- Mar 3, 2017
- 1,623
- 5,894
- 136
Well, someone leaked to them (especially MLID) a lot of correct info on Mi300 some 15 months ago. That should count for something.MI200's were a simple, hacky way to win CORAL-2 bids (and they won more than that!).
He's an idiot.
Also an idiot.
That's late, MI300 concept and configs were known as of mid'21 by people who should.Well, someone leaked to them (especially MLID) a lot of correct info on Mi300 some 15 months ago
His leak was in April 2022, with pretty much everything he said ended up being right on the money. Which is kind of rare.That's late, MI300 concept and configs were known as of mid'21 by people who should.
even a moron like him with his head so far up his keester with only his ginormous chin helping him from eating himself bum first like some kind of insect can guess a few things right from time to time.Well, someone leaked to them (especially MLID) a lot of correct info on Mi300 some 15 months ago. That should count for something.
Zen3 was 19% so that's not well above.should be well above what Zen 3 brought, so my guess this time is between 20 and 25%
They're not going up at all.ST clocks maybe up to touch 6Ghz, MT boost probably the same or +100Mhz vs Zen 4.
Seems plausible for staying within 170W socket due to supposed pretty big Cac increase. I'd not be surprised if they end up being some 100mhz lower.They're not going up at all.
if thats really the chase and this "high 2k" 1T CB R23 leak is correct (as well as Keplers +30% ST lead over Raptor Refresh), than the IPC has to increase in the range of 35-40% - this would be a monumental achievement considering it's basically the same manufacturing node as Zen 4. AMD never had this kind of lead over intel befor...Zen3 was 19% so that's not well above.
They're not going up at all.
Which Paul?Good ole Paul dishing out facts. What was it the last time? >30% IPC?
Eyeball PaulWhich Paul?
Any particular source to this claim?They're not going up at all.
if thats really the chase and this "high 2k" 1T CB R23 leak is correct (as well as Keplers +30% ST lead over Raptor Refresh), than the IPC has to increase in the range of 35-40% - this would be a monumental achievement considering it's basically the same manufacturing node as Zen 4. AMD never had this kind of lead over intel befor...
Good ole Paul dishing out facts. What was it the last time? >30% IPC?
Large IPC jumps are not free.I mean, i would not expect any massive jumps given the same node thing, but at least 100~300 MHz as result of the node being more mature and whatnot...
The Acer CEO calls them Astonishing Miracle Delivery for a reason. 😉Large IPC jumps are not free.
So this means its just your own conjecture, but there is otherwise no source to support that?Large IPC jumps are not free.
That's doubtful.
If you look back at my projections from months ago that happens to be in the same ballbark as leaked now, the difference is minuscule percentages (1.3%):
Page 45 - Discussion - Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)
Page 45 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.forums.anandtech.com
The difference can only be between clockspeed and IPC and the resulting multiplication of both should remain the same. This is simplified rule of thumb, actual calculation probably only off by a little anyway.
Since my setting was 8% clock at best and 15% IPC : 1.08x1.15=1.242 you can play around within this range.
For example if IPC were to be 25% then clock will have to be 1.242/1.25=0.9936 so that's obviously very unlikely to happen with marketing in mind.
If clockspeed is only raised by 0.1 GHz (from Ryzen 9 7950X 4.5/5.7 GHz) or a change of 1.7% then IPC would 22.1% at best.
It seems to me 15-20% is likely what's going to be marketed.
Btw. I didn't post it before but for 8c I had 8c 24185 quite a difference from the leaked 23k so clockspeed must be lower for the lower level offerings.
Talk to people and you'll find the source quick enough.So this means its just your own conjecture, but there is otherwise no source to support that?
That’s not how it works. There’s a super secret handshake club to enter the AMD fraternity where you are then allowed to view this benchmark data showing a >=30% IPC gain.Talk to people and you'll find the source quick enough.
I am talking to you, does anything stop you from revealing the source?Talk to people and you'll find the source quick enough.
Not quite.There’s a super secret handshake club to enter the AMD fraternity where you are then allowed to view this benchmark data showing a >=30% IPC gain.
Yea.Everybody has seen it and talks about it but nobody will tell you where to find it.
I mean, yea.does anything stop you from revealing the source?