Jump to content
Search In
  • More options...
Find results that contain...
Find results in...
VegetableStu

The Core Wars: Endgame (2019) (10980XE vs 3960X/3970X)

link related: a feature overview of every announced TRX40 motherboard

https://www.anandtech.com/show/15121/the-amd-trx40-motherboard-overview-

Recommended Posts

6 minutes ago, VegetableStu said:

a Ryzen with RAM on the IF topology next to the chiplet O_O

It's actually in the chiplet die itself. Each CCX has 32MB L3 cache. 

 

I do think an L4 cache of eDRAM would be interesting though. Slap 256mb of that stuff on the package and watch as trips down memory lane dwindle even further. 


R9 3900x; 64GB RAM | RTX 2080 | 1.5TB Optane P4800x

1TB ADATA XPG Pro 8200 SSD | 2TB Micron 1100 SSD
HD800 + SCHIIT VALI | Topre Realforce Keyboard

Link to post
Share on other sites
13 minutes ago, comander said:

It's actually in the chiplet die itself. Each CCX has 32MB L3 cache. 

 

I do think an L4 cache of eDRAM would be interesting though. Slap 256mb of that stuff on the package and watch as trips down memory lane dwindle even further. 

 

I think he was speculating on possabilities. And yeah throwing a decent amount of really high speed on chip memory at things would doubtless have some interesting effects.

Link to post
Share on other sites
8 minutes ago, comander said:

It's actually in the chiplet die itself. Each CCX has 32MB L3 cache. 

Each CCD has 32MB of ram, each CCX has 16MB. Which is still a lot compared to historic quantities. Guess 7nm enables that for you.

 

8 minutes ago, comander said:

I do think an L4 cache of eDRAM would be interesting though. Slap 256mb of that stuff on the package and watch as trips down memory lane dwindle even further. 

While I think L4 could be interesting, I think it has to be a lot bigger than 256MB. In context, that's hardly bigger than the combined L3 at 128MB. Where it might offer best benefit is if has sufficient bandwidth to feed all dies at once, and enable on-package sharing of data. I understand currently one CCX can't directly access another CCX's cache. You have to go back to ram for data even if it is otherwise on package. L4 could then mitigate the limited bandwidth of ram for bigger work allowing all cores more easy access to the same data even if off CCX.

 

Note Zen 3 rumour is L3 cache could become shared within CCD, not only at CCX level as current. Doesn't directly help with inter-CCD sharing though.


Main rig: Asus Maximus VIII Hero, i7-6700k stock, Noctua D14, G.Skill Ripjaws V 3200 2x8GB, Gigabyte GTX 1650, Corsair HX750i, In Win 303 NVIDIA, Samsung SM951 512GB, WD Blue 1TB, HP LP2475W 1200p wide gamut

Gaming system: Asrock Z370 Pro4, i7-8086k stock, Noctua D15, Corsair Vengeance LPX RGB 3000 2x8GB, Gigabyte RTX 2070, Fractal Edison 550W PSU, Corsair 600C, Optane 900p 280GB, Crucial MX200 1TB, Sandisk 960GB, Acer Predator XB241YU 1440p 144Hz G-sync

Ryzen rig: Asrock B450 ITX, R5 3600, Noctua D9L, G.SKill TridentZ 3000C14 2x8GB, Gigabyte RTX 2070, Corsair CX450M, NZXT Manta, WD Green 240GB SSD, LG OLED55B9PLA

VR rig: Asus Z170I Pro Gaming, i7-6700T stock, Scythe Kozuti, Kingston Hyper-X 2666 2x8GB, Zotac 1070 FE, Corsair CX450M, Silverstone SG13, Samsung PM951 256GB, Crucial BX500 1TB, HTC Vive

Gaming laptop: Asus FX503VD, i5-7300HQ, 2x8GB DDR4, GTX 1050, Sandisk 256GB + 480GB SSD

Total CPU heating: i7-8086k, i3-8350k, i7-7920X, 2x i7-6700k, i7-6700T, i5-6600k, i3-6100, i7-5930k, i7-5820k, i7-5775C, i5-5675C, 2x i7-4590, i5-4570S, 2x i3-4150T, E5-2683v3, 2x E5-2650, E5-2667, R7 3700X, R5 3600, R5 2600, R7 1700

Link to post
Share on other sites
6 minutes ago, porina said:

Each CCD has 32MB of ram, each CCX has 16MB. Which is still a lot compared to historic quantities. Guess 7nm enables that for you.

 

While I think L4 could be interesting, I think it has to be a lot bigger than 256MB. In context, that's hardly bigger than the combined L3 at 128MB. Where it might offer best benefit is if has sufficient bandwidth to feed all dies at once, and enable on-package sharing of data. I understand currently one CCX can't directly access another CCX's cache. You have to go back to ram for data even if it is otherwise on package. L4 could then mitigate the limited bandwidth of ram for bigger work allowing all cores more easy access to the same data even if off CCX.

 

Note Zen 3 rumour is L3 cache could become shared within CCD, not only at CCX level as current. Doesn't directly help with inter-CCD sharing though.

 

I believe AMD themselves have allready confirmed zen3 will be moving from 2 4 core CCX's per chiplet to 1 8 core CCX per chiplet.

Link to post
Share on other sites
2 hours ago, VegetableStu said:

btw:

 

2019-11-25-image-4.jpg

do not get "Value HEDT for Content Creators". the NUMA improvements are worth more than the price difference

AMD deserve a medal for bringing "Mainstream gaming for $149" back again. :)

I got my Pheni x3 750~ for around £140, and even unlocked a core and overclocked it, it lasted 5 years... been on this second hand 3470 (from a 3440 or whatever) as an upgrade for around the same time too. Why? It cost me like £50... and anything "better" was clocking in at £200 and a new motherboard (sometimes deals for less, but for a whole new rig, and 10% FPS/performance improvement, I just could not be bothered).

 

Finally, something sub £200, AND a massive upgrade in the Ryzen 5 series. :)

Link to post
Share on other sites
Link to post
Share on other sites
9 minutes ago, CarlBar said:

I believe AMD themselves have allready confirmed zen3 will be moving from 2 4 core CCX's per chiplet to 1 8 core CCX per chiplet.

I hadn't seen that. That's why I was putting it into the rumour bucket. Got a reference?


Main rig: Asus Maximus VIII Hero, i7-6700k stock, Noctua D14, G.Skill Ripjaws V 3200 2x8GB, Gigabyte GTX 1650, Corsair HX750i, In Win 303 NVIDIA, Samsung SM951 512GB, WD Blue 1TB, HP LP2475W 1200p wide gamut

Gaming system: Asrock Z370 Pro4, i7-8086k stock, Noctua D15, Corsair Vengeance LPX RGB 3000 2x8GB, Gigabyte RTX 2070, Fractal Edison 550W PSU, Corsair 600C, Optane 900p 280GB, Crucial MX200 1TB, Sandisk 960GB, Acer Predator XB241YU 1440p 144Hz G-sync

Ryzen rig: Asrock B450 ITX, R5 3600, Noctua D9L, G.SKill TridentZ 3000C14 2x8GB, Gigabyte RTX 2070, Corsair CX450M, NZXT Manta, WD Green 240GB SSD, LG OLED55B9PLA

VR rig: Asus Z170I Pro Gaming, i7-6700T stock, Scythe Kozuti, Kingston Hyper-X 2666 2x8GB, Zotac 1070 FE, Corsair CX450M, Silverstone SG13, Samsung PM951 256GB, Crucial BX500 1TB, HTC Vive

Gaming laptop: Asus FX503VD, i5-7300HQ, 2x8GB DDR4, GTX 1050, Sandisk 256GB + 480GB SSD

Total CPU heating: i7-8086k, i3-8350k, i7-7920X, 2x i7-6700k, i7-6700T, i5-6600k, i3-6100, i7-5930k, i7-5820k, i7-5775C, i5-5675C, 2x i7-4590, i5-4570S, 2x i3-4150T, E5-2683v3, 2x E5-2650, E5-2667, R7 3700X, R5 3600, R5 2600, R7 1700

Link to post
Share on other sites
1 hour ago, comander said:

This is what happens when you only really need 8 threads and each thread gets 32mb cache. 

 

Haha. Basically 0 trips to the RAM. 

 

I'm speculating of course. I don't know if the os is smart enough to go that. 

We think the Tomb Raider result is due to the 4ch memory removing a bottleneck in the specific area that Steve tests in. It's a heavy CPU demanding area, compared to the canned benchmark, so it's very possible the heavy CPU load is some sort of load/deload scripts running in the background. Others have gotten the CPUs right around each other, so the little bit to get over the top could have been the memory choice or motherboard auto-OC features. 

Link to post
Share on other sites
20 minutes ago, porina said:

I hadn't seen that. That's why I was putting it into the rumour bucket. Got a reference?

And AMD engineer gave a talk at a HPC conference in the UK. Entire thing got put on YT, and a bunch of us watched it. It got taken down, but we know that AMD had samples back in the office by early Summer. We also know they're moving to 8c CCX or no-CCX structure. (Single alignment per chiplet.) Lastly, we know that the DDR5 platform will be called SP5 and if you had requests about it, you could still give AMD input.

Link to post
Share on other sites

No one going to complain here that AMD offered the same core count at noticeably higher costs this generation of threadripper? I mean if this was intel or nvidia the forums would be screaming about it.

 

Anyways, looks cool. That interconnect power consumption is insane though, and dx86 and avx512 workloads still show their massive performance uplifts. 

 

Good job AMD (I do mean that). Looking forward to Zen3


LINK-> Kurald Galain:  The Night Eternal 

Top 5820k, 980ti SLI Build in the World*

CPU: i7-5820k // GPU: SLI MSI 980ti Gaming 6G // Cooling: Full Custom WC //  Mobo: ASUS X99 Sabertooth // Ram: 32GB Crucial Ballistic Sport // Boot SSD: Samsung 850 EVO 500GB

Mass SSD: Crucial M500 960GB  // PSU: EVGA Supernova 850G2 // Case: Fractal Design Define S Windowed // OS: Windows 10 // Mouse: Razer Naga Chroma // Keyboard: Corsair k70 Cherry MX Reds

Headset: Senn RS185 // Monitor: ASUS PG348Q // Devices: Galaxy S9+ - XPS 13 (9343 UHD+) - Samsung Note Tab 7.0 - Lenovo Y580

 

LINK-> Ainulindale: Music of the Ainur 

Prosumer DYI FreeNAS

CPU: Xeon E3-1231v3  // Cooling: Noctua L9x65 //  Mobo: AsRock E3C224D2I // Ram: 16GB Kingston ECC DDR3-1333

HDDs: 4x HGST Deskstar NAS 3TB  // PSU: EVGA 650GQ // Case: Fractal Design Node 304 // OS: FreeNAS

 

 

 

Link to post
Share on other sites
20 minutes ago, Curufinwe_wins said:

No one going to complain here that AMD offered the same core count at noticeably higher costs this generation of threadripper? I mean if this was intel or nvidia the forums would be screaming about it.

 

Anyways, looks cool. That interconnect power consumption is insane though, and dx86 and avx512 workloads still show their massive performance uplifts. 

 

Good job AMD (I do mean that). Looking forward to Zen3

While I can understand what you are saying, the benchmarks don't lie and it's clear that even tho you give up a few cores, it still destroys its higher-core predecessors.

 

Also as I think linus or GN said, first and second gen threadripper weren't flawless. Especially the second gen WX cpu's were a more hit-and-miss because it all came down to the workload. However this generation feels like they finally got rid of all those flaws and it really looks like a polished platform from the start. (I guess launch would be a better word but eh). This time in workloads that don't scale well it can keep up very well and workloads that should scale well do actually scale very well. And not just a few but basically all of them as it should be.

 

 


If you want my attention, quote meh! D: or just stick an @samcool55 in your post :3

Spying on everyone to fight against terrorism is like shooting a mosquito with a cannon

Link to post
Share on other sites
29 minutes ago, Curufinwe_wins said:

No one going to complain here that AMD offered the same core count at noticeably higher costs this generation of threadripper? I mean if this was intel or nvidia the forums would be screaming about it.

 

Anyways, looks cool. That interconnect power consumption is insane though, and dx86 and avx512 workloads still show their massive performance uplifts. 

 

Good job AMD (I do mean that). Looking forward to Zen3

Yes, but 3950x does nearly everything the 10980XE does. That's what the 3950x is for. Threadripper is in a league of it's own, and it's not targeted at us in particular. It's targeted at people who don't care about the price and where as Linus said, time is money.

 

And the 3960x and 3970x tend to offer more performance in those multicore workloads than maybe people would have been expecting. Rather than the 25-30% higher speeds we'd expect over the 2990WX, we're seeing 50%+. And a lot of the issues that 2nd gen Threadripper had are gone. I think that deserves a price bump, especially given that it's targeted at those people where price is irrelevant.

Link to post
Share on other sites

That MSpaint threadripper. what does those "finger guns" mean?


Intel Xeon E5 1650 v3 @ 3.5GHz 6C:12T / CM212 Evo / Asus X99 Deluxe / 16GB (4x4GB) DDR4 3000 Trident-Z / Samsung 850 Pro 256GB / Intel 335 240GB / WD Red 2 & 3TB / Antec 850w / RTX 2070 / Win10 Pro x64

HP Envy X360 15: Intel Core i5 8250U @ 1.6GHz 4C:8T / 8GB DDR4 / Intel UHD620 + Nvidia GeForce MX150 4GB / Intel 120GB SSD / Win10 Pro x64

 

HP Envy x360 BP series Intel 8th gen

AMD ThreadRipper 2!

5820K & 6800K 3-way SLI mobo support list

 

Link to post
Share on other sites
15 minutes ago, Curufinwe_wins said:

No one going to complain here that AMD offered the same core count at noticeably higher costs this generation of threadripper? I mean if this was intel or nvidia the forums would be screaming about it.

 

Anyways, looks cool. That interconnect power consumption is insane though, and dx86 and avx512 workloads still show their massive performance uplifts. 

 

Good job AMD (I do mean that). Looking forward to Zen3

How do you figure? If I remember correctly the 32 and 24 core 2nd gen threadrippers launched at similar prices and with this you are getting a significant jump in performance and 2x the amount of L3 cashe. Yeah you go from 1800 to 2000 but in the grand scheme of things 200 dollars at that price point is not all that much and it definitely has the performance increase to warrant its price. I find it weird that people would complain about getting a 32 core cpu that is good at just about anything for only 2000. That is a steal tbh and to say the price is unfair would be kinda ridiculous. I think people forget that you haven't been able to get anything like this for that price. 

Link to post
Share on other sites
3 minutes ago, Brooksie359 said:

How do you figure? If I remember correctly the 32 and 24 core 2nd gen threadrippers launched at similar prices and with this you are getting a significant jump in performance and 2x the amount of L3 cashe. Yeah you go from 1800 to 2000 but in the grand scheme of things 200 dollars at that price point is not all that much and it definitely has the performance increase to warrant its price. I find it weird that people would complain about getting a 32 core cpu that is good at just about anything for only 2000. That is a steal tbh and to say the price is unfair would be kinda ridiculous. I think people forget that you haven't been able to get anything like this for that price. 

Every other company gets lambasted on this forum for producing next generation equivalent products in their stack at the same price, let alone higher. Regardless of how much more performance there is along side it. Just pointing that out. 

 

It's a great product. I've shared the concerns I have with numa awareness, and the interconnect power scaling being huge (hilarious that people just a few days ago slammed the shit out of Ryan Shrout for daring to say that other parts of the CPU needed optimizing as well, and that interconnect optimizations would be a key path moving forward. Then a 100W interconnect chiplet comes around driving max multicore power budget down to 3-5W per core. Interconnect power is now almost 40% of total budget.) 

 

 

Other than that, (and even with it) it's great. As I said looking forward to Zen3.  Just wanted to give those caveats. The praise for the chip is stated much better in other places, so I didnt bother repeating it. Sorry that that made it seem like I wasn't very positive on it overall.


LINK-> Kurald Galain:  The Night Eternal 

Top 5820k, 980ti SLI Build in the World*

CPU: i7-5820k // GPU: SLI MSI 980ti Gaming 6G // Cooling: Full Custom WC //  Mobo: ASUS X99 Sabertooth // Ram: 32GB Crucial Ballistic Sport // Boot SSD: Samsung 850 EVO 500GB

Mass SSD: Crucial M500 960GB  // PSU: EVGA Supernova 850G2 // Case: Fractal Design Define S Windowed // OS: Windows 10 // Mouse: Razer Naga Chroma // Keyboard: Corsair k70 Cherry MX Reds

Headset: Senn RS185 // Monitor: ASUS PG348Q // Devices: Galaxy S9+ - XPS 13 (9343 UHD+) - Samsung Note Tab 7.0 - Lenovo Y580

 

LINK-> Ainulindale: Music of the Ainur 

Prosumer DYI FreeNAS

CPU: Xeon E3-1231v3  // Cooling: Noctua L9x65 //  Mobo: AsRock E3C224D2I // Ram: 16GB Kingston ECC DDR3-1333

HDDs: 4x HGST Deskstar NAS 3TB  // PSU: EVGA 650GQ // Case: Fractal Design Node 304 // OS: FreeNAS

 

 

 

Link to post
Share on other sites

It's not a war when one side has 10,000 tanks, 10,000 planes, 5,000 subs and 1,000,000 troops while the other has nothing but civilians who've been called up for national service. That would be called a massacre.

 

Intel are currently relying on national service and their population is elderly and slowly becoming arthritic.

 


Main Rig:-

Ryzen 7 3800X | Asus ROG Strix X570-F Gaming | 16GB Team Group Dark Pro 3600Mhz | Samsung 970 Evo 500GB NVMe | Sapphire 5700 XT Pulse | Corsair H115i Platinum | WD Black 1TB | WD Green 4TB | EVGA SuperNOVA G3 650W | Asus TUF GT501 | Samsung C27HG70 1440p 144hz HDR FreeSync 2 | Windows 10 Pro X64 |

 

Server:-

Raspberry Pi 4 Model B running OMV Arrakis and an 8TB Seagate USB 3.0 external HDD

Link to post
Share on other sites

Aside from sword waving and halo products, HEDT doesn't really make sense anymore for 95% of users. 

 

Sad in a way, it's completely sucked any interest I had in it. 


Current PC:

Spoiler

*WORK IN PROGRESS*

 

Mothballed PC:

Spoiler

 

CPU: Intel i5 4690k Cooler: Corsair H100i V2 Motherboard: MSI Z97i AC ITX

RAM: Crucial Ballistix 16GB DDR3 Storage: Kingston Fury 240GB GPU: Asus Strix GTX 970

PSU: Thermaltake TR2 Case: Phanteks Enthoo Evolv ITX

Monitor: Dell P2214H x2 Mouse: Logitech MX Master Keyboard: G.Skill KM780 Cherry MX Red

 

 

Link to post
Share on other sites
41 minutes ago, Curufinwe_wins said:

 Sorry that that made it seem like I wasn't very positive on it overall.

Sometimes when you point out the hypocrisy of the forum, people ironically conclude you are being negative toward a product they were admiring with said bias.

 

Go figure.


QuicK and DirtY. Read the CoC it's like a guide on how not to be moron.  Also I don't have an issue with the VS series.

Sometimes I miss contractions like n't on the end of words like wouldn't, couldn't and shouldn't.    Please don't be a dick,  make allowances when reading my posts.

Link to post
Share on other sites
8 minutes ago, mr moose said:

Sometimes when you point out the hypocrisy of the forum, people ironically conclude you are being negative toward a product they were admiring with said bias.

 

Go figure.

I mean it's a natural human reaction (even if unfortunate), but regardless I could have probably stated it better originally as well.

 

**shrugs**


LINK-> Kurald Galain:  The Night Eternal 

Top 5820k, 980ti SLI Build in the World*

CPU: i7-5820k // GPU: SLI MSI 980ti Gaming 6G // Cooling: Full Custom WC //  Mobo: ASUS X99 Sabertooth // Ram: 32GB Crucial Ballistic Sport // Boot SSD: Samsung 850 EVO 500GB

Mass SSD: Crucial M500 960GB  // PSU: EVGA Supernova 850G2 // Case: Fractal Design Define S Windowed // OS: Windows 10 // Mouse: Razer Naga Chroma // Keyboard: Corsair k70 Cherry MX Reds

Headset: Senn RS185 // Monitor: ASUS PG348Q // Devices: Galaxy S9+ - XPS 13 (9343 UHD+) - Samsung Note Tab 7.0 - Lenovo Y580

 

LINK-> Ainulindale: Music of the Ainur 

Prosumer DYI FreeNAS

CPU: Xeon E3-1231v3  // Cooling: Noctua L9x65 //  Mobo: AsRock E3C224D2I // Ram: 16GB Kingston ECC DDR3-1333

HDDs: 4x HGST Deskstar NAS 3TB  // PSU: EVGA 650GQ // Case: Fractal Design Node 304 // OS: FreeNAS

 

 

 

Link to post
Share on other sites
14 minutes ago, Curufinwe_wins said:

I mean it's a natural human reaction (even if unfortunate), but regardless I could have probably stated it better originally as well.

 

**shrugs**

Maybe, but I don't think there is a way to tell someone they are biased without that bias effecting their ability to hear it.

 

 


QuicK and DirtY. Read the CoC it's like a guide on how not to be moron.  Also I don't have an issue with the VS series.

Sometimes I miss contractions like n't on the end of words like wouldn't, couldn't and shouldn't.    Please don't be a dick,  make allowances when reading my posts.

Link to post
Share on other sites
55 minutes ago, Curufinwe_wins said:

I mean it's a natural human reaction (even if unfortunate), but regardless I could have probably stated it better originally as well.

 

**shrugs**

(14:21 if the time stamp thing doesn't work)

 

The internet is a fun place. 🙃

Link to post
Share on other sites
4 hours ago, porina said:

While I think L4 could be interesting, I think it has to be a lot bigger than 256MB. In context, that's hardly bigger than the combined L3 at 128MB. Where it might offer best benefit is if has sufficient bandwidth to feed all dies at once, and enable on-package sharing of data. I understand currently one CCX can't directly access another CCX's cache. You have to go back to ram for data even if it is otherwise on package. L4 could then mitigate the limited bandwidth of ram for bigger work allowing all cores more easy access to the same data even if off CCX.

Fair warning, my thought process was more for 8 or 16C parts. The amount would certainly need to be higher for higher core count variants with more internal cache. 


R9 3900x; 64GB RAM | RTX 2080 | 1.5TB Optane P4800x

1TB ADATA XPG Pro 8200 SSD | 2TB Micron 1100 SSD
HD800 + SCHIIT VALI | Topre Realforce Keyboard

Link to post
Share on other sites
8 hours ago, Curufinwe_wins said:

Every other company gets lambasted on this forum for producing next generation equivalent products in their stack at the same price

 

But it's not equivalent, it's not even close to equivalent. If this only had a 15-20% performance gain over the previous 32 core part you probably would see people giving AMD a hard time. But when 50%+ is happening with alarming regularity people don;t really care because whilst it may only be 32 cores still it's performing like it has a lot more cres than it actually does.

Link to post
Share on other sites

I want to see Intel cherry pick some tests and claims its 10980XE is "UP TO" 65% better than TR 3960X/3970X.

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now


×