# R7 hardware results

## 2026-09-16: successful retest and release promotion

The owner requested the first unified TF030/68000 core that had booted and
suspected the earlier disk/performance issue was the CF card. The exact R7
BIN identified below was restored as `MINIMIG1.BIN` at 18:40 EDT, verified
after copying, and the SD card safely unmounted. The previous R3 image was
backed up. Firmware and SD configuration were unchanged.

The owner then reported that it works and explicitly requested promotion to
release. **R7 is now the unified TF030/68000 release**, tag
`r7-unified-20260916`; [download and provenance](../releases/r7-unified/README.txt).
The BIN is unchanged from the original R7 handoff and earlier shared boot.

No new throughput figure, test duration or card configuration was supplied.
The successful retest supersedes the earlier pending-release status. The
CF-card explanation remains the owner's suspected cause, and the earlier
measurements below remain historical observations rather than a new result.
The new report does not separately establish TF536 or PiStorm hardware support.

## Image under test

- Version: **PI FMG260915R760Mhz**.
- Source commit: `dc45bf45d1992da9ec3eb57e0f43b1d46df13390`.
- BIN SHA-256: `9a522b53ceba70566123a1886a08010fbecf0c2a042faa37727f3436fa7df67d`.
- Copied as `MINIMIG1.BIN` on 2026-09-15, with checksum verification and safe
  unmount. The previous R5 image was backed up.

The source/BIN package remains the original handoff snapshot. This log records
subsequent hardware feedback without changing that image or its identity.

## 2026-09-15: TF534 Normal boot

User report: “TF 534 is booting in NORMAL mode!”

This confirms a successful Normal-mode boot on the user's TF534 board with R7.
The same user previously reported that both original-configuration TF534 and
TF536 cards failed to boot R5 in Normal. The R7 report does not yet establish
repeat-boot reliability or application or memory stability. The initial report
covered Normal only; the subsequent Turbo report is recorded below.

| Configuration | Reported R7 hardware result |
| --- | --- |
| TF534, Normal | Boots; broader testing pending |
| TF534, Turbo | Boots/operates; earlier storage measurements below, followed by the successful September 16 retest |
| TF536, Normal / Turbo | Not yet reported |
| 68SEC000 | Boots; clock mode not yet specified |

The user subsequently reported: “btw 68SEC000 also booted”. This confirms a
68SEC000 boot with the same R7 candidate. The clock mode and application or
memory-test results were not specified in that report.

## 2026-09-15: TF030 Turbo operation

User report: “Fantastic, even TF030 works in TURBO!”

This adds user-reported TF030 Turbo operation with the same R7 image. The user
subsequently confirmed this is the same TF534. Workloads and duration were not
specified; this report does not yet establish extended game or memory stability.

## 2026-09-15: TF storage throughput problem

After reporting Turbo operation, the user reported that the SD card on the TF
card was extremely slow, around **5 KB/s**. A subsequent Normal-mode comparison
gave **364 KB/s**. The user expects closer to **2,000 KB/s**. These are reported
figures. The user confirmed **SysInfo on the same TF534** for the comparison.
The exact adapter, firmware and SysInfo version remain to be identified. The
expected figure is a user target, not an established R7 result.

Normal is approximately 73 times faster than the reported Turbo result, but
also falls short of the stated target. Storage performance is therefore an
open issue; successful boot does not qualify the TF storage path. Investigation
is comparing socket-clock dependencies, interrupt/register transactions and
the latest AmiCube bridge. No replacement image has been produced for this issue.
The [focused source and simulation investigation](r7-tf-storage.txt) did not
reproduce the suspected early write-retirement fault with the tested TF534 RTL;
the throughput cause remains unresolved.

Next checks: TF applications/games, memory and repeated resets in Normal and
Turbo; remaining card-specific boot coverage; SEC000 Alien Breed in Normal and
Turbo. Record fresh-boot versus mode-switch behavior as testing proceeds.

## 2026-09-15: cache bypass does not change storage throughput

User report: “Cache-off: 7MHz 364kb, 60MHz 5KB. Not a factor”.

After identifying the existing Video/Scanlines Cache Off control, the user
repeated SysInfo on the TF534. Both values match the earlier cache-enabled
reports. This comparison argues against cache-hit responses as the cause of
the Turbo throughput collapse. Cache capacity changes are not the next fix.
The requested setup kept the TF on its onboard 50 MHz oscillator; the user
did not separately restate the jumper setting in this result.

A source-level SPI test independently verified that command A2 bypasses both
ROM and slow-RAM hits while Turbo stays active, and an ordinary Amiga reset
retains that setting. Fourteen reads passed; disconnecting the bypass wire
produced the expected policy failure. Local evidence is in
`build/r7-cache-toggle-probe` in the main checkout. R7's image is unchanged.
