learners permit
[H]ard|Gawd
- Joined
- Jun 15, 2005
- Messages
- 1,799
I run raid 0 on a pair of sabrent rocket nvme drives on X570. Definitely a pain to get setup but flawless since then and really fast writes particularly.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
AMD Ryzen 9 4950X "Vermeer" Tested, the Sample Boosts to 4.8 GHz
- Thread starter erek
- Start date
M
mls1995
Guest
I have 2x sabrent rocket nvme drives on X570 too but I'm not using RAID. Is there really a benefit? Seems like it would be more hassle than it's worth to me.
I'm very happy with my 3900x but I guess I could see myself upgrading along with a new video card (2080 to 3080 or whatever). I don't think it makes any financial sense but when has that ever stopped me before.
cjcox
2[H]4U
- Joined
- Jun 7, 2004
- Messages
- 2,937
While I do (now) generally prefer Dell, they aren't perfect. You can get a bad Broadcom and/or Intel NIC from Dell (of course, just as easily gotten from others as well)
IdiotInCharge
NVIDIA SHILL
- Joined
- Jun 13, 2003
- Messages
- 14,675
If you have to ask the question, the answer is generally 'no'.
The biggest benefit to faster storage is lower storage latency. Some of that comes from the far higher transfer rates of NVMe, especially as the size of data to be read increases, but most of that comes from using flash in the first place, and then from using native PCIe signalling instead of SATA for the protocol.
Where it would make a difference would be if your use case(s) involved very large reads and / or writes. Just note that the scale of tranfers needed to produce a measurable increase in performance over a base NVMe drive are generally pretty large!
IdiotInCharge
NVIDIA SHILL
- Joined
- Jun 13, 2003
- Messages
- 14,675
And we have / did, but generally I'll take a bad NIC over most any other part failing
Yet, the A0 revision also had an OPN.... Figure that one out. And zen2 had an OPN listed for engineering samples also. OPN is just an assigned code for the specific processor line. Not sure why you think this wouldn't be the case this time when this was the exact case last time.
Edit:.Just wanted to share link for zen2 ES with OPN...
https://www.techspot.com/amp/news/78452-amd-next-12-core-cpu-appears-benchmark-database.html
learners permit
[H]ard|Gawd
- Joined
- Jun 15, 2005
- Messages
- 1,799
Just off top my head as I remember reads were like 5200Mb/s and writes were 6400Mb/s in one of the disk tests. I would have to rerun the test though to confirm it as it's been several months ago.
Don't bother, he does the same thing every launch. He did it with Rome at the first leak, when the clocks were 1.4/2.2 or something similar, and not a peep when not even the 64 core parts were clocked that low, or anything close.
Rockenrooster
Gawd
- Joined
- Apr 11, 2017
- Messages
- 955
Well, he very clearly spells out he is Anti-AMD. Hard to believe anything he says about AMD...
Huh? 800mhz umm im done replying after that comment
My 3600x stablizes at 3875MHz all-core at 100% load in a stress test like prime95 small fft.
Oh I know, I just like to point it out that way anyone else reading isn't mislead... I don't expect he'll change his ways, I just don't want it to spread
.Anyways, a 5950x hitting 4.8ghz is good news either way, especially if it's anywhere close to a 15% IPC on the architecture uplift. I am hoping they are creating the new naming scheme to fix their mistake on the APU naming so the APU and normal CPU will be the same architect as each other. If they name this series 5xxx the next APU to come out will HOPEFULLY keep the 5xxx branding and no more confusing naming.
IdiotInCharge
NVIDIA SHILL
- Joined
- Jun 13, 2003
- Messages
- 14,675
AMDs boost clocks are not Intel's boost clocks -- that's the biggest issue with reports of higher boost clocks on AMD parts and making a comparison IMO, unfortunately, and it really doesn't tell us if AMD will close the single-thread gap with the release of Zen 3 CPUs.
cybereality
[H]F Junkie
- Joined
- Mar 22, 2008
- Messages
- 8,789
I got better results using AMD's default boost mode, than with manual overclocking.
I don't recall the exact numbers, it was on a 2700X machine and I believe I got to 4.6 GHz (or was it 4.7) all core (with a lot of juice).
Which sounds great on paper, but the machine actually ran worse. Some synthetic benchmarks did get a boost, but for most tasks, including gaming, it was a wash and not worth the heat/cooling issues.
So I think the default boost technology does work well.
I don't recall the exact numbers, it was on a 2700X machine and I believe I got to 4.6 GHz (or was it 4.7) all core (with a lot of juice).
Which sounds great on paper, but the machine actually ran worse. Some synthetic benchmarks did get a boost, but for most tasks, including gaming, it was a wash and not worth the heat/cooling issues.
So I think the default boost technology does work well.
Yeah, I know. But note that at least that is above the rated base clock, and with one of the toughest torture tests, so it's not too bad at least. I do wish it ran faster, of course.
That's what the flash memory used on NVMe excels at.
Broadcrap NICs are something I try to avoid. I've dealt with horrendous Dell / Broadcom related issues that far outstrip any of the bullshit I've endured with HP's crappy storage controllers. That said, I prefer Dell servers, just with Intel NICs.
I doubt it. Ryzen and its descendants simply don't clock as well as Intel's architectures do. We've seen this time and time again.
So, 11th gen is moving to pcie 4.0, does that include the DMI link to the PCH? Seems this has been an issue for a long time with Intel (RAID 0 with 2 NVME drives sharing the pcie 3.0 x4 lanes with other things as well). I guess they want you to move up to workstation grade, but they don't even have much for you there.
Well, if they are getting 4.8ghz (nobody knows how the different SKUs will boost yet) and that rumoured IPC increase is anywhere near 10%+, I think it will help close that last bit of gap. Hopefully they can get the latency issues resolved, we saw what difference a single ccx did with the 3300x vs 3100... Huge difference even at identical clocks. I'm hopeful but not banking on it
.learners permit
[H]ard|Gawd
- Joined
- Jun 15, 2005
- Messages
- 1,799
I think that will make the 4800X the the perfect desktop core to have for gaming and general purpose stuff at least for me anyway.
So, 11th gen is moving to pcie 4.0, does that include the DMI link to the PCH? Seems this has been an issue for a long time with Intel (RAID 0 with 2 NVME drives sharing the pcie 3.0 x4 lanes with other things as well). I guess they want you to move up to workstation grade, but they don't even have much for you there.
Well, if they are getting 4.8ghz (nobody knows how the different SKUs will boost yet) and that rumoured IPC increase is anywhere near 10%+, I think it will help close that last bit of gap. Hopefully they can get the latency issues resolved, we saw what difference a single ccx did with the 3300x vs 3100... Huge difference even at identical clocks. I'm hopeful but not banking on it
Honestly, I have no way of knowing at this time. Just because the PCIe controller in the chipset is PCIe 4.0 capable doesn't mean that the DMI link will be, or that the chipset will be fully enabled as PCIe 4.0 compatible. I've heard only one rumor that Intel will still use DMI 3.0, but that the link count is going to double. I don't know how true this is as I've not been able to verify or disprove this in any way shape or form. We'll have to wait until some more definitive information comes out to know for sure.
I use to use it all the time with spinning rust. The only problem I have encountered was the 10k raptor drives themselves taking a shit. With how fast SSD are now it makes little sense to do.
How does latency affect word and such? I don't fully understand ram stats like that, but am genuinely curious how it affects real world work (rather than just tuning for a benchmark because more is better). I just got some new ram for my ryzen so hoping to tune it.
I always ran RAID 0 arrays for my OS and games. I run several spinning disks in a RAID 1 array for redundancy. That's about all I do with it now. I no longer bother on the OS and games drives due to the speed of even middle of the road NVMe SSD's.
MrCaffeineX
[H]ard|Gawd
- Joined
- Aug 22, 2011
- Messages
- 1,604
I concur with Dan_D. I've migrated plenty of workstations that were using Intel RAID1, 5, and 10 to newer motherboards without issue. I've also broken individual disks out of Intel RAID1 arrays and tossed them into other Intel-based systems just to see what they would do and they worked fine after Sysprepping to clear out the duplicate names/ids/etc.
I can't imagine they can double the DMI link on the z490 if it wasn't built in to start, but if it's pcie 4.0 compatible maybe it can run at 2x bandwidth like AMD has. Something tells me this won't be the case and it'll just ungrade to have pcie to the CPU links (aka, GPU). Who knows though without more info. I feel they are just making a faux pass at pcie 4.0 just to hold over their investors and to be able to say they have it also until pcie 5.0 comes around.
I believe the rumor was concerning another chipset or variant, but I could be mistaken. As I said I hadn't verified any of that so I can't speak authoritatively towards the validity of it. And, if I did have official information about it, I'd probably be under NDA anyway.
Yeah, I hear you, I'm just saying if people are buying now they are probably stuck with what they have. Another chipset it's possible they may upgrade the DMI link finally, either 3.0 x8 or 4.0 x4 (basically doubling bandwidth one way or another) and be on par with AMD finally. Really this only affects a minority of people and work loads, but hey
. My server runs RAID 10, my desktop just has a single NVME and then 2 spinners in RAID 0. Seems odd that AMD has had 24 lanes (20 direct to cpu, 4 for chipset link) for so long and Intel is still sitting with 16 + DMI x4. I was hoping the 10 series was going to fix this deficiency even if it was still 3.0, but alas... still waiting.ps. Yeah, NDA's are fun, I haven't had to sign one in a long time (when it was still ATI). I used to get access to engineering hardware when programmable shaders were first coming out (both nvidia and AMD). It's fun putting in a GPU that is the model AFTER the one that's in pre-sale, haha. Not so much anymore though.
Rockenrooster
Gawd
- Joined
- Apr 11, 2017
- Messages
- 955
4.9GHz? https://videocardz.com/newz/amd-16-core-zen3-ryzen-9-4950x-engineering-sample-boosts-up-to-4-9-ghz
4% clock bump + 15% IPC + improved latency seems like a pretty decent upgrade
4% clock bump + 15% IPC + improved latency seems like a pretty decent upgrade
IdiotInCharge
NVIDIA SHILL
- Joined
- Jun 13, 2003
- Messages
- 14,675
Mostly, it's not much of a deficiency beyond spec-sheet comparisons. Even if you have the drives attached, you're still just talking about peak transfer rates, which you'd need to be able to use continuously at those speeds to really make a difference in some workload completion time.
And then you'd have to make the case for using consumer instead of HEDT / workstation hardware
IdiotInCharge
NVIDIA SHILL
- Joined
- Jun 13, 2003
- Messages
- 14,675
If they can get the sustained all-core speeds well above the middling 4.0 to 4.2 that they currently top out at, that'd be about right for me. Then I just need to find a pair of 32GB DDR4-3600 C16 DIMMs that'll run at speed...
German Muscle
Supreme [H]ardness
- Joined
- Aug 2, 2005
- Messages
- 6,943
You guys ready to see AMD CPUs cost nearly as much as Intels? I sure am. They cant/wont keep their prices at the floor when the chips performance is at the ceiling.
Axman
VP of Extreme Liberty
- Joined
- Jul 13, 2005
- Messages
- 17,307
They probably can't raise prices to match Intel, especially not with OEMs. Intel can play the rebate game with OEMs for years to come. They will build a bridge between their currently flagging architecture to a future good architecture with money.
The only way this doesn't happen--and I wouldn't count it out entirely--is if Intel doesn't succeed at making a good product in the future. They aren't in the same position as they were in 15 years ago with a backup chip design team, not that I have heard of, anyway. Unless you count Atom, but I don't think Atom is another potential successor the way Core supplanted Pentium.
whateverer
[H]ard|Gawd
- Joined
- Nov 2, 2016
- Messages
- 1,815
Yeah, especially when it's also expected to see up-to 10% performance boost from going single CCX. That's the magic behind the 3300X
With that much of a clock boost,Tiger Lake just became irrelevant.
Yeah, I agree, most people would never notice. The only time a normal person would even notice is when transferring (sequenctial) data from 1 nvme to another nvme. You benchmark and get 3500MB/s on both drives individually, but get less than 2,000MB/s (which, lets be honest is still ridiculously fast). It's just for those trying to push the limits or are on the cusp of a HEDT build but aren't ready to make the leap yet. That being said, I was hoping for some sort of advancement... DMI 2.0 had 4GB/s connection back in 1155 (3rd series intel chips).... they "upgraded" to DMI 3.0 with Skylake, which brought the bandwidth down to 3.93GB/s.... here we are over 11 years later still stuck with the same PCH bandwidth (actually, it went slightly down). I guess if it ain't broke don't fix it. If 99% of the population wouldn't see a difference, not much point in changing things, but people running 10900k's have not point in trying to run an NVME RAID is just... dissapointing even if mostly academic and only really going to be noticeable in a benchmark.Mostly, it's not much of a deficiency beyond spec-sheet comparisons. Even if you have the drives attached, you're still just talking about peak transfer rates, which you'd need to be able to use continuously at those speeds to really make a difference in some workload completion time.
And then you'd have to make the case for using consumer instead of HEDT / workstation hardware
atp1916
[H]ard|DCoTM x1
- Joined
- Jun 18, 2004
- Messages
- 5,004
So it's ok for Intel to normalize the industry at said price/perf tiering for what, a solid decade and then some - and now all the sudden it's a huge red flag / issue if their competitor possibly follows suite?
I could only hope AMD does such a thing. At least we'd have more than 4c/8T / PCIe 3.0 / DDR4.
Yeah, welcome to Intel fanboism
. If Intel does it, it's fine, if people even think AMD might do it, they're evil. Like it's AMD's sole purpose in life to keep Intel prices cheap. Even then their prices went down, for example 1700x was $399 to 3700x was $349. If the 4700x (5700x?) is back up to $399 and runs a good bit faster than the 3700x... it'll probably be sold out for the first 3-6 months. If the price is $349, it'll still be sold out. What sense does it make to not sell it for as much as you can? It's not a charity just as Intel isn't a charity.
SPARTAN VI
[H]F Junkie
- Joined
- Jun 12, 2004
- Messages
- 8,762
Assuming it's still on DDR4, I wonder what FCLK/MCLK we're looking at to maintain that 1:1 ratio. DDR4-4000?
IdiotInCharge
NVIDIA SHILL
- Joined
- Jun 13, 2003
- Messages
- 14,675
It's not chip designs that are the issue -- it's the fabs to produce them. AMD was lucky to partner with TSMC, who had the breakthrough that Intel (and GloFo and Samsung and...) haven't yet made. Otherwise AMD will still be at pre-Skylake levels of performance.
Now, supposing that Intel does get their fab tech online, AMD and TSMC will have a fight on their hands, and based on the respective companies' history (that is, Intel and TSMC), it's more likely for Intel to fix their fabs than for TSMC to continue to lead.
Well AMD kinda already did that to motherboards for high end. I didn't think I would buy an AMD board thats 370 lol. But I did. They reset that market because they have good CPUs. But I honestly don't see them raising pricing on the CPUs to intel's level. You will likely see them stick to what they have + 50 may be. But They already have increased their margins pretty damn good so I highly doubt they need to match intel with pricng. Intel has just been little greedy without competition.
You say that, as Intel releases information about them being a year behind on 7nm (although, they said the 1 year behind is only going to affect them by 6 months?). Going by recent Intel mis-steps, I think it's going to take a little bit longer to get themselves back into gear.
IdiotInCharge
NVIDIA SHILL
- Joined
- Jun 13, 2003
- Messages
- 14,675
I say that based on established history, because Intel has been lying to themselves first all these years. I don't expect to get concrete information out of them until they have shipable products in the pipeline.