Same. I just don't like to see a lot of meat left on the bone, lol. 80C in a stess test on air seems.......limiting.
-
Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
Arrow Lake leaks + post-launch discussion
- Thread starter mackerel
- Start date
OP
- Joined
- Mar 7, 2008
- Thread Starter
- #62
Just for fun, about 2 months ago I made an estimate on CB R23 score for the top Arrow Lake based on leaked clocks and some assumptions about core performance. My guess was 43450. This was posted on LTT forum as a status update on my profile. I can't link directly to that post but it can be searched or expand attached image for screenshot.
Why bring this up now? There's been a recent leak of scores as follows:
42286 - this result in on hwbot and it is ambiguous if it is really stock as Videocardz claims. The submitter has what looks like hardware adjustment tools in the background of the screenshot, and if benchmate is accurate the CPU is running hot enough there may be throttling going on.
45563 - result at 250W power limit which is the Intel recommended performance profile, so this is more likely "stock" if this result is correct. Temps under 80C so shouldn't be throttling.
46289 - unlimited power limit, peaking at reported 370W, temps up to 90C.
Two more days until we get widespread independent testing.
videocardz.com
Why bring this up now? There's been a recent leak of scores as follows:
42286 - this result in on hwbot and it is ambiguous if it is really stock as Videocardz claims. The submitter has what looks like hardware adjustment tools in the background of the screenshot, and if benchmate is accurate the CPU is running hot enough there may be throttling going on.
45563 - result at 250W power limit which is the Intel recommended performance profile, so this is more likely "stock" if this result is correct. Temps under 80C so shouldn't be throttling.
46289 - unlimited power limit, peaking at reported 370W, temps up to 90C.
Two more days until we get widespread independent testing.
Unlocked Intel Core Ultra 9 285K approaches 370W power draw during Cinebench test - VideoCardz.com
Intel Core Ultra 9 285K put through extreme testing According to new screenshots shared on social media, the Core Ultra 9 285K will consume between 360-370W when fully unlocked. This power draw can be observed during a multi-core Cinebench R23 test, with all P-Cores locked at 5.6 GHz and all...
videocardz.com
- Joined
- Nov 14, 2002
- Location
- Nashville
yikes, 370w!
At that kind of power I could run 2x 7800x3d's and still produce less heat.
At that kind of power I could run 2x 7800x3d's and still produce less heat.
![]-[itman](/forums/data/avatars/m/9/9309.jpg?1650325391){kind=avatar}
Just for fun, about 2 months ago I made an estimate on CB R23 score for the top Arrow Lake based on leaked clocks and some assumptions about core performance. My guess was 43450. This was posted on LTT forum as a status update on my profile. I can't link directly to that post but it can be searched or expand attached image for screenshot.
Why bring this up now? There's been a recent leak of scores as follows:
42286 - this result in on hwbot and it is ambiguous if it is really stock as Videocardz claims. The submitter has what looks like hardware adjustment tools in the background of the screenshot, and if benchmate is accurate the CPU is running hot enough there may be throttling going on.
45563 - result at 250W power limit which is the Intel recommended performance profile, so this is more likely "stock" if this result is correct. Temps under 80C so shouldn't be throttling.
46289 - unlimited power limit, peaking at reported 370W, temps up to 90C.
Two more days until we get widespread independent testing.
Unlocked Intel Core Ultra 9 285K approaches 370W power draw during Cinebench test - VideoCardz.com
Intel Core Ultra 9 285K put through extreme testing According to new screenshots shared on social media, the Core Ultra 9 285K will consume between 360-370W when fully unlocked. This power draw can be observed during a multi-core Cinebench R23 test, with all P-Cores locked at 5.6 GHz and all...
View attachment 368939
The 45563 score isn’t stock, it’s manually overlocked.
Even 3x 7800x3d
... and 9800x3d will be released soonI wouldn't touch Intel since the first Ryzen release if not for tests and reviews.
- Joined
- Nov 14, 2002
- Location
- Nashville
Man... I love so many aspects of Intel. VPro is awesome, quicksync is useful, their iGPUs are honestly very good for non gaming. I really love the T series CPU's for moderate low power applications.Even 3x 7800x3d
I wouldn't touch Intel since the first Ryzen release if not for tests and reviews.
I just hope that within reasonable power and thermal constraints they can pull some good value for that middle market. If it turns out that at 35w or 65w their CPU's are still pulling 90% of the performance as these suckers juiced to the moon, then they could still pull out a win.
OP
- Joined
- Mar 7, 2008
- Thread Starter
- #67
That's with no power limit and we don't know if that was a one off peak or how close the average of the run was.
Looking again, the linked poster did say "with bios optimisations" but it looks like they kept Intel recommended power limit. The same poster used the 42286 example as unoptimised but that is way above recommended power limit, so can't be stock either. Following the links further didn't shed any more light on the original source of what those optimisations might be.
The nearest competitor is the 9950X which has a stock power limit of 230W, so Intel recommended power limit of 250W is not so much different. We'll have to see how they perform at those limits and/or at same power. 2 days...
9950x stock power limit is 200 W. I think the 42286 score is probably close to stock. I think the 285k will hit a hard wall around 250 W and pushing higher power won't increase performance hardly at all unless you do some manual tuning. I think ARL will perform better in CB24 from the scores that have leaked.
In other news, word on the street (twitter) is that ARL's gaming performance is a decent regression from RPL. They are saying it is comparable to ADL (12th gen) in gaming performance. Of course, take it with a healthy dose of salt as it's just rumors for now. It would kind of line up with Intel's claims (if you read between the lines) as they claimed ARL was on par with a 9950x in gaming while giving the 285k faster memory and using APO on supported games, which would mean in a more apples to apples comparison, it probably is slower than vanilla Zen 4 for gaming. Only a couple more days for reviews to drop to know for sure.
OP
- Joined
- Mar 7, 2008
- Thread Starter
- #69
PPT is usually set 35% over TDP, which at 170W TDP would put it at 230W. I didn't double check it at the time so there's always a chance AMD throws a curveball. Trying to verify that value now, TechPowerUp says it is 225W, Tom's Hardware says it is 230W.
AMD Ryzen 9 9950X Review - The New Flagship
With 16 cores and 32 threads, the Ryzen 9 9950X, powered by AMD’s Zen 5 architecture, is the fastest desktop processor we've ever tested. In our review, it breezed through application workloads and delivered high FPS rates in gaming. But at $650, it doesn’t come cheap.
CB 2024 is significantly affected by ram performance unlike R23 and earlier which hardly responds to it unless you're after 1% for competitive benchmarking. So if ARL has a better ram subsystem, that could help those scores.
Hallock already stated to expect about 5% behind 7800X3D. If accurate, that would still leave it competitive to Zen 4/5 non-3D. The perf delta is pretty close for a lot of models at the top anyway, so small changes could alter ranking a lot. It is a given that AMD retains the halo position, but the battle could still be won or lost on pricing at the lower tiers. Basically a similar battle AMD GPUs have to face vs Nvidia.
I'd imagine the IPC increase rather than the multi-tile latency helps it there. It may be able to run faster... but, it also may need too. I think the uncore is lower between gens too by several hundred Mhz.
I've seen the same leaks and, from them, a 285K does outperform a 14900K in CB24 using the same RAM/Timings.
I can't tell you what my results are, but I will share that I have spent an inordinate amount of time double checking them.
PPT is usually set 35% over TDP, which at 170W TDP would put it at 230W. I didn't double check it at the time so there's always a chance AMD throws a curveball. Trying to verify that value now, TechPowerUp says it is 225W, Tom's Hardware says it is 230W.
AMD Ryzen 9 9950X Review - The New Flagship
With 16 cores and 32 threads, the Ryzen 9 9950X, powered by AMD’s Zen 5 architecture, is the fastest desktop processor we've ever tested. In our review, it breezed through application workloads and delivered high FPS rates in gaming. But at $650, it doesn’t come cheap.www.techpowerup.com
CB 2024 is significantly affected by ram performance unlike R23 and earlier which hardly responds to it unless you're after 1% for competitive benchmarking. So if ARL has a better ram subsystem, that could help those scores.
Hallock already stated to expect about 5% behind 7800X3D. If accurate, that would still leave it competitive to Zen 4/5 non-3D. The perf delta is pretty close for a lot of models at the top anyway, so small changes could alter ranking a lot. It is a given that AMD retains the halo position, but the battle could still be won or lost on pricing at the lower tiers. Basically a similar battle AMD GPUs have to face vs Nvidia.
It’s 200W. Not sure why they are reporting otherwise. You can check actual power consumption measurements.
ARL has a better memory controller, but the cache and memory latency is way higher, largely due to the move to tiles and the significant reduction in ring frequency. The added latency will hurt gaming as it is a latency sensitive workload and doesn’t really care about bandwidth above “normal” DDR5 speeds.
Hallock also claimed MTL was way more efficient than Zen 4, which was a lie.
OP
- Joined
- Mar 7, 2008
- Thread Starter
- #73
Looking forward to it.
In quick further searching, I did find two places stating a 9950X PPT of 200W, but I also found more saying 230W. Do we have an AMD slide or similar stating this anywhere to settle this? I'll search deeper later.
Without going back to find out exactly what was said at the time, and how it was tested, it is possible to be efficient in some conditions and not others.
I didn't like the guy when he was at AMD. I don't know if he's any different now at Intel. That's (technical) marketing for you.
Less than that now. Is the embargo time under embargo? I think most tech go live times tend to be early afternoon UK time, which might be 8 or 9am Eastern?You can't tell only for two more days
The first rule of review club is: you do not talk about the embargo.
Looking forward to it.
In quick further searching, I did find two places stating a 9950X PPT of 200W, but I also found more saying 230W. Do we have an AMD slide or similar stating this anywhere to settle this? I'll search deeper later.
Without going back to find out exactly what was said at the time, and how it was tested, it is possible to be efficient in some conditions and not others.
I didn't like the guy when he was at AMD. I don't know if he's any different now at Intel. That's (technical) marketing for you.
Less than that now. Is the embargo time under embargo? I think most tech go live times tend to be early afternoon UK time, which might be 8 or 9am Eastern?
The first rule of review club is: you do not talk about the embargo.
You can check power measurements (it won't go over 200 W at stock) or Ryzen Master which lists the stock PPT at 200 W.
For Hallock's lies, it was during the Intel MTL preview event. They showed a bunch of benchmarks where MTL had huge leads over Zen 4 and Hallock promised they were using approximately the same amount of power. Turns out MTL was using like 50% or more power during their testing.
Here is the video. You can compare with independent testing here.
Edit: ARL should be much improved over RPL for power consumption, though, and do much better for mobile and lower power levels in general.
Last edited:
OP
- Joined
- Mar 7, 2008
- Thread Starter
- #75
Thanks. I'll take mobo default as an indicator, as AMD have been better than Intel at keeping settings in line. I wonder if those other sites listing 230W did what I did then. I don't recall if there were other exceptions but PPT has been 35% over TDP as far as I can remember.
Apart from that, we wait for ARL results and a bit longer for whatever 9000X3D in November.
Yes, the 9950x is the lone exception to the PPT/TDP ratio that I am aware of so most likely those outlets just assumed it was the same as well.
Less than that now. Is the embargo time under embargo? I think most tech go live times tend to be early afternoon UK time, which might be 8 or 9am Eastern?
The first rule of review club is: you do not talk about the embargo.
I had to sign some stuff until Oct 24th but without any hours. However, I have only mobo and CUDIMM samples. Most will be delayed anyway. Because of delays (a typhoon in Taiwan and other things), I'm still deep in the X870/E mobos.
I talk about things I can that are already listed by manufacturers on their product pages. We can't talk about results and product details under the signed NDA, but we can share some thoughts in general.
A user at overclockers.uk got his CPU early and has posted a couple of results. The results are really bad, seems like an early BIOS issue (hopefully).
For reference, a 7800x3d scores around 260 fps, if not higher.
MT is about 10% faster than a 13700k at stock, but this is drawing close to 10% more power than a 13700k at stock. The single core score is significantly slower than a 13700k (~2125 pts). Could be that the single core run is bouncing around cores or stuck on an e-core which is causing the lower score.
Additionally, there are leaks of significant stability issues (games crashing, BSOD). Supposedly it is firmware related and not a hardware issue.
For reference, a 7800x3d scores around 260 fps, if not higher.
MT is about 10% faster than a 13700k at stock, but this is drawing close to 10% more power than a 13700k at stock. The single core score is significantly slower than a 13700k (~2125 pts). Could be that the single core run is bouncing around cores or stuck on an e-core which is causing the lower score.
Additionally, there are leaks of significant stability issues (games crashing, BSOD). Supposedly it is firmware related and not a hardware issue.
I know the hour, but not sure if i can say anything, lol. We wont see anything here for Z890 on launch day, but my mobo review will be up at Tom's.
... worth noting, this is 1080p with everything on low so it's really leaning on the CPU. I'd imagine when more GPU gets involved (higher settings), higher res, it won't look as bad (just a guess, I haven't tested on high res).
I've been good across the 3 boards I've played with (for what little that is worth).
I think when we see a wider swath of games the difference won't be as much, but I'd imagine there will be outliers like this...
... worth noting, this is 1080p with everything on low so it's really leaning on the CPU. I'd imagine when more GPU gets involved (higher settings), higher res, it won't look as bad (just a guess, I haven't tested on high res).
OP
- Joined
- Mar 7, 2008
- Thread Starter
- #80
ARL pricing is up at a major UK retailer. They handled the NV FE sales.
14900KS £630
7950X3D £580
9950X £560
285K £549
7950X £480
7900X3D £440
14900K £440
7800X3D £430
14900KF £420
9900X £400
265K £380
265KF £370
14700K £350
14700KF £330
14700 £315
9700X £310
245K £290
14700F £290
7700X £290
245KF £279
7700 £275
14600K £250
14500 £230
9600X £229
14600KF £206
7600X £200
14400 £181
14400F £170
No price on 7900 and 7600
14900 non-K costs more than K
Gives some context of where they fit in price wise, as that will influence how their performance is judged.
14900KS £630
7950X3D £580
9950X £560
285K £549
7950X £480
7900X3D £440
14900K £440
7800X3D £430
14900KF £420
9900X £400
265K £380
265KF £370
14700K £350
14700KF £330
14700 £315
9700X £310
245K £290
14700F £290
7700X £290
245KF £279
7700 £275
14600K £250
14500 £230
9600X £229
14600KF £206
7600X £200
14400 £181
14400F £170
No price on 7900 and 7600
14900 non-K costs more than K
Gives some context of where they fit in price wise, as that will influence how their performance is judged.
