? If you need something small, why not consider X470 mITX? They have all the features of the X470 boards, some are better than the B450 boards and you lose nothing.

 

[IT Troll]

If remember correctly from some pages back, it was said that the Asus guy assumes as chip fixed the patched motherboards supporting those CPUs.

I've read that several times and still can't understand it. The Asus rep said the CPU has no hardware fixes and just has the fixed microcode. He is wrong. The CPU has hardware fixes for Meltdown and L1TF.

 

[IT Troll]

Look, my question is why you need mATX? If you need something small, why not consider X470 mITX? They have all the features of the X470 boards, some are better than the B450 boards and you lose nothing.

Mostly so I can reuse my 4 sticks of RAM and so I have the option of fitting a sound card if the on-board is pants. Sure I could buy new RAM and an external DAC, but then you lose the value for money argument.

 

[jigger]

There isn't mATX AM4 Taichi (there is a nice Threadripper board though). The B350/B450 Mortar is one that only has heatsinks on half of the VRM. For audio I only want a nice clean headphone out, but with good driver support for surround sound etc. Quality on-board works better for me, or failing that a dedicated soundcard, rather than more boxes on the desk.

I've spent a lot of time in the Motherboards, but no one was singing any praises.
https://forums.overclockers.co.uk/threads/about-to-give-up-on-am4-matx-any-news.18822747/

I will see if I can find some happy (but not deluded) mATX owners in some of the Ryzen threads.

Most onboard sound chips will have you covered. If you want quality then you need spend money on head phones, DACs and maybe an amp to suit.

 

[Panos]

Mostly so I can reuse my 4 sticks of RAM and so I have the option of fitting a sound card if the on-board is pants. Sure I could buy new RAM and an external DAC, but then you lose the value for money argument.

Many motherboards don't like 4 RAM Sticks with Ryzen. Especially the B450 ones.

 

[IT Troll]

Most onboard sound chips will have you covered. If you want quality then you need spend money on head phones, DACs and maybe an amp to suit.

There is quite a lot of difference between budget on-board and the higher end boards.

Many motherboards don't like 4 RAM Sticks with Ryzen. Especially the B450 ones.

Yeah, believe me I have looked into this. There just doesn't seem to be a Ryzen solution for me which wouldn't cost more than going Intel.

 

[Vince]

Many motherboards don't like 4 RAM Sticks with Ryzen. Especially the B450 ones.

I'm running 8 sticks on gen 1 tr and it's been at 3200 cas 14 for around a week now following a thread here. I am almost sure that with strong b-die and a strong board (I have x399 Taichi and 8x8gb 3466 gskill trident z) that you could hit 3200 across 4 sticks. Does it take a bit more tuning and is a bit harder to hit? Sure does, but I bet most chips especially second gen ones with better imc should do that happily.

 

[DragonQ]

I seen it differently.

Things came to light in this video its clear that GN (and probably other reviewers) went to great length to maximise AMD's results in their reviews, something which I previously complained about. This interview is as much as about maintaining their own credibility as anything else.

Most reviewers got similar scores though, the PT ones are huge outliers. Remember when Anandtech got **** for getting very odd results that showed Intel in a bad light in certain games? It didn't match anyone else's results and was eventually discovered to be due to them using a non-default HPET setting on their systems.

This is very similar to PT using a non-standard 4c/8t configuration for their Ryzen test - it produces outlying results that are not credible.

One example is that he revealed in GN testing on ryzen2 chips he pushed fan's to 100% instead of AUTO, that I do not remember been disclosed in his reviews, and the average user is not going to do that, especially when noise is important. He openly admitted when explaining his reason in that it has impact on bench testing results.
Another thing I moaned about, is most of the reviewers were using intel stock configuration, against ryzen2 scores, which to me was wrong, as they were putting in effort to maximise ryzen2 performance but decided to not manually OC the K processors on comparison scores as well as disable MCE.

Huh? Firstly, he said he puts fan speeds to maximum for all tests. Whether you agree with this or not, it is at least fair.

Every credible reviewer I have seen makes it clear when they are using stock or overclocked parts. Intel's chips have had a lot more overclocking headroom (until recently anyway) so obviously they typically win those tests. Can you point out anyone that compared an overclocked Ryzen 2 against a stock Intel chip without making it clear?

MCE is an interesting one but IMO leaving it disabled on a stock configuration makes more sense than enabling it. It's a factory overclocking setting added by the motherboard manufacturer that doesn't take into account thermals or TDP. PBO is completely different because:
a) it's a documented and advertised feature by AMD;
b) it is typically enabled by default;
c) Ryzen is clever enough to only use it when there is thermal/TDP headroom.

PBO is analogous in this sense to Intel's normal Turbo Boost functionality and both should be enabled in "stock" configurations.

He raised the point about not using the same exact GPU for testing to remove variability, but I think unless PT deliberately gave the AMD rigs the worst binned GPUs I have no complaint, He mentioned that in the most extreme cases there is a 5% swing in performance, but if the test is not GPU bottlenecked for most of the test, then the impact on the test itself shouldnt be anywhere near 5%. I accept his point but I dont see it as a major one. The GPU silicon lottery e.g. would not cause the impact we seen on the graphs only probably 1-2% of it.

Probably true, this a minor point on his list I think.

On the cooler, his reply was rationale, he is imitating an average gamer not a high end gamer, and the AMD chips come with cooler so they get used, intel does not so an after market cooler "has" to be used. I accept that answer personally.

So you accept an answer that suggests emulating a "typical gamer" is more important than a fair test, whilst at the same time complaining about other reviewers' tests not being fair? What is your view on them disabling half of the R7 2700X cores? I'm pretty sure that isn't something a typical gamer would do, nor does it make for a fair comparison.

ON the ram,xmp issue. Steve had a point when asking why ram was XMP enabled but then downclocked but I also feel the answer returned was reasonable, the bit on the secondary and tertiary timings been board, chipset and bios dependent for reliability I feel steve wasnt been realistic, the vast majority of gamers wont be manually tuning timings. This video is really interesting as it shows how unrealistic reviewers can get. Info on this video that isnt disclosed in reviews. Also there is no rule stating that all reviewers should be like robots testing in the exact same way. e.g. lets say you got 10 AMD systems, they all run at 2933 ram speed, but 6 run at 3200, and 2 run at 3600. A reviewer like GN I expect would use the binned parts and post a test result at 3600 to show it in its best light, that to me is flawed, a better result is what ALL chips can do in that respect.

So why use XMP for the Intel machines?

 

[Panos]

I'm running 8 sticks on gen 1 tr and it's been at 3200 cas 14 for around a week now following a thread here. I am almost sure that with strong b-die and a strong board (I have x399 Taichi and 8x8gb 3466 gskill trident z) that you could hit 3200 across 4 sticks. Does it take a bit more tuning and is a bit harder to hit? Sure does, but I bet most chips especially second gen ones with better imc should do that happily.

True, but he has Hynix ram from gazillion years ago, that he has to use on mATX board, which are all B450s.
If he was considering X470 ATX/mITX board, or the X370 CH6 there shouldn't be a single consideration that the ram might not work.

I agree on good quality ram, since I had the CH6 with the 1800X back last year, with Samsung B 3600C16 ram and worked out of the box at rated speed. When at the same moment everyone else was crying cannot make their ram work with the Ryzen 1 systems.

 

[Vince]

True, but he has Hynix ram from gazillion years ago, that he has to use on mATX board, which are all B450s.
If he was considering X470 ATX/mITX board, or the X370 CH6 there shouldn't be a single consideration that the ram might not work.

I agree on good quality ram, since I had the CH6 with the 1800X back last year, with Samsung B 3600C16 ram and worked out of the box at rated speed. When at the same moment everyone else was crying cannot make their ram work with the Ryzen 1 systems.

This thread is clearly moving too fast for me as I missed that he is looking to use old slow ram on one of these new fangled chips. Even so I'm finding it really hard to believe that the Intel system comes in cheaper. You could probably build a tr 1950x cheaper than the 9900k at today's prices. Of course I missed all of the other odd bits as well like the b450 requirement.

Even taking the memory out of the equation I'm really struggling to believe it. It seems mad given the difference in processor cost alone gets you 16gb fast ddr4 with change.

 

[chrcoluk]

Believe he said in yesterdays video that they do that for all chips they test?

yeah it doesnt matter if intel or AMD, I am saying they shouldnt be doing that. Sorry if it sounded like I think it should just be an AMD thing.

 

[chrcoluk]

Most reviewers got similar scores though, the PT ones are huge outliers. Remember when Anandtech got **** for getting very odd results that showed Intel in a bad light in certain games? It didn't match anyone else's results and was eventually discovered to be due to them using a non-default HPET setting on their systems.

This is very similar to PT using a non-standard 4c/8t configuration for their Ryzen test - it produces outlying results that are not credible.

If a result is different it doesnt mean its not credible, it just means a different methodology may have been used.

There is variables that can affect results such as what windows updates are installed, if meltdown mitigation is enabled, and of course if effort has been put in to hide a products problems such as tweaking bios's and so forth.

Huh? Firstly, he said he puts fan speeds to maximum for all tests. Whether you agree with this or not, it is at least fair.

Disagree, if one product has bigger issues with thermal throttling than the other, then putting a fan to a non stock configuration that benefits one product over the other is not fair. Its only fair if the default config is 100%. Note e.g. most gpu's get tested on default fan curve's.

Every credible reviewer I have seen makes it clear when they are using stock or overclocked parts. Intel's chips have had a lot more overclocking headroom (until recently anyway) so obviously they typically win those tests. Can you point out anyone that compared an overclocked Ryzen 2 against a stock Intel chip without making it clear?

Not right now as I am a lazy sod, but if I remember right the "majority" of reviewers used non OC results in their graphs, I remember OC3D been one of the exceptions who did not.

Now my feeling is mixed on this one, if you specifically testing OOB, then you should not be overclocking intel manually, and thats fair enough. However in my opinion if you as a reviewer are fiddling with timings to get round ryzen issues, enabling bios to enhance performance on ryzen, and so forth, then you should also be overclocking the intel parts.

MCE is an interesting one but IMO leaving it disabled on a stock configuration makes more sense than enabling it. It's a factory overclocking setting added by the motherboard manufacturer that doesn't take into account thermals or TDP. PBO is completely different because:
a) it's a documented and advertised feature by AMD;
b) it is typically enabled by default;
c) Ryzen is clever enough to only use it when there is thermal/TDP headroom.

If AMD bios was all left at defaults, no special ram tuning to get round issues, I agree keep MCE at its default, but as I said above, it wasnt, it was tweaked.

So you accept an answer that suggests emulating a "typical gamer" is more important than a fair test, whilst at the same time complaining about other reviewers' tests not being fair? What is your view on them disabling half of the R7 2700X cores? I'm pretty sure that isn't something a typical gamer would do, nor does it make for a fair comparison.


So why use XMP for the Intel machines?

I accepted that reasoning for most of the stuff yes, only the 64gig of ram I found completely out of whack.

Using XMP probably should not have been done at all, I dont know why they even touched XMP, they should have just left JEDEC timings in place. I am guessing what happened they initially tried to use XMP/DOCP, then hit stability issues with the 64gig ram, and then manually turned down the clock speed of the memory leaving the timings in place, steve said this in his cooldown bit as well so he agrees with me I think.

Ironically I am a real life situation with such a config, I have 3200 XMP timings enabled on my ram, but if you see in my sig, I manually downclocked to 3000mhz (with those timings in place) to get stability.

 

[Panos]

I accepted that reasoning for most of the stuff yes, only the 64gig of ram I found completely out of whack.

That the 2700X was running as a 2400G was right for you then and not out of whack?

 

[chrcoluk]

That the 2700X was running as a 2400G was right for you then and not out of whack?

No as it was using AMD's gaming mode, I accepted his reason for enabling it plus he said it improved results for half of the games.

Enabling something called gaming code is not the same as going in a bios and purposely gimping something which the public seem to be implying he did.

 

[IT Troll]

True, but he has Hynix ram from gazillion years ago, that he has to use on mATX board, which are all B450s.

This thread is clearly moving too fast for me as I missed that he is looking to use old slow ram on one of these new fangled chips.

Is my RAM really that old and slow? To the extent that I would really notice it in real word use. It seems crazy be buying another 32GB of DDR4 now, given the current pricing of RAM and the fact that DDR5 will be along in a year or so.

32GB of RAM more that wipes out the difference in CPU cost (I have a £500 pre-order), then there would be a sound card or DAC, and a new cooler (if I went TR).

 

[Panos]

Is my RAM really that old and slow? To the extent that I would really notice it in real word use. It seems crazy be buying another 32GB of DDR4 now, given the current pricing of RAM and the fact that DDR5 will be along in a year or so.

32GB of RAM more that wipes out the difference in CPU cost (I have a £500 pre-order), then there would be a sound card or DAC, and a new cooler (if I went TR).

No your ram is Hynix. That's the issue trying to get a B450 mATX board. Could be hit/miss depending the board. X470 you won't have any issue with it.

 

[gavinh87]

No your ram is Hynix. That's the issue trying to get a B450 mATX board. Could be hit/miss depending the board. X470 you won't have any issue with it.

What tosh. It isn't the board, its the IMC.

 

[Combat squirrel]

Am I reading ocuk right when I see its a 8 core 8 thread cpu???? What??? May as well get 6/12 and have more threads!!??